Machinery Malfuntion Diagnosis and Corretion

Machinery Malfunction
Diagnosis and Correction

Vibration Analysis and Troubleshooting
for the Process Industries
Robert C. Eisenmann, Sr., P.E.

President
MACHINERY
DIAGNOSTICS,
Inc.
Minden,
Nevada
Principal Mechanical
Engineer
- Sulzer Hickham
Inc.
- La Porte,
Texas
and
Robert C. Eisenmann, Jr.

Manager of Rotating Equipment HAHN & CLAY Houston, Texas
Rotating Equipment Technical Authority - BP Products North America - Houston, Texas
PTR
Cliffs,
New by:
Jersey
07632
The original
HardPrentice
Copy formatHall,
of this Englewood
book was previously
published
Pearson
Education, Inc.
Copyright Assigned to Robert C. Eisenmann, Sr. by Hewlett-Packard effective June 6, 2005.
Library of Congress Cataloging-in-Publication Data
Author: This space is reserved for Library of Congress

Cataloging-in-Publication Data, which PTR will insert.
Original Library of Congress Cataloging in Publication Data

Eisenmann, Robert C. 1943Machinery malfunction diagnosis and correction: vibration analysis and troubleshooting for
the process industries / Robert C. Eisenmann, Sr., and Robert C. Eisenmann, Jr.
cm -- (Hewlett- Packard professional
books)
Acquisitionsp.
editor:
Editorial assistant:
Includes
index.
Cover
design: bibliographical references andCover
design director: Eloise Starkweather-Muller
Copy
Editor:0-13-240946-1
Art production manager: Gail Cocker-Bogusz
ISBN
Manufacturing
Manager:
Alexis
R.
Heydt
Illustrations
by: Robert
C. Eisenmann,Robert
Sr.
1. Machinery -- Monitoring. 2. Machinery
-- Vibration.
I. Eisenmann,
C., 1970Production
team:
Sophie
Papanikolaou,
Jane
Bonnell,
Lisa
Iarkowski,
John
Morgan,
Dit
Mosco,
II. Title. III. Series.
Mary Rottino, Ann Sullivan, Harriet Tellem, and Camille Trentacoste.
TJ153.E355 1997
Proofreaders:
621.8'16-dc21
97-31974
This book was composed with FrameMaker.
CIP
1997 PTR Prentice Hall

Prentice-Hall, Inc.
A Paramount Communications Company
Englewood Cliffs, New Jersey 07632
The publisher offers discounts on this book when ordered in bulk
quantities. For more information, contact Corporate Sales Department,
PTR Prentice Hall, 113 Sylvan Avenue, Englewood Cliffs, NJ 07632.
Phone: 201-592-2863; FAX: 201- 592-2249.
All rights reserved. No part of these templates may be
reproduced, in any form or by any means,
without permission in writing from the publisher.
Printed in the United States of America
10 9 8 7 6 5 4 3 2 1
Author: This space reserved for OCR ISBN to be inserted by PTR.
Prentice-Hall International (UK) Limited, London

Prentice-Hall of Australia Pty. Limited, Sydney
Prentice-Hall Canada Inc., Toronto
Prentice-Hall Hispanoamericana, S.A., Mexico
Prentice-Hall of India Private Limited, New Delhi
Prentice-Hall of Japan, Inc., Tokyo
Simon & Schuster Asia Pte. Ltd., Singapore
Editora Prentice-Hall do Brasil, Ltda., Rio de Janeiro
To Mary Rawson Eisenmann,

Wife and Mother
Who Always Kept The Home Fires Burning
While The Boys Went Off To Play With Their Machines
Table of Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Chapter 1 - Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Machinery Categories 4
Chapter Descriptions 5
Bibliography 8
Chapter 2 - Dynamic Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Malfunction Considerations and Classifications 9
Fundamental Concepts 10
Vector Manipulation 21
Undamped Free Vibration 28
Case History 1: Piping System Dynamic Absorber 31
Free Vibration with Damping 37
Forced Vibration 45
Case History 2: Steam Turbine End Cover Resonance 55
Torsional Vibration 58
Bibliography 66
Chapter 3 - Rotor Mode Shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

Mass and Support Distribution 67
Case History 3: Two Stage Compressor Rotor Weight Distribution 72
Inertia Considerations and Calculations 74
Damping Influence 96
Stiffness Influence 105
Critical Speed Transition 120
Mode Shape Measurement 130
Case History 4: Vertical Generator Mode Shape 137
Analytical Results 142
Case History 5: Eight Stage Compressor Mode Shape Change 143
Bibliography 148
Chapter 4 - Bearings and Supports . . . . . . . . . . . . . . . . . . . . . . . . 149

Fluid Film Radial Journal Bearings 150
Case History 6: Shaft Position In Gas Turbine Elliptical Bearings
Fluid Film Radial Bearing Clearance Measurements 165
Case History 7: Expander Journal Bearing Clearance 174
161
iv
Bearing Supports Measurements and Calculations 179

Case History 8: Measured Steam Turbine Bearing Housing Stiffness 181
Case History 9: Measured Gas Turbine Bearing Housing Stiffness 185
Bearing Housing Damping 187
Fluid Film Thrust Bearings 188
Rolling Element Bearings 193
Before Considering Bearing Redesign 196
Bibliography 198
Chapter 5 - Analytical Rotor Modeling . . . . . . . . . . . . . . . . . . . . 199

Modeling Overview 199
Undamped Critical Speed 201
Case History 10: Mode Shapes for Turbine Generator Set 206
Case History 11: Torsional Analysis of Power Turbine and Pump 208
Stability and Damped Critical Speed Calculations 213
Case History 12: Complex Rotor Damped Analysis 217
Forced Response Calculations 222
Case History 13: Gas Turbine Response Correlation 226
Case History 14: Charge Gas Compressor with Internal Fouling 230
Case History 15: Hybrid Approach To A Vertical Mixer 236
Bibliography 242
Chapter 6 - Transducer Characteristics . . . . . . . . . . . . . . . . . . . 243

Basic Signal Attributes 244
Proximity Displacement Probes 253
Velocity Coils 272
Piezoelectric Accelerometers 278
Pressure Pulsation Transducers 285
Specialized Transducers 288
Aspects of Vibration Severity 294
Bibliography 302
Chapter 7 - Dynamic Signal Characteristics . . . . . . . . . . . . . . . 303

Electronic Filters 303
Time and Orbital Domain 316
Time and Frequency Domain 333
Case History 16: Steam Turbine Exhaust End Bearing Dilemma 343
Signal Summation 347
Case History 17: Opposed Induced Draft Fans 349
Amplitude Modulation 353
Case History 18: Loose and Unbalanced Compressor Wheel 356
Frequency Modulation 359
Case History 19: Gear Box with Excessive Backlash 362
Bibliography 364
vi
Chapter 8 - Data Acquisition and Processing . . . . . . . . . . . . . . . 365

Vibration Transducer Suite 365
Recording Instrumentation 369
Data Processing Instrumentation 379
Data Presentation Formats 383
Bibliography 394
Chapter 9 - Common Malfunctions . . . . . . . . . . . . . . . . . . . . . . . 395

Synchronous Response 395
Mass Unbalance 398
Bent or Bowed Shaft 400
Case History 20: Repetitive Steam Turbine Rotor Bow 402
Eccentricity 406
Case History 21: Seven Element Gear Box Coupling Bore 407
Shaft Preloads 410
Resonant Response 416
Case History 22: Re-Excitation of Compressor Resonance 419
Machinery Stability 422
Case History 23: Warehouse Induced Steam Turbine Instability 429
Case History 24: Pinion Whirl During Coastdown 432
Mechanical Looseness 435
Case History 25: Loose Steam Turbine Bearing 438
Rotor Rubs 440
Cracked Shaft Behavior 443
Case History 26: Syngas Compressor with Cracked Shaft 446
Foundation Considerations 449
Case History 27: Floating Induced Draft Fan 451
Case History 28: Structural Influence of Insufficient Grout 454
Bibliography 458
Chapter 10 - Unique Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459

Parallel Shaft - Two Element Gear Boxes 459
Case History 29: Herringbone Gear Box Tooth Failure 466
Epicyclic Gear Boxes 470
Case History 30: Star Gear Box Subsynchronous Motion 477
Process Fluid Excitations 483
Case History 31: Boiler Feed Water Pump Splitter Vane Failures 496
Case History 32: Hydro Turbine Draft Tube Vortex 499
Electrical Excitations 507
Case History 33: Motor With Unsupported Stator Midspan 515
Case History 34: Torsional Excitation From Synchronous Motor 519
Reciprocating Machines 522
Case History 35: Hyper Compressor Plunger Failures 526
Bibliography 534
vii
Chapter 11 - Rotor Balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535

Before Balancing 536
Standardized Measurements and Conventions 539
Combined Balancing Techniques 545
Linearity Requirements 547
Case History 36: Complex Rotor Non-Linearities 548
Single Plane Balance 552
Case History 37: Forced Draft Fan Field Balance 560
Two Plane Balance 564
Case History 38: Five Bearing, 120 MW Turbine Generator Set 575
Weight Sequence Variation 586
Case History 39: Three Bearing Turbine Generator at 3,600 RPM 588
Case History 40: Balancing A 36,330 RPM Pinion Assembly 597
Three Plane Balance 606
Static-Couple Corrections 616
Multiple Speed Calculations 618
Response Prediction 619
Trim Calculations 622
Balancing Force Calculations 623
Balance Weight Splitting 626
Weight Removal 628
Shop Balancing 629
Bibliography 636
Chapter 12 - Machinery Alignment . . . . . . . . . . . . . . . . . . . . . . . . 637

Pre-Alignment Considerations 638
Optical Position Alignment 649
Case History 41: Hyper Compressor Position Alignment 654
Laser Position Alignment 658
Optical and Laser Bore Alignment 660
Wire Bore Alignment 663
Case History 42: Hyper Compressor Bore Alignment 667
Shaft Alignment Concepts 669
Rim and Face Shaft Alignment 673
Reverse Indicator Shaft Alignment 681
Optics, Lasers, and Wires for Shaft Alignment 691
Hot Alignment Techniques 692
Case History 43: Motor to Hot Process Pump Alignment 697
Bibliography 702
viii
Chapter 13 - Applied Condition Monitoring . . . . . . . . . . . . . . . . 703

Maintenance Philosophies 703
Condition Monitoring 705
Machinery Performance 706
Vibration Response Data 708
Bearing Temperature Data 711
Data Trending 712
Case History 44: Four Pole Induction Motor Bearing Failure 714
Case History 45: Cracked Gas Compressor Intermittent Instability 718
Case History 46: High Stage Compressor Loose Thrust Collar 721
Pre-Startup Inspection and Testing 724
Startup Inspection and Testing 732
Case History 47: Turbine Solo Operation with Tapered Journal 735
Case History 48: Coupled Turbine Generator Startup 736
Case History 49: Heat Soak and Load Stabilization 739
Bibliography 742
Chapter 14 - Machinery Diagnostic Methodology . . . . . . . . . . . 743

Diagnostic Objectives 744
Mechanical Inspection 744
Test Plan Development 745
Data Acquisition and Processing 746
Data Interpretation 749
Conclusions and Recommendations 750
Corrective Action Plan 750
Case History 50: Steam Turbine Electrostatic Voltage Discharge
Case History 51: Barrel Compressor Fluidic Excitation 758
Case History 52: High Speed Pinion Instability 766
Conclusions on Diagnostic Methodology 770
Bibliography 770
751
Chapter 15 - Closing Thoughts and Comments . . . . . . . . . . . . . 771

Economic Reality 772
Corporate Considerations 773
Presentation of Results 778
Silver Bullets 780
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 781
A Machinery Diagnostic Glossary
B Physical Properties 795
C Conversion Factors 797
D Index 801
781
Preface
hen my son graduated from Texas

A&M University, he was understandably eager to start working, and begin earning a livable salary. He accepted a maintenance engineering position at a large
chemical complex, and embarked upon learning about process machinery. In the
months and years that followed, he and his colleagues had many questions concerning a variety of machinery problems. From my perspective, most of these
problems had been solved twenty or thirty years ago. However, it was clear that
the new engineering graduates were devoting considerable effort attempting to
unravel mysteries that had already been solved.
The obvious question that arises might be stated as: How come the new
engineers cannot refer to the history files instead of reworking these issues? A partial answer to this question is that the equipment files often do not provide any
meaningful historical technical data. Major corporations are reluctant to spend
money for documentation of engineering events and achievements. Unless the
young engineers can find someone with previous experience with a specific malfunction, they are often destined to rework the entire scenario.
Although numerous volumes have been published on machinery malfunctions, there are very few technical references that address the reality of solving
field machinery problems. This general lack of usable and easily accessible information was a primary force in the development of this text. The other significant
driving force behind this book was the desire to coalesce over thirty-three years
of experience and numerous technical notes into some type of structured order
that my son, and others could use for solving machinery problems.
This is a book about the application of engineering principles towards the
diagnosis and correction of machinery malfunctions. The machinery under discussion operates within the heavy process industries such as oil refineries, chemxi
xii
ical plants, power plants, and paper mills. This machinery consists of steam, gas
and hydro turbines, motors, expanders, pumps, compressors, and generators,
plus various gear box configurations. This mechanical equipment covers a wide
variety of physical characteristics. The transmitted power varies from 50 horsepower, to units in excess of 150,000 horsepower. Rotational speeds range from
128 to more than 60,000 revolutions per minute. There is a corresponding wide
range of operating conditions. Fluid temperatures vary from cryogenic levels of
minus 150F, to values in excess of plus 1,200F. The operating pressures range
from nearly perfect vacuums to levels greater than 40,000 pounds per square
inch. Physically, the moving elements may be only a few feet long, and weigh less
than 100 pounds or they may exceed 200,000 pounds, and cover the length of a
football field. In virtually all cases, these process machines are assembled with
precision fits and tolerances. It is meaningful to note that the vibration severity
criteria for many of these machines are less than the thickness of a human hair.
In some respects, it is amazing that this equipment can operate at all.
When the number of individual mechanical components are considered, and the
potential failure mechanisms are listed, the probabilities for failures are staggering. Considerable credit must be given to the designers, builders, and innovators
of this equipment. They have consistently produced machines that are constantly evolving towards units of improved efficiency, and extended reliability.
The majority of machinery problems that do occur fall into what I call the
ABC category. These common problems are generally related to Alignment, Balance, and incorrect Clearances (typically on bearings). Due to the continual
appearance of these malfunctions, an entire chapter within this text has been
devoted to each of these subjects. Machines also exhibit other types of failures,
and a sampling of common plus unique problems are described within this book.
Some people might view this document as a textbook. Others might consider this to be a reference manual, and still other individuals might use this
book for troubleshooting. It has also been suggested that this book be categorized
as a how to do it manual. Since 52 detailed case histories are combined with
numerous sample calculations and examples, each of these descriptions are accurate and applicable. In the overview, the contents of this book cover a variety of
machinery malfunctions, and it engages the multiple engineering disciplines
that are required to solve real world problems. Regardless of the perception, or
the final application, this is a book about the mechanics, measurements, calculations, and diagnosis of machinery malfunctions. I sincerely hope that this text
will provide some meaningful help for students, for new graduates entering this
field, as well as provide a usable reference for seasoned professionals.
Finally, I would like to extend my deepest personal thanks to John Jensen
of Hewlett Packard for the inspiration, encouragement, and opportunity to write
this book. I am further indebted to John for his detailed and thorough review of
much of the enclosed material. I would also like to thank Ron Bosmans, Dana
Salamone, and Pamela Puckett for their constructive comments and corrections.
Robert C. Eisenmann, Sr., P.E.
October 1997
Introduction
ior
al
B
K ehav
ysi
c
tio
nta
me
tru e
x
E
Ph
Ins
achinery development has been synonymous with technological progress. This growth has resulted in an evolutionary
trend in industrial equipment that moves towards increased complexity, higher
speeds, and greater sophistication. The water wheel has evolved into the hydroelectric plant, the rudimentary steam engine has grown into the gas turbine, and
coarse mechanical devices have been replaced by elegant electronic circuits.
Throughout this evolution in technology, new industries and vocations have
developed. In recent decades, the Machinery Diagnostician has appeared within
most maintenance engineering organizations. These individuals generally possess an extensive knowledge of the machinery construction. They understand
repair procedures, and they have a working knowledge of the peripheral equipment. This includes familiarity with the lube and seal oil system, the processing
scheme, and the machine controls. Diagnosticians are generally knowledgeable
of the machinery monitoring or surveillance instrumentation that covers everything from transducers to the data logging computers. Furthermore, when a
problem does appear on a piece of equipment, it generally falls under the jurisdiction of the machinery diagnostician to resolve the difficulty, and recommend
an appropriate course of corrective action. This requirement imposes another set
of demands. That is, these individuals must be familiar with problem solving
techniques and proven methodology for correcting the machinery malfunction.
Clearly, the diagnostician must be qualified in many technical disciplines.
As depicted in the adjacent diagram, the basic areas of expertise include knowledge of machinery, knowledge of physical behavior, plus knowledge of instrumentation. The machinery background must be
thorough, and it must allow the diagnostician to focus upon
realistic failure mechanisms rather than esoteric theories.
The category of physical behavior embraces technical fields
w l e dg
such as: statics, dynamics, kinematics, mechanics of
no
materials, fluid dynamics, heat transfer, mathematics,
and rotordynamics. Knowledge in these areas must be
fully integrated with the instrumentation aspects of
pe
c
ri e n
the electronic measurements required to document
Machinery
and understand the machinery motion.
Chapter-1
Competence in these three areas is only achieved by a combination of

knowledge and field experience. Acquiring knowledge often begins with specific
technical training. For instance, all academic institutions provide the mathematics and physics necessary to grasp many physical principles. A few universities
provide an introduction to the world of analytical rotordynamics. Unfortunately,
academia is often burdened by the necessity to obtain research grants, and generate complex general solutions for publication. Certainly the college level contributions to this field are significant, and the global solutions are impressive.
However, the working machinery diagnostician often cannot use generalized concepts for solving everyday problems. To state it another way, integral calculus is
absolutely necessary for success in the classroom, but it is reasonably useless for
most activities performed on the compressor deck.
Within the industrial community, a variety of training programs are available. Instrumentation vendors provide courses on the application and operation
of their particular devices. Similarly, machinery vendors and component suppliers have various courses for their clientele. Although these training courses are
oriented towards solutions of field problems, they typically display shortcomings
in three areas. First, the industrial courses are limited in scope to three or four
days of training. This time frame is acceptable for simple topics, but it is inadequate for addressing complex material. Second, industrial training courses are
restricted to the instruments or devices sold by the vendor providing the training. Although this approach is expected by the attendees, it does limit the depth
and effectiveness of the training. The third problem with vendor training resides
in the backgrounds of the training specialists. Although these people are usually
well qualified to represent the products of the vendor, they often lack an understanding of the realities within an operating plant. Clearly, the smooth presentation of fifty computer generated slides has no relationship to the crucial decisions
that have to be made at 2:00 AM regarding a shaking machine.
Another disturbing trend seems to permeate the specialized field of vibration analysis. Within this technical area, there have been long-term efforts by
some vendors to train people to solve problems based entirely on simplistic vibratory symptoms. This is extraordinarily dangerous, and the senior author has
encountered numerous instances of people reaching the wrong conclusions based
upon this approach. Many problems display similar vibratory symptoms, and
additional information is usually required to sort out the differences. In all cases,
the measurements must be supplemented with the logical application of physical
laws. In addition, the machinery construction and operation must be examined
and understood in order to develop an accurate assessment of the malfunction.
Very few professional organizations provide a comprehensive and integrated approach targeted to the topic of machinery diagnosis. The text contained
herein attempts to provide a pragmatic and objective overview of machinery malfunction analysis. The three fundamental areas of physical behavior, machinery,
and instrumentation knowledge are integrated throughout this book. The structure of this text is directed towards developing a basic understanding of fundamental principles. This includes the applicability of those principles towards
machinery, plus the necessary instrumentation and computational systems to
describe and understand the actual behavior of the mechanical equipment.

It should be recognized that acquiring basic knowledge does not guarantee
that the diagnostician will be qualified to engage and solve machinery problems.
As previously stated, experience is mandatory to become proficient in this field.
Although the preliminary knowledge may be difficult to obtain, the experience
portion may be even harder to acquire. This is particularly true for the individual that works in an operating complex that contains a limited assortment of
mechanical equipment. For this diagnostician, the ability to develop a wellrounded background may be hampered due to an absence of mixed machinery
types, and associated problems. References such as the excellent series of books
by Heinz Bloch1 provide detailed machinery descriptions, procedures, and guidelines. If the diagnostician is not familiar with a particular machine, this is the
one available source that will probably answer most mechanical questions.
In a further attempt to address the experience issue, this text was prepared
with 52 field case histories interspersed throughout the chapters. These case
studies are presented with substantial details and explanations. The logical
steps of working through each particular problem are reviewed, and the encountered errors as well as the final solutions are presented. It is the authors hope
that these field examples on major process machinery will provide additional
insight, and enhance the experience level of the machinery diagnostician.
The equipment discussed in this text resides within process industries such
as oil refining, pipeline, chemical processing, power generation, plus pulp and
paper. The specific machines discussed include pumps, blowers, compressors, and
generators that vary from slow reciprocating units to high speed centrifugal
machines. The prime movers appear in various configurations from induction
motors, to cryogenic and hot gas expanders, hydro-turbines, multistage steam
turbines, and large industrial gas turbines. In some cases the driver is directly
coupled to the driven equipment, and in other trains an intermediate gear box is
included. Some of the discussed machinery was installed decades ago, and other
mechanical equipment was examined during initial field commissioning.
It is an objective of this text to assist in understanding, and to demonstrate
practical solutions to real world machinery problems. This book is not designed
to be mathematically rigorous, but the presented mathematics is considered to
be accurate. In all cases, the original sources of the mathematical derivations are
identified. This will allow the reader to reference back to the original technical
work for additional information. Significant equations in this text are numerically identified, and highlighted with an outline box such as equation (2-1).
Developmental and supportive equations are sequentially numbered in each
chapter. In addition, intermediate results plus numeric sample calculations are
also presented. These examples are not assigned equation numbers. In essence,
this book is structured to supplement a formal training presentation, and to provide an ongoing reference.
1 Heinz P. Bloch, Practical Machinery Management for Process Plants, Vol. 1 to 4 (Houston, TX:
Gulf Publishing Company, 1982-1989).
Chapter-1
MACHINERY CATEGORIES
It is organizationally advantageous to divide process machinery into three
categories. Typically, these individual machinery categories are administered
under a singular condition monitoring program since they share a common technology. However, the allocation of resources among the three segments varies in
direct proportion to the process criticality of the mechanical equipment.
The first segment covers the large machinery within an operating plant.
These main equipment trains are generally critical to the process. In most
instances the plant cannot function without these machines. For example, the
charge gas compressor in an ethylene plant, or a syngas compressor in an ammonia plant fall into this category. This equipment typically ranges between 5,000
and 50,000 horsepower. Operating speeds vary from 200 to 60,000 RPM, and
fluid film bearings are normally employed. Most of the machinery problems presented within this text reside within this critical category.
Machines of this class are typically equipped with permanently installed
proximity probe transducer systems for vibration and position measurements,
plus bearing temperature pickups, and specialized transducers such as torque
sensors. Historically, the field transducers are hard wired to continuous monitoring systems that incorporate automated trip features for machinery protection.
These monitoring systems are also connected to process and/or dedicated computer systems for acquisition of static and dynamic data at predetermined sample rates. These data acquisition computer systems provide detailed information
concerning the mechanical condition of the machinery.
The second major group of machines are categorized as essential units.
They are physically smaller than the critical units, they normally have lower
horsepower ratings, and they are usually installed with full backup or spare
units. Machines within this category include trains such as product pumps,
boiler feed water pumps, cooling water pumps, etc. Individual units in this category may not be critical to the process but it is often necessary to keep one out
of two, or perhaps two out of three units running at all times. It should be recognized that a particular service may be considered as essential equipment when a
fully functional main and spare unit are in place. However, if one unit fails, plant
operation then depends upon the reliability of the remaining train. In this manner, an essential train may be rapidly upgraded to the status of a critical unit.
These essential machinery trains are usually instrumented in a manner
similar to the critical units previously discussed. Shaft sensing proximity probe
systems, and thermocouples are hard wired to monitoring systems. These monitoring systems may be integrated with computerized trending systems. Due to
the similarity of construction and installation of the critical and the essential
machines, the text contained herein is directly applicable to essential units.
The third group of machines are referred to as general purpose equipment. These units are physically smaller, and they generally contain rolling element bearings. These machines are often installed with full backups, or they are
single units that are non-critical to the process. Machines within this category
have minimal vibration or temperature measuring instrumentation perma-
Chapter Descriptions
nently installed. This equipment is often monitored with portable data loggers,
and the information tracked with dedicated personal computer systems. In many
instances, small machines are not subjected to detailed analytical or diagnostic
procedures. An in-depth analysis might cost more than the original purchase
price of the equipment. Although there are not many direct references to small
machinery within this book, the techniques and physical principles discussed for
large machines are fully appropriate for these smaller units.
The technology necessary to understand the behavior of process machinery
has been evolving for many years. For example, dedicated machinery monitoring
systems are being replaced by direct interfaces into Distributed Control Systems
(DCS) for trending of general information. Detailed dynamic data is simultaneously acquired in a separate diagnostic computer system. This improvement in
data trending and resolution allows a better assessment of machinery malfunctions. In addition, numerous developments in the areas of rotor dynamics, aerodynamics, blade design, cascade mechanics, metallurgy, fabrication, testing, plus
optimizing bearing and support designs have all combined to provide a wealth of
knowledge. Understanding these individual topics and the interrelationship
between design parameters, mechanical construction, vibratory behavior, position between elements, and the array of electronic measurements and data processing can be an intimidating endeavor.
In support of this complex requirement for knowledge plus experience, this
book has been prepared. To provide continuity through the chapters, various facets of several basic types of industrial machines are examined. It is understood
that one text cannot fully cover all of the material requested by all of the readers.
However, it is anticipated that the information presented within this text will
provide a strong foundation of technical information, plus a source for future reference. The specific topics covered in this book are summarized as follows.
CHAPTER DESCRIPTIONS
The following chapter 2 on dynamic motion begins with a general classification of machinery vibration problems. A review of the fundamental concepts
provides a foundation that extends into a description of a simple undamped
mechanical system. The addition of damping, plus the influence of forced vibration are discussed. Although the majority of the emphasis is placed upon lateral
motion, the parallel environment of torsional vibration is introduced. Finally, the
theoretical concepts are correlated with actual measured machinery vibratory
characteristics for lateral and torsional behavior.
Rotor mode shapes are discussed in chapter 3. This topic begins with a
review of static deflection, followed by the influence of rotor mass, and the distribution of mass and supports. Various aspects of inertia of mechanical systems
are discussed, and critical distinctions are identified. Next, system damping, and
effective support stiffness are discussed, and their influence upon the deflected
mode shapes are demonstrated. The physical transition of a rotor across a critical speed, or balance resonance region is thoroughly explained. These basic con-
Chapter-1
cepts are then extended into measured and calculated rotor mode shapes. In
addition, the construction of interference maps are introduced, and a variety of
illustrations are used to assist in a visualization of these important concepts.
Chapter 4 addresses machinery bearings and supports in rotating systems. This includes an introduction to oil film bearing characteristics, and some
computational techniques. This is followed by proven techniques for determination of radial fluid film bearing clearances, plus the measurement of bearing
housing coefficients. Fluid film thrust bearings are also discussed, and the characteristics of rolling element bearings are reviewed. Appropriate case histories
are included within this chapter to assist in explanation of the main concepts.
Analytical rotor modeling is introduced in chapter 5. This is a continuation of the machinery behavior concepts initiated in the previous chapters. These
concepts are applied to the development of an undamped critical speed analysis
for lateral and torsional behavior. This is followed by the inclusion of damping to
yield the damped response, plus a stability analysis of the rotating system. Further refinement of the machinery model allows the addition of dimensional forcing functions to yield a synchronous response analysis. This step provides
quantification and evaluation of the transient and steady state vibration
response characteristics of the machinery. Finally, the validity and applicability
of these analytical techniques are demonstrated by six detailed case histories
distributed throughout the chapter.
Chapter 6 provides a discussion of transducer characteristics for the
common measurement probes. A traditional industrial suite of displacement,
velocity, acceleration, and pressure pulsation probes are reviewed. The construction, calibration, and operating characteristics of each transducer type are subjected to a comprehensive discussion. In addition, the specific advantages and
disadvantages of each standard transducer are summarized. Specialized transducers are also identified, and their general applications are briefly discussed.
Finally, the topic of vibration severity and the establishment of realistic vibration limits is discussed.
Dynamic signal characteristics are presented in chapter 7. This section
addresses the manipulation and examination of dynamic vibration signals with a
full range of electronic filters. In addition, an explanation of combining time
domain signals into orbits, and the interrelationship between the time and frequency domain characteristics are examined. Finally, common signal combinations such as signal summation, amplitude modulation, and frequency
modulation are discussed. In all cases, appropriate examples are presented.
Chapter 8 covers data acquisition and processing in terms of the
instrumentation systems required for accurate field data acquisition, plus the
processing of the data into useful hard copy formats. Sample forms are included
to facilitate documentation of field measurements. In addition, the functions and
necessary compatibility issues between instruments and transducers are discussed, and operational guidelines are offered. This chapter concludes with an
overview of the most useful machinery data presentation formats.
Based upon the concepts discussed in the previous sections, chapter 9 discusses the origin of many of the common malfunctions experienced by process
Chapter Descriptions
machinery. The topics include synchronous (rotational speed) excitations such as

unbalance, bowed shafts, eccentricity, and resonant responses. The influence of
preloads, machinery stability, mechanical looseness, rubs, and cracked shafts are
discussed. In addition, foundation considerations are reviewed from several perspectives. These general problems are applicable to all rotating machines, and
several case histories are included to illustrate these fundamental mechanisms.
Chapter 10 addresses the unique behavior of different types of machinery. Excitations associated with gear boxes, electrical frequencies, and fluid excitations are included. In addition, the behavioral characteristics of traditional
reciprocating machines, plus hyper compressors are reviewed. Although this
group does not cover all of the potential sources of excitation, it does provide a
useful summary of problems that occur with regularity on many types of
machines. Again, a series of fully descriptive field case histories are distributed
throughout the chapter.
Rotational speed vibration is the dominant motion on most industrial
machines. Chapter 11 is devoted to an in-depth discussion of this synchronous
behavior, and the direct application of these concepts towards rotor balancing.
This chapter begins with the initial thought process prior to balancing, and the
standardized measurements and conventions. The concept of combined balancing techniques are presented, and the machinery linearity requirements are
identified. The development of balancing solutions are thoroughly discussed for
single plane, two plane, and three plane solutions. In addition, static-couple solutions using two plane calculations are presented, and multiple speed calculations
are discussed. The use of response prediction, and trim balance calculations are
reviewed, and several types of supportive calculations are included. Again, field
case histories are provided to demonstrate the applicability of the rotational
speed analysis, and rotor balancing techniques on process machines.
The last portion of chapter 11 deals with shop balancing machines, techniques, and procedures. Although the fundamental concepts are often similar to
field balancing, the shop balancing work is generally performed at low rotative
speeds. This shop balancing discussion includes additional considerations for the
various types of machinery rotors, and common balance specifications.
Machinery alignment persists as one of the leading problems on process
machinery, and this topic is covered in chapter 12. Alignment is discussed in
terms of the fundamental principles for casing position, casing bore, and shaft
alignment. Each type of machinery alignment is discussed, and combined with
explanations of several common types of measurements and calculations. This
includes dial indicator readings, optical alignment, wire alignment, plus laser
alignment, proximity probes, and tooling balls. The applicability of each technique is addressed, and suitable case histories are provided to demonstrate the
field use of various alignment techniques.
The concepts of applied condition monitoring within an operating plant
are discussed in chapter 13 of this text. This chapter was based upon a tutorial
by the senior author to the Texas A&M Turbomachinery Symposium in Dallas,
Texas. The first portion of this chapter describes the logic and evolution of condition monitoring, and the typical parameters involved. These concepts are illus-
Chapter-1
trated with machinery problems detected during normal operation. The second
part of this chapter reviews the turnaround checks and calibrations that should
be performed on the machinery control and protection systems. The third portion
of this chapter covers the application of condition monitoring during a post-overhaul startup of a machinery train. Again, case studies are used to illustrate the
main points of the transient vibratory characteristics.
Chapter 14 address a machinery diagnostic methodology that may be
used for diagnosis of complex mechanical problems. This chapter was based upon
a paper prepared by the senior author for an annual meeting of the Vibration
Institute in New Orleans, Louisiana. This topic discusses the fundamental tools,
successful techniques, and the seven-step process used for evaluation of machinery problems. Again, specific field case histories are included to illustrate some of
the germane points of this topic.
The final chapter 15 is entitled closing thoughts and comments, and it
addresses some of the other obstacles encountered when attempting to solve
machinery problems. This includes candid observations concerning the problems
of dealing with multiple corporate entities, plus the politics encountered within
most operating plants. In many instances, an acceptable solution is fully dependent upon a proper presentation of results that combine economic feasibility
with engineering credibility.
The appendix begins with a machinery diagnostic glossary for the specialized language and terminology associated with this business. For reference
purposes, a list of the physical properties of common metals and fluids, plus a
table of conversion factors are included. The technical papers and books cited
within this text are identified with footnotes, and summarized in a bibliography
at the end of each chapter. In addition, a detailed index is provided in the last
appendix section that includes technical topics, corporate references, and specific
authors referenced throughout this book.
It is the authors hope that the material included within this book will be
beneficial to the machinery diagnostician, and that this text will serve as an
ongoing technical reference. To paraphrase the words of Donald E. Bently (circa
1968), founder and owner of Bently Nevada Corporation we just want to make
the machinery run better To this objective, we have dedicated our professional
careers and this manuscript.
BIBLIOGRAPHY
1. Bloch, Heinz P., Practical Machinery Management for Process Plants, Vol. 1 to 4,
Houston, TX: Gulf Publishing Company, 1982-1989.
Dynamic Motion
any mechanical problems are initially

recognized by a change in machinery vibration amplitudes. In order to understand, and correctly diagnose the vibratory characteristics of rotating machinery,
it is essential for the machinery diagnostician to understand the physics of
dynamic motion. This includes the influence of stiffness and damping on the frequency of an oscillating mass as well as the interrelationship between frequency, displacement, velocity, and acceleration of a body in motion.
MALFUNCTION CONSIDERATIONS AND CLASSIFICATIONS

Before examining the intricacies of dynamic motion, it must be recognized
that many facets of a mechanical problem must be considered to achieve a successful and acceptable diagnosis in a timely manner. For instance, the following
list identifies some of the related considerations for addressing and realistically
solving a machinery vibration problem:
Economic Impact
Machinery Type and Construction
Machinery History Trends Failures
Frequency Distribution
Vibratory Motion Distribution and Direction
Forced or Free Vibration
The economic impact is directly associated with the criticality of the

machinery. A problem on a main process compressor would receive immediate
attention, whereas a seal problem on a fully spared reflux pump would receive a
lower priority. Clearly, the types of machinery, the historical trends, and failure
histories are all important pieces of information. In addition, the frequency of
the vibration, plus the location and direction of the motion are indicators of the
problem type and severity. Traditionally, classifications of forced and free vibration are used to identify the origin of the excitation. This provides considerable
insight into potential corrective actions. For purposes of explanation, the following lists identify some common forced and free vibration mechanisms.
9
10
Chapter-2
Forced Vibration Mechanisms
Free Vibration Mechanisms
Mass Unbalance
Misalignment
Shaft Bow
Gyroscopic
Gear Contact
Rotor Rubs
Electrical Excitations
External Excitations
Oil Whirl
Oil or Steam Whip
Internal Friction
Rotor Resonance
Structural Resonances
Acoustic Resonances
Aerodynamic Excitations
Hydrodynamic Excitations
Forced vibration problems are generally solved by removing or reducing the

exciting or driving force. These problems are typically easier to identify and solve
than free vibration problems. Free vibration mechanisms are self-excited phenomena that are dependent upon the geometry, mass, stiffness, and damping of
the mechanical system. Corrections to free vibration problems may require physical modification of the machinery. As such, these types of problems are often difficult to correct. Success in treating self-excited problems are directly related to
the diagnosticians ability to understand, and apply the appropriate physical
principles. To address these fundamental concepts of dynamic motion, including
free and forced vibration, the following chapter is presented for consideration.
It should be mentioned that much of the equation structure in this chapter
was summarized from the classical textbook by William T. Thomson1, entitled
Mechanical Vibrations. For more information, and detailed equation derivation,
the reader is encouraged to reference this source directly. The same basic equation structure is also described in his newer text entitled Theory of Vibration
with Applications2. Regardless of the vintage, at least one copy of Thomson
should be part of the reference library for every diagnostician.
FUNDAMENTAL CONCEPTS
Initially, consider a simple system consisting of a one mass pendulum as
shown in Fig. 2-1. Assume that the pendulum mass M is a concrete block suspended by a weightless and rigid cable of length L. Further assume that the system operates without frictional forces to dissipate system energy. Intuitively, if
the pendulum is displaced from the vertical equilibrium position, it will oscillate
back and forth under the influence of gravity. The mass will move in the same
path, and will require the same amount of time to return to any specified reference point. Due to the frictionless environment, the amplitude of the motion will
remain constant. The time required for one complete oscillation, or cycle, is
called the Period of the motion. The total number of cycles completed per unit of
1 William Tyrell Thomson, Mechanical Vibrations, 2nd Edition, 9th Printing, (Englewood
Cliffs, New Jersey: Prentice Hall, Inc., 1962), pp.1-75
2 William T. Thomson, Theory of Vibration with Applications, 4th Edition, (Englewood Cliffs,
New Jersey: Prentice Hall, 1993), pp. 1-91.
Fundamental Concepts
11
time is the Frequency of the oscillation. Hence, frequency is simply the reciprocal
of the period as shown in the following expression:
1
Frequency = ------------------Period
(2-1)
The box around this equation identifies this expression as a significant or

important concept. This same identification scheme will be used throughout this
text. Within equation (2-1), period is a time measurement with units of hours,
minutes or seconds. Frequency carries corresponding units such as Cycles per
Hour, Cycles per Minute (CPM), or Cycles per Second (CPS or Hz). Understandably, the oscillatory motion of the pendulum is repetitive, and periodic. As shown
in Marks Handbook3, Fourier proved that periodic functions can be expressed
with circular functions (i.e., a series of sines and cosines) where the frequency
for each term in the equation is a multiple of the fundamental. It is common to
refer to periodic motion as harmonic motion. Although many types of vibratory
motions are harmonic, it should be recognized that harmonic motion must be
periodic, but periodic motion does not necessarily have to be harmonic.
Stationary I-Beam
cos
W=MG
Cable Length - L
sin
Mass
A
C
B
Negative
Fig. 21 Oscillating Pendulum Displaying Simple Harmonic Motion
Max. Neg. Displ.

Zero Velocity
Max. Pos. Accel.
Equilibrium
Zero Displacement
Maximum Velocity
Zero Acceleration
Positive
Max. Pos. Displ.
Zero Velocity
Max. Neg. Accel.
3 Eugene A. Avallone and Theodore Baumeister III, Marks Standard Handbook for Mechanical Engineers, Tenth Edition, (New York: McGraw-Hill, 1996), pp. 2-36.
12
Chapter-2
In a rotating system, such as a centrifugal machine, frequency is normally

expressed as a circular rotational frequency . Since one complete cycle consists
of one revolution, and one revolution is equal to 2 radians, the following conversion applies:
= 2 Frequency = 2 F
(2-2)
Combining (2-1) and (2-2), the rotational frequency may be expressed in

terms of the Period as follows:
2
= ------------------Period
(2-3)
The frequency units for in equation (2-3) are Radians per Second, or Radians per Minute. Again, this is dependent upon the time units selected for the
period. Although these are simple concepts, they are continually used throughout this text. Hence, a clear and definitive understanding of period and frequency are mandatory for addressing virtually any vibration problem.
Returning to the pendulum of Fig. 2-1, a gravitational force is constantly
acting on the mass. This vertical force is the weight of the block. From physics it
is known that weight W is equal to the product of mass M, and the acceleration of
gravity G. As the pendulum oscillates through an angular displacement , this
force is resolved into two perpendicular components. The cosine term is equal
and opposite to the tension in the string, and the sine component is the Restoring
Force acting to bring the mass back to the vertical equilibrium position. For
small values of angular displacement, sin is closely approximated by the angle
expressed in radians. Hence, this restoring force may be represented as:
Restoring Force W
(2-4)
Similarly, the maximum distance traveled by the mass may also be determined from plane geometry. As shown in Fig. 2-1, the cable length is known, and
the angular displacement is specified by . The actual change in lateral position
for the mass is the distance from A to B, or from B to C. In either case, this distance is equal to L sin. Once more, for small angles, sin in radians, and the
total deflection from the equilibrium position may be stated as:
Deflection L
(2-5)
This repetitive restoring force acting over the same distance has a spring
like quality. In actuality, this characteristic may be defined as the horizontal
stiffness K of this simple mechanical system as follows:
Force
Stiffness = K = -----------------------------Deflection
(2-6)
If equations (2-4) and (2-5) are substituted into (2-6), and if the weight W is
replaced by the equivalent mass M times the acceleration of gravity G, the following expression is produced:
13
Force
W
W
MG
K = ------------------------------ --------------- = ----- = ----------------Deflection L
L
L
(2-7)
Later in this chapter it will be shown that the natural frequency of oscillation for an undamped single degree of freedom system is determined by equation
(2-44) as a function of mass M and stiffness K. If equation (2-7) is used for the
stiffness term within equation (2-44), the following relationship results:
=
K
------ =
M
MG 1
----------------- ------ =
L
M
G
---L
(2-8)
Equation (2-8) is often presented within the literature for describing the
natural frequency of a simple pendulum. A direct example of this concept may be
illustrated by considering the motion of the pendulum in a grandfathers clock.
Typically, the pendulum requires 1.0 second to travel one half of a stroke, or 2.0
seconds to transverse a complete stroke (i.e., one complete cycle). The length L of
the pendulum may be determined by combining equations (2-3) and (2-8):
2
= ------------------- =
Period
G
---L
If the period is represented in terms of the pendulum length L, the above

expression may be stated as:
L
Period = 2 ---G
(2-9)
Equation (2-9) is a common expression for characterizing a simple pendulum. The validity of this equation may be verified in technical references such as
Marks Handbook4. For the specific problem at hand, equation (2-9) may be
solved for the pendulum length. Performing this manipulation, and inserting the
gravitational constant G, plus the period of 2.0 seconds, the following is obtained:
2
G Period
( 386.1 Inches/Second ) ( 2.0 Seconds )
L = --------------------------------- = -------------------------------------------------------------------------------------------- = 39.12 Inches
2
2
4
4
Thus, the pendulum length in a grandfathers clock should be 39.12 inches.
This value is accurate for a concentrated mass, and a weightless support arm. In
an actual clock, the pendulum is often ornate, and weight is distributed along
the length of the support arm. This makes it difficult to accurately determine the
location of the center of gravity of the pendulum mass. Nevertheless, even rough
measurements reveal that the pendulum length is in the vicinity of 40 inches. In
addition, clock makers normally provide a calibration screw at the bottom of the
pendulum to allow the owner to adjust the clock accuracy. By turning this adjustment screw, the effective length of the pendulum may be altered. From the previ4 Eugene A. Avallone and Theodore Baumeister III, Marks Standard Handbook for Mechanical Engineers, Tenth Edition, (New York: McGraw-Hill, 1996), p. 3-15.
14
Chapter-2
ous equations, it is clear that changing the pendulum length will alter the period
of the pendulum. By moving the weight upward, and decreasing the arm length,
the clock will run faster (i.e., higher frequency with a shorter period). Conversely,
by lowering the main pendulum mass, the length of the arm will be increased,
and the clock will run slower (i.e., a lower frequency with a longer period).
Although the grandfather clock is a simple application of periodic motion, it
does provide a realistic example of the fundamental concepts. Additional complexity will be incorporated later in this text when the behavior of a compound
pendulum is discussed. It should be noted that a compound pendulum is a
mechanical system that normally contains two degrees of freedom. This additional flexibility might be obtained by adding flexible members such as springs,
or additional masses to a simple system. In a two mass system, each mass might
be capable of moving independently of the other mass. For this type of arrangement, each mass must be tracked with an independent coordinate system, and
this would be considered as a two degree of freedom system.
The number of independent coordinates required to accurately define the
motion of a system is termed the Degree of Freedom of that system. Process
machinery displays many degrees of freedom, and accurate mathematical
description of these systems increases proportionally to the number of required
coordinates. However, in the case of the simple pendulum, only one coordinate is
required to describe the motion and the pendulum is a single degree of freedom system exhibiting harmonic motion. More specifically, this is an example of
basic dynamic motion where the restoring force is proportional to the displacement. This is commonly referred to as Simple Harmonic Motion (SHM). Other
devices such as the undamped spring mass (Fig. 2-7), the torsional pendulum
(Fig. 2-25), the particle rotating in a circular path, and a floating cork bobbing up
and down in the water at a constant rate are all examples of SHM.
Before expanding the discussion to more complex systems, it is desirable to
conclude the discussion of the simple pendulum. Once again, the reader is
referred back to the example of the oscillating pendulum depicted in Fig. 2-1. On
this diagram, it is meaningful to mentally trace the position of the mass during
one complete cycle. Starting at the vertical equilibrium position B, the displacement is zero at time equal to zero. One quarter of a cycle later, the mass has
moved to the maximum positive position C. This is followed by a zero crossing at
point B as the mass approaches the maximum negative value at position A. The
last quarter cycle is completed as the mass returns from the A location back to
the original equilibrium, or center rest point B.
Intuitively, the mass achieves zero velocity as it swings back and forth to
the maximum displacement points A and C (i.e., the mass comes to a complete
stop). In addition, the maximum positive velocity occurs as the mass moves
through point B from left to right, combined with a maximum negative velocity
as the mass moves through B going from right to left. Finally, the mass must deaccelerate going from B to C, and accelerate from C back to point A. Then the
mass will de-accelerate as it moves from A back to the original equilibrium point
B that displays zero lateral acceleration.
Another way to compare and correlate the displacement, velocity, and accel-
15
eration characteristics of this pendulum would be a time domain examination.

Although a meaningful visualization of the changes in displacement, velocity,
and acceleration with respect to time may be difficult a mathematical description simplifies this task. For instance, assume that the periodic displacement of
the mass may be described by the following fundamental equation relating displacement and time:
Displacement = D sin ( 2 F t )
where: Displacement
D
F
t
=
=
=
=
(2-10)
Instantaneous Displacement
Maximum Displacement (equal to pendulum position A or C)
Frequency of Oscillation
Time
In a rotating system, such as a centrifugal machine, this expression can be

simplified somewhat by substituting the rotational frequency that was previously defined in equation (2-2) to yield:
Displacement = D sin ( t )
(2-11)
The instantaneous velocity of this periodic motion is the time derivative of

displacement. Velocity may now be determined as follows:
Velocity =
d
Displacement = D cos ( t )
dt
By converting the cosine to a sine function, expression (2-12) is derived:

Velocity = D sin ( t + 2 )
(2-12)
Note that velocity leads displacement by /2 or 90. Another way to state

the same concept is that displacement lags behind velocity by 90 in the time
domain. The same procedure can now be repeated to examine the relationship
between velocity and acceleration. Since acceleration is the time rate of change
of velocity, the first time derivative of velocity will yield acceleration. The same
result may be obtained by taking the second derivative of displacement with
respect to time to obtain acceleration:
Acceleration =
d
dt
2
2
Displacement = D sin ( t )
By adding to the sine term, the negative sign is removed, and the following expression is obtained:
2
Acceleration = D sin ( t + )
(2-13)
Acceleration leads displacement by or 180, and it leads velocity by 90. It

may also be stated that displacement lags acceleration by 180 in time. The relationship between displacement, velocity, and acceleration may be viewed graphically in the polar coordinate format of Fig. 2-2. This diagram reveals that
16
Chapter-2
Velocity Vector
Vel. = D sin( t + /2)
Displacement Vector
Displ. = D sin( t)
t + /2
Time
t
t+
Phase
Acceleration Vector
Accel.= D 2 sin( t + )
Fig. 22 Timing Relationship Between Displacement Velocity, and Acceleration
mechanical systems in motion do display a consistent and definable relationship

between frequency, and the respective displacement, velocity, and acceleration of
the body in motion.
Understanding the timing between vectors is mandatory for diagnosing
machinery behavior. It is very easy to become confused between terms such as
leading and lagging, and the diagnostician might inadvertently make a 90 or a
180 mistake. In some instances, this type of error might go unnoticed. However,
during rotor balancing, a 180 error in weight placement might result in excessive vibration or even physical damage to the machine. This type of error is
totally unnecessary, and it may be prevented by establishing and maintaining a
consistent timing or phase convention.
From Fig. 2-2, it is noted that time is shown to increase in a counterclockwise direction. If this diagram represented a rotating shaft, time and rotation
would move together in a counterclockwise direction. As discussed in succeeding
chapters, phase is measured from the peak of a vibration signal backwards in
time to the reference trigger point. This concept is illustrated in Fig. 2-3 that
depicts a rotating disk with a series of angles marked off at 45 increments.
Assume that the disk is turning counterclockwise on the axial centerline. If this
rotating disk is observed from a stationary viewing position, the angles will move
past the viewing point in consecutive order.
That is, as the disk turns, the angles progress in a 0-45-90-135-180-225270-315 consecutive numeric order past the fixed viewing position. However, if
the angles increased with rotation, the observed viewing order would be backwards. Since this does not make good physical sense, the direction of numerically
increasing angles are always set against shaft rotation as in Fig. 2-3. This angular convention will be used throughout this text, and vector angles will always be
considered as degrees of phase lag. This convention applies to shaft and casing
vibration vectors, balance weight vectors, balance sensitivity vectors, plus all
17
225
Time
and
Rotation
270
180
315
135
of
is
n
Ax tatio
Ro
Stationary
Viewing
Position
0
90
45
Angle or
Phase
Direction
Fig. 23 Traditional Angle Designation On A Rotating Disk
analytically calculated vectors. In short, angles and the associated phase are
measured against rotation based upon this physical relationship.
For proper identification, phase angles should be specified as a phase lag, or
provided with a negative sign. In most cases, it is convenient to ignore the negative sign, and recognize that these angles are phase lag values. Using this convention, phase between the 3 vibration vectors in Fig. 2-2 may be converted by:
Phase displacement = Phase velocity + 90
(2-14)
Phase displacement = Phase acceleration + 180
(2-15)
Phase velocity = Phase acceleration + 90
(2-16)
If a velocity phase angle occurs at 225, it is determined from (2-14) that

the displacement phase angle is computed by: 225+90=315. Similarly, the
velocity phase may be converted to an acceleration phase from equation (2-16)
as: 225-90=135. If the phase lag negative sign is used, the angle conversions in
equations (2-14) to (2-16) must also be negative (i.e., -90 and -180). In either
case, consistency is necessary for accurate and repeatable results.
In addition to phase, the vibration magnitude of an object may be converted
from displacement to velocity or acceleration at a constant frequency. This
requires a conversion of units within the motion equations (2-12) and (2-13). For
example, consider the following definition of English units for these parameters:
D = Displacement Mils,peak to peak = Mils,p-p
V = Velocity Inches/Second,zero to peak = IPS,o-p
A = Acceleration Gs,zero to peak = Gs,o-p
F = Frequency Cycles/Second (Hz)
Reinstalling 2F for the frequency , and considering the peak values of the
18
Chapter-2
terms (i.e., sin=1), equation (2-12) may be restated as follows:

V = D = D 2 F
Since velocity is generally defined as zero to peak (o-p), and displacement is
typically considered as peak to peak (p-p), the displacement value must be halved
to be consistent with the velocity wave. Applying the appropriate physical unit
conversions, the following expression evolves:
D
1 Inch
Radians
Cycles
V = ---- Mils ------------------------- 2 ------------------- F -----------------
2
Cycle
Second
1, 000 Mils
Which simplifies to the following common equation:
DF
V = ---------------318.31
(2-17)
Next, consider the relationship between acceleration and displacement as

described by equation (2-13), and expanded with proper engineering units to the
following expression:
D
1 Inch
2
Radians
Cycles 2
A = D = ---- Mils ------------------------- 2 ------------------- F -----------------
2
Cycle
Second
1, 000 Mils
2
D F Inches
A = ------------------ -------------------2 50.661 Second
Acceleration units for the above conversion are Inches/Second2. Measurement units of Gs can be obtained by dividing this last expression by the acceleration of gravity as follows:
2
1G
D F Inches
A = ------------------ -------------------2- ----------------------------------------------2 50.661 Second 386.1 Inches/Second
This conversion expression may be simplified to the following format:
2
DF
F 2
A = ------------------ = D -------------
19, 560
139.9
(2-18)
The relationship between acceleration and velocity may be stated as:

A = V = V 2 F
Expanding this expression, and including dimensional units, the following
equation for converting velocity at a specific frequency to acceleration evolves:
1G
Inches
Radians
Cycles
A = V --------------- 2 ------------------- F ----------------- ----------------------------------------------2-
Second
Cycle
Second 386.1 Inches/Second
19
Simplifying this expression, the following common equation is derived:

VF
A = --------------61.45
(2-19)
The last three equations allow conversion between displacement, velocity,

and acceleration at a fixed frequency measured in Cycles per Second (Hz). A set
of expressions for frequency measured in Cycles per Minute (CPM) may also be
developed. Since machine speeds are measured in Revolutions per Minute
(RPM), this additional conversion is quite useful in many instances. Performing
this frequency conversion on equations (2-17), (2-18), and (2-19) produces the
next three common conversion equations:
D RPM
V = -------------------------19, 099
(2-20)
RPM 2
A = D ---------------
8, 391
(2-21)
V RPM
A = -------------------------3, 687
(2-22)
The vibration units for equations (2-20), (2-21), and (2-22) are identical to
the English engineering units previously defined. However, the frequency for
these last three equations carry the units of Revolutions per Minute (i.e., RPM or
Cycles per Minute).
The simultaneous existence of three parameters (i.e., displacement, velocity, and acceleration) to describe vibratory motion can be confusing. This is further complicated by the fact that instrumentation vendors are often specialized
in the manufacture of a single type of transducer. Hence, one company may promote the use of displacement probes, whereas another vendor may strongly
endorse velocity coils, and a third supplier may cultivate the application of accelerometers. The specific virtues and limitations of each of these types of transducer systems are discussed in greater detail in chapter 6 of this text. However,
for the purposes of this current discussion, it is necessary to recognize that displacement, velocity, and acceleration of a moving body are always related by the
frequency of the motion.
This relationship between variables may be expressed in various ways. For
example, consider an element vibrating at a frequency of 100 Hz (6,000 CPM)
and a velocity of 0.3 IPS,o-p. From equation (2-17) the relationship between velocity and displacement may be used to solve for the displacement as follows:
318.31 0.3 IPS o-p
318.31 V
D = --------------------------- = ------------------------------------------- = 0.955 Mils p-p
F
100 Hz
Similarly, the equivalent acceleration of this mechanical element may be
determined from equation (2-19) in the following manner:
20
Chapter-2
0.3 IPS o-p 100 Hz

VF
A = --------------- = ------------------------------------------ = 0.488 Gs o-p
61.45
61.45
10
Displacement
H H
0.1
0.01
1000
60,000
100
6,000
10
600
1
60
Vibration Amplitude (Mils, IPS, G's)
100
10000
600,000
Frequency (Cycles/Minute)
Velocity
H H H
B
B
B
J B
B J
HB H H
HJ H H
J
J
Acceleration
20000
1,200,000
Thus, the displacement and acceleration amplitudes for this velocity may
be computed for any given frequency. Another way to view this interrelationship
between parameters is to extend this calculation procedure to a large range of
frequencies, and plot the results as shown in Fig. 2-4. Within this diagram, the
velocity is maintained at a constant magnitude of 0.3 IPS,o-p and the displacement and acceleration amplitudes calculated and plotted for several frequencies
between 1 and 20,000 Hz (60 and 1,200,000 CPM).
H H H
J
J
20,000
Frequency (Hertz)
10,000
1,000
100
10
0.001
Fig. 24 Equivalent Displacement, Velocity, and Acceleration Amplitudes V. Frequency
Fig. 2-4 shows that displacement is large at low frequencies, and acceleration is larger at high frequencies. From a measurement standpoint, displacement would be used for lower frequencies, and acceleration would be desirable
for high frequency data. Again, specific transducer characteristics must also be
considered, and the reader is referred to chapter 6 for additional details on the
actual operating ranges of transducers.
For purposes of completeness, it should be recognized that the circular functions previously discussed can be replaced by an exponential form. For instance,
equation (2-23) is a normal format for these expressions:
Displacement = D e
it
(2-23)
In this equation, i is equal to the square root of minus 1 and e is the natural log base that has a value of 2.71828. This expression will satisfy the same
equations, and produce identical results to the circular formats. However, it is
Vector Manipulation
21
sometimes easier to manipulate equations using an exponential form rather

than a circular function. For reference, the relationship between the exponential
and the circular function is shown as follows:
it
= cos ( t ) + i sin ( t )
(2-24)
The cos(t) term is often referred to as the Real, or the In-Phase component. The i sin(t) term is the projection of the vector on the imaginary axis.
This is normally called the Imaginary, or the Quadrature component. These
terms are used interchangeably. It should be understood that the form, and not
the intent of the equations has been altered. It should also be mentioned that
both the Real and Imaginary (In-Phase and Quadrature) components must satisfy the equation of motion for the mechanical system.
VECTOR MANIPULATION
Many physical characteristics of machines are described with vectors. A
magnitude is joined with a directional component to provide a parameter with
real physical significance. These vector quantities are routinely subjected to various types of mathematical operations. More specifically, the addition, subtraction, multiplication, and division of vectors must be performed as an integral
part of vibration and modal analysis, rotor balancing, analytical modeling, plus
instrumentation calibration.
For reference purposes, it is necessary to define the methods used for vector
manipulation. The different vector operations may be performed with a hand
held calculator, they may be executed with the math tools incorporated in
spreadsheets, or they may be included as subroutines into computer programs.
In addition, some Dynamic Signal Analyzers (DSA) use vector math as part of
the signal processing and computational capabilities. In all cases, these fundamental math operations must be performed in a consistent manner.
From an explanatory standpoint, the specific vector equations will be
shown, and a numeric example will be presented for each type of operation. The
examples will be performed with circular coordinates, however an exponential
form will provide an identical solution. For consistency, the following pair of
polar coordinate vectors will be used throughout this series of explanations:
V a = A
(2-25)
V b = B
(2-26)
The first vector (2-25) has of a magnitude A, occurring at an angle . Similarly, the second vector (2-26) has an amplitude of B, and an angle . As previously discussed, these vectors may be represented in a Cartesian coordinate (XY) system by the following pair of equations:
22
Chapter-2
V a = V a + V a = A cos + A sin
x
V b = V b + V b = B cos + B sin
x
Multiplying the amplitude by the cosine and sine of the associated angle
will allow conversion from polar to rectangular coordinates. The cosine term represents the magnitude on the X-Axis, and the sine term identifies the amplitude
on the Y-Axis. From the last pair of equations, the individual Cartesian amplitudes for each vector component may be summarized as:
V a = A cos
(2-27)
V a = A sin
(2-28)
V b = B cos
(2-29)
V b = B sin
(2-30)
This conversion of the initial vectors now provides the format to allow the
addition and subtraction of two vector quantities. Vector addition is performed
by summing the individual X and Y components, and converting from Cartesian
back to polar coordinates. The summation of X-Axis components is achieved by
adding equations (2-27) and (2-29) in the following manner:
V add = V a + V b = A cos + B cos
x
(2-31)
Similarly, the Y-Axis summation component is obtained by addition of the

previously described equations (2-28) and (2-30) as follows:
V add = V a + V b = A sin + B sin
y
(2-32)
The X and Y summation components are now be converted back into polar
coordinates of amplitude and angle as shown in equations (2-33) and (2-34):
V add =
( V add ) + ( V add )
x
V add
add = atan ----------------y
V add x
(2-33)
(2-34)
Vector addition is used with many different types of calculations. For

instance, consider the installation of two weights into the balance ring on a turbine rotor. If the weights are both installed in the same hole, the effective weight
correction would be the simple sum of the two weights. However, if the weights
are screwed into two different holes, the effective balance correction must be
Vector Manipulation
23
90
Effective Weight Vector
r
Vadd =
88.6 Grams @ 69
Second Weight Vector
r
Vb = 40 Grams @ 80
First Weight Vector
r
Va =
50 Grams @ 60
0
180
Fig. 25 Vector Addition

Of Two Balance Weights
270
determined by vector addition of the two individual weight vectors. For demonstration purposes, assume that 50 Grams was inserted into a hole at 60, and 40
Grams was installed at the 80 hole as described in Fig. 2-5. The initial weight
vectors are represented with equations (2-25) and (2-26) as:
V a = A = 50 Grams 60
V b = B = 40 Grams 80
The summation of horizontal vector components in the X-Axis is determined with equation (2-31):
V add = A cos + B cos
x
V add = 50 cos 60 + 40 cos 80

x
V add = 50 0.500 + 40 0.174 = 25.00 + 6.95 = 31.95 Grams

x
Similarly, the summation of vertical vector components in the Y-Axis may

be computed with equation (2-32) as follows:
V add = A sin + B sin
y
V add = 50 sin 60 + 40 sin 80

x
V add = 50 0.866 + 40 0.985 = 43.30 + 39.39 = 82.69 Grams

x
The calculated X and Y balance weights identify the combined effect of both
weights in the horizontal and vertical directions. These weights are actually X
and Y coordinates that may be converted to a polar coordinate magnitude using
equation (2-33) in the following manner:
24
Chapter-2
2
( V add ) + ( V add )
V add =
( 31.95 ) + ( 82.69 ) =
V add =
7, 858 = 88.64 Grams
Finally, the angle of the resultant vector may now be determined from
equation (2-34) as shown:
V add
add = atan ----------------y
V add x
82.69
add = atan ------------- = atan ( 2.588 ) = 68.9
31.95
Thus, the 50 Gram weight installed at 60 plus the 40 Gram weight at 80
are vectorially equivalent to 88.6 Grams at 69 (as shown in Fig. 2-5). The magnitude of this vector sum is the net effective weight that should be used for additional balancing calculations such as centrifugal force. The effective angle of this
weight pair is necessary information for intermediate balancing response calculations, as well as the documentation of final results. For more information on
this type of calculation, please refer to chapter 11 of this text.
The same basic approach is used for vector subtraction, with one significant difference. Instead of adding Cartesian coordinates, the X and Y components are subtracted. That is, by subtracting the B vector from the A vector, the
X-Axis change is obtained by subtracting equation (2-29) from (2-27):
V sub = V a V b = A cos B cos
x
(2-35)
In a similar manner, the Y-Axis component is obtained by subtraction of the

previously identified equation (2-30) from (2-28) as follows:
V sub = V a V b = A sin B sin
y
(2-36)
Calculation of the differential vector is achieved with equations (2-37) and

(2-38) that are identical in form to the vector addition conversions:
V sub =
( V sub ) + ( V sub )
x
V sub
sub = atan ---------------y
V sub x
(2-37)
(2-38)
This type of vector computation is extremely useful for performing routine

tasks such as runout subtraction on proximity probe displacement signals. For
instance, Fig. 2-6 displays a synchronous vibration vector at full operating speed
of 2.38 Mils,p-p at an angle of 134. Assume that the slow speed 1X runout was
Vector Manipulation
25
Runout Corrected Vector
r
Vsub =
2.01 Mils,p-p @ 157

90
Runout Vector
Initial Vector
r
Vb = 0.94 Mils,p-p @ 78
r
Va = 2.38 Mils,p-p @ 134
Fig. 26 Vector Subtraction Of Shaft Runout From
Running Speed Vector
180
270
measured to be 0.94 Mils,p-p, at 78. Subtraction of the slow roll from the full
speed vector yields a compensated, or a runout corrected vector.
Mathematically, the initial vibration at running speed may be identified as
the A vector, and the slow roll runout may be represented by the B vector. Substitution of the defined vibration vectors into equations (2-25) and (2-26) provides
the following vectors for subtraction:
V a = A = 2.38 Mils p-p 134
V b = B = 0.94 Mils p-p 78
The difference between horizontal X-Axis vector components is determined
with equation (2-35) in the following manner:
V sub = A cos B cos
x
V sub = 2.38 cos 134 0.94 cos 78

x
V sub = 2.38 ( 0.695 ) 0.94 0.208 = 1.654 0.196 = 1.850 Mils p-p
x
Similarly, the difference of vector components in the vertical Y-Axis may be

computed with equation (2-36):
V sub = A sin B sin
y
V sub = 2.38 sin 134 0.94 sin 78

y
V sub = 2.38 0.719 0.94 0.978 = 1.711 0.919 = 0.792 Mils p-p
y
The negative value for the horizontal component is perfectly normal, and
acceptable. This negative sign, combined with the positive sign on the vertical
component, identifies that the final vector will reside in the upper left polar
quadrant (i.e., angle between 90 and 180). The computed X and Y coordinates
may now be converted to polar coordinates using equation (2-37) to determine
the magnitude of the runout corrected vector:
26
Chapter-2
V sub =
V sub =
( V sub ) + ( V sub )
x
( 1.850 ) + ( 0.792 ) =
4.050 = 2.01 Mils p-p
The angle of the runout compensated vector may now be calculated from
equation (2-38) as follows:
V sub
sub = atan ---------------y
V sub x
0.792
sub = atan ---------------- = atan ( 0.428 ) = 23.2
1.850
sub = 23.2 + 180 = 156.7
The 180 addition to the angle is a quadrant correction. Thus, subtracting a
runout of 0.94 Mils,p-p at 78 from the full speed vector of 2.38 Mils,p-p at 134
yields a runout compensated vector quantity of 2.01 Mils,p-p at 157. This calculated result is in full agreement with the vector diagram shown in Fig. 2-6. This
compensated vector represents the actual dynamic motion (i.e., vibration) of the
shaft. For more information on runout compensation, please refer to chapters 6,
7, 8, and 11.
The major complexity associated with vector addition and subtraction is
due to the necessity for converting from polar to Cartesian coordinates, performing a simple operation, and then converting from Cartesian back to polar coordinates. Fortunately, this multiple conversion is not required for vector
multiplication and division.
Vector multiplication of two vector quantities may be executed by simply
multiplying amplitudes, and adding the respective phase angles as follows:
V mul = V a V b = ( A B ) ( + )
(2-39)
This manipulation is easy to perform, and the only cautionary note resides
with the value of the angle. In many cases, this may exceed 360, due to the size
of angles and . When a full circle has been exceeded (i.e., final angle greater
than 360), the size of the angle may be reduced by 360 to yield a physically
meaningful angle between 0 and 360.
Vector multiplication is necessary in the machinery diagnosis business. For
example, consider the situation of determining the required balance weight to
correct the 1X vibration response of a machine. Presuming that the unit has a
properly defined balance sensitivity vector, the required balance weight and
angle can be determined from equation (2-39). This requires a vector multiplication between the measured vibration, and the sensitivity vector. For demonstration purposes, assume that the measured vibration vector is 2.0 Mils,p-p at an
angle of 40. Further assume that the rotor balance sensitivity vector is equal to
Vector Manipulation
27
150.0 Grams/Mil,p-p at an angle of 190. Based on this data, the operable vectors
for this vector manipulation are identified as:
V a = A = 2.0 Mils p-p 40
V b = B = 150 Grams/Mil p-p 190
Multiplication of these two vectors is performed with equation (2-39) as:
V mul = ( A B ) ( + )
V mul = ( 2.0 Mils p-p 150 Grams/Mil p-p ) ( 40 + 190 ) = 300 Grams 230
This vector product indicates that the installation of a 300 Gram weight at
an angle of 230 will balance the measured synchronous response of 2.0 Mils,p-p
at 40. Naturally, the accuracy of this value is dependent upon the correctness of
the balance sensitivity vector.
As described in further detail in chapter 11, a vector summation between
the calculated vibration from the weight, plus the current vibration vector will
result in a predicted vibration vector with the weight attached. An additional
vector summation with the shaft runout will produce an uncompensated 1X vector. For a perfectly linear mechanical system, this would be the vibration amplitude and phase displayed by a synchronous tracking filter. Although this
discussion is somewhat premature within the sequence of this text, the main
point is that vector calculations may involve a string of manipulations to achieve
the necessary result.
Vector division represents the final category of vector math. Referring
back to the initial vectors, equations (2-25) and (2-26), vector division is performed by dividing the amplitudes, and subtracting the angles as follows:
Va
A
V div = ------- = ---- ( )
B
Vb
(2-40)
This kind of manipulation is also easy to perform, and again a cautionary

note resides with the final value of the angle. In many cases, this angle may drop
below 0, due to the relative size of angles and . When the zero point is crossed
(i.e., negative angle), the size of the angle may be increased by 360 to yield a
physically meaningful angle between 0 and 360.
Vector division is widely used for various types of machinery calculations.
For instance, the computation of a balance sensitivity vector requires the division of a calibration weight vector by a differential vibration response vector. The
technical details associated with this calculation are in chapter 11. However,
from a pure computational standpoint, consider the following initial pair of vectors for division.
28
Chapter-2
V a = A = 400 Grams 230

V b = B = 5.00 Mil p-p 60
Division of these two vectors is performed with equation (2-40) as follows:
Va
A
V div = ------- = ---- ( )
B
Vb
400 Grams
V div = -------------------------- ( 230 60 ) = 80.0 Grams/Mil p-p 170
5.00 Mil
p-p
This calculation identifies a single balance sensitivity vector based upon a
measured differential response vector of 5.00 Mils,p-p at an angle of 60. This vector change in shaft vibration response was due to the installation of a 400 Gram
weight at an angle of 230. Vector division of the weight by the differential vibration vector yields the balance sensitivity vector of 80.0 Grams/Mil,p-p at an angle
of 170. This unbalance sensitivity vector may now be used to compute balance
corrections in a manner similar to the earlier example of vector multiplication.
These simplified rules for vector multiplication and division may be verified
by performing the same operations using exponential functions instead of the
presented polar coordinates. The results will be identical, and this will reinforce
the concept that the vector math may be successfully executed using either exponential or circular functions. In all cases, these vector manipulations are continually used throughout the field of machinery analysis, and these procedures
must be mastered to allow progression to the real machinery topics.
UNDAMPED FREE VIBRATION

Expanding upon the concepts of the previous section, again consider the
single mass pendulum of Fig. 2-1. Within this earlier mechanical system, the
mass of the concrete block was identified as the only significant element in the
system. If this concrete block remains constant, and if the weightless cable is
replaced by a coil spring, the simple spring mass system of Fig. 2-7 is produced.
Assume that the spring is suspended from a totally rigid I-Beam, and consider
the mass to be confined to movement only in the vertical direction. Since damping is not involved, this is considered as an undamped mechanical system. In
addition, there are no external forces applied to this system, so it must be classified as a system that exhibits free vibration when it is displaced, and allowed to
oscillate in the vertical plane. The resultant motion is defined as undamped free
vibration of this one degree of freedom mechanical system.
If this physical example is converted into a traditional physics diagram, the
sketches shown in Fig. 2-8 evolve. The left diagram shows the main mechanical
elements, and the right sketch displays the Free Body Diagram. Normally, this
mechanical system would remain at rest (i.e., no motion). For this system to
Undamped Free Vibration
29
Spring Force =
- Stiffness x Displacement
Stationary I-Beam
Spring with
Stiffness = K
Coil Spring
Mass = M
Equilibrium
D=0
Displacement = +D
Mass
Gravity Force =
Mass x Acceleration
Fig. 27 Spring Mass
Mechanical System
Fig. 28 Equivalent Spring Mass Mechanical System And

Associated Free Body Diagram
move, some type of initial disturbance is required. Furthermore, when this

mechanical system is in motion, the free body diagram (Fig. 2-8) reveals two
active forces; a spring force, and the gravitational term. The general equation of
motion for this body is simply the equality of active forces as follows:
( Mass Acceleration ) = ( Stiffness Displacement )
By rearranging terms, the following summation of forces is obtained:
( Mass Acceleration ) + ( Stiffness Displacement ) = 0
Substituting a simpler alpha identification for each of the four variables,
the equation of motion for this simple spring mass system may be stated in the
manner that W. T. Thomson used:
( M A) + ( K D) = 0
(2-41)
If equation (2-41) is divided by the mass, the resultant expression contains

a system mechanical constant (i.e., K/M), plus the interrelated acceleration and
displacement of the body:
K
A + ------ D = 0
M
(2-42)
Equation (2-42) can be satisfied by either of the previously discussed circular or exponential functions. For simplicity, assume that an exponential function
as defined in equation (2-23) is substituted into (2-42) to yield the following version of the equation of motion:
30
Chapter-2
2
D e
it
it
K
+ ------ D e
= 0
M
Extracting the common terms from this equation, the following is obtained:
De
it
2 K
+ ------ = 0
(2-43)
Equation (2-43) is satisfied for all values of time t when the terms within
the brackets are equated to zero:
K
2 + ----- = 0
M
This may now be solved for the natural or critical frequency c as follows:
c =
K
-----M
(2-44)
Another common form of this expression is obtained by converting the rotational frequency c units of Radians per Second to Cycles per Second in accordance with equation (2-2) to yield the following:
1
K
F c = ------ -----2
M
(2-45)
Clearly, the frequency of oscillation is a function of the spring constant, and

the mass. This is the undamped natural frequency of the mechanical system. It
is also commonly called the undamped critical frequency, and the subscript c
has been added to identify frequencies c and Fc. In all cases, following an initial
disturbance, the mass will oscillate (or vibrate) at this natural frequency, and the
amplitude of the motion will gradually decay as a function of time. This reduction in amplitude is due to energy dissipation within a real mechanical system.
Although this result is simple in format, it does represent an extraordinarily important concept in the field of vibration analysis. That is, the natural frequency of a mechanical resonance will respond to an alteration of the stiffness
and the mass. Often, the diagnostician has limited information on the effective
stiffness, or equivalent mass of the mechanical system. However, changes in
stiffness or mass will behave in the manner described by equation (2-44). In
many instances, this knowledge of the proper relationship between parameters
will allow a respectable solution to a mechanical problem.
Initially, the existence of a unique natural frequency that is a function of
the mechanical system mass and stiffness may appear to be only of academic
interest. In reality, there are field applications of this physical relationship that
may be used to provide solutions for mechanical problems. For instance, if a
mechanical system is excited by a periodic force at a frequency that approaches a
natural resonant frequency of the mechanical system the resultant vibratory
31
motion may be excessive, or even destructive. Three potential solutions to this

type of problem were identified by J. P. Den Hartog5, in his text Mechanical
Vibrations. Quoting from page 87 of this book:
In order to improve such a situation, we might first attempt to eliminate
the force. Quite often this is not practical or even possible. Then we may change
the mass or the spring constant of the system in an attempt to get away from the
resonance condition, but in some cases this is also impractical. A third possibility
lies in the application of the dynamic vibration absorber, invented by Frahm in
1909The vibration absorber consists of a comparatively small vibratory system
k, m attached to the main mass M. The natural frequency k m of the attached
absorber is chosen to be equal to the frequency of the disturbing force. It will be
shown that then the main mass M does not vibrate at all, and that the small system k, m vibrates in such a way that its spring force is at all instances equal and
opposite to Po sin t. Thus there is no net force acting on M and therefore the
mass does not vibrate
In his text book, Den Hartog proceeds to derive a detailed equation set that
supports the above statements. He also examines torsional systems, and damped
vibration absorbers. Thomson6 also discussed the utilization of both lateral and
torsional vibration absorbers. However, for this discussion, the application of a
simple lateral undamped spring mass vibration absorber will be reviewed. The
fundamental engineering principles behind an absorber installation are illustrated with the following case history.
Case History 1: Piping System Dynamic Absorber
The mechanical system under consideration consists of a pair of product
transfer pumps that were subjected to a modification of the discharge piping to
span across a new roadway. These essential pumps were motor driven at a constant speed of 1,780 RPM. The pumps had a successful eight year operating history, with only minor seal problems, and one coupling failure. During a plant
revision, the pump discharge piping was rerouted to a new pipe rack. Due to the
design of the new rack, the discharge line was poorly supported, and problems
began to appear on both pumps shortly after the piping modification.
Multiple seal failures were combined with repetitive bearing, and coupling
failures. These two pumps that previously received maintenance attention only
once or twice a year were now subjected to overhauls on a monthly basis. This
increased maintenance passed unnoticed for a long time. Unfortunately, one
night the main pump failed when the spare pump was out for repairs. This coincidence of mechanical failures forced a plant outage, and this event focused management attention upon the reduced reliability of these pumps.
Vibration analysis of the pumps and the associated piping revealed a domi5 J.P. Den Hartog, Mechanical Vibrations, 4th edition, (New York: McGraw-Hill Book Company,
1956), p. 87.
6 William T. Thomson, Theory of Vibration with Applications, 4th Edition, (Englewood Cliffs,
New Jersey: Prentice Hall, 1993), pp. 150-159.
32
Chapter-2
nant motion at the pump running speed of 1,780 RPM. Comparison with historical data revealed 1X vibration amplitudes on the pump and motor were ten to
twenty times higher than previously measured. This machinery abnormality was
coincident with vertical vibration levels in excess of 25 Mils,p-p at the middle of
the unsupported discharge line (i.e., midspan of the road crossing).
A temporary brace was fabricated, and placed below the discharge line.
This support reduced the piping vibration, and also resulted in a drop in the
pump synchronous motion. Considering the positive results of this test, and
some preliminary calculations on the natural frequency of the piping span, it
was concluded that the pump running speed was very close to a lateral natural
frequency of the new discharge pipe.
Since a brace in the middle of the road was unacceptable as a long-term
solution, other possibilities were examined and discarded. Finally, the application of a tuned spring mass vibration absorber was considered as a potential and
practical solution. For this problem a simple horizontal cantilevered vibration
absorber was designed to resemble the diagram in Fig. 2-9.
Sliding Overhung Mass

Spring
Pipe I.D.
Fig. 29 Typical Tuned

Spring Mass Vibration
Absorber Assembly
For Piping System
Fabricated
Pipe Saddle
This device consists of a fabricated pipe saddle that is securely bolted to the
outer diameter of the discharge pipe. It is physically located at the point of highest vibration (i.e., center of the piping span). Since the pipe vibrates vertically,
the absorber is positioned horizontally so that the cantilevered weight may also
vibrate vertically. In this case, the spring consists of flat bar stock that has the
most flexible axis placed in the direction of the desired motion. The overhung
mass is bolted to the flat bar stock spring, and it may be moved back and forth to
allow adjustment of the natural frequency.
By inspection of this damper assembly, it is apparent that the stiffness and
mass of the spring, plus the overhung mass are equivalent to a simple spring
mass system. The problem in designing an appropriate vibration absorber is now
reduced to a reasonable selection of physical dimensions to obtain a natural frequency of 1,780 CPM for this installed assembly.
Several approaches may be used to determine an acceptable set of absorber
33
dimensions. For example, a Finite Element Analysis (FEA) could be performed.

However, an FEA approach may become unnecessarily complicated and time
consuming. Use of published beam natural frequency equations may also be considered. However, one must be careful of published canned equations where the
assumptions and boundary conditions may not be clearly explained or understood. Fortunately, a practical approach for performing these calculations was
presented by John D. Raynesford7, in his Hydrocarbon Processing article on this
subject. In this article, he considered the system as a simple spring mass assembly. The dimensions of the spring were combined with the overhung mass to provide the basic elements for the absorber design. Specifically, Raynesford
considered the total static deflection Ytotal of the vibration absorber to be associated with the weight W, mass M, gravitational constant G, and the spring constant K of the assembly in the following manner:
W
MG
K = ---------------- = ----------------Y total
Y total
(2-46)
This is the same general stiffness relationship that was previously applied
to the simple pendulum in equation (2-7). If equation (2-46) for stiffness is placed
into the previously developed natural frequency equation (2-44), the following
substitution and changes may be performed:
c =
K
------ =
M
MG 1
----------------- ------ =
Y total M
G
---------------Y total
Solving for the total deflection Ytotal, the following equation is obtained:
G
Y total = -----2
c
(2-47)
The total end point deflection of the vibration absorber was presumed to be
due to a combination of the uniformly distributed weight of the spring, plus the
cantilevered mass on a weightless beam. Traditional deflection equations for
these two elements may be extracted from various references. For example,
deflection of a beam with a uniformly distributed load may be obtained from references such as Shigley8, or Roark9 as follows:
3
W spring L
Y spring = ----------------------------------8EI
(2-48)
7 John D. Raynesford, Use Dynamic Absorbers to Reduce Vibration, Hydrocarbon Processing,

Vol. 54, No. 4, (April 1975), pp. 167-171.
8 Joseph E. Shigley and Charles R. Mischke, Standard Handbook of Machine Design, (New
York: McGraw-Hill Book Company, 1986), pp 11.5-11.6.
9 Warren C. Young, Roarks Formulas for Stress & Strain, 6th edition, (New York: McGraw-Hill
Book Co., 1989), pp. 100-102.
34
Chapter-2
where: Yspring
Wspring
L
E
I
=
=
=
=
=
End Deflection of Spring (Inches)

Weight of Spring (Pounds)
Length of Spring (Inches)
Modulus of Elasticity (= 30 x 106 Pounds / Inch2 for steel)
Spring Area Moment of Inertia (Inches4)
Similarly, the deflection of a cantilevered mass on a weightless beam may

be extracted from either Shigley, or Roark, as follows:
3
Y mass
where:
W mass L
= ------------------------------3EI
(2-49)
Ymass = End Deflection at Mass (Inches)

Wmass = Weight of Mass (Pounds)
The total deflection due to the weight of the spring plus the cantilevered
mass is obtained by superposition (addition) of these well proven beam deflection
equations as follows:
Y total = Y spring + Y mass
(2-50)
Substituting equations (2-47), (2-48), and (2-49) into the total deflection
equation (2-50) yields the following combined result:
3
W spring L
W mass L
G
------ = ----------------------------------- + ------------------------------2
8EI
3EI
c
(2-51)
At this point, the Raynesford article begins a trial and error solution to
arrive at the vibration absorber dimensions. Another way to obtain a set of realistic dimensions is to pursue a further simplification of the equation. For
instance, equation (2-51) may be solved for the weight of the overhung mass as:
3 G E I 3
W mass = -------------------------------- --- W spring
L3 c2 8
(2-52)
In equation (2-52), the area moment of inertia I for the flat bar stock used
for the spring is determined by the next equation for a rectangular cross section:
3
bh
I = --------------12
where:
(2-53)
b = Width of Rectangular Spring (Inches)

h = Height of Rectangular Spring (Inches)
As always, the spring weight is calculated simply by multiplying volume by

the material density as follows:
35
W spring = b h L
where:
(2-54)
= Material Density (= 0.283 Pounds/Inches3 for steel)
Equations (2-53) and (2-54) will now be substituted back into (2-52), and
simplified to yield the following expression for the overhung mass:
G E b h 3 3 b h L
- -----------------------------------------W mass = -----------------------------------
8
4 L3 c2
This expression contains the known quantities of the acceleration of gravity
G, the modulus of elasticity E, the density of the spring material . If a spring is
constructed from flat stock that is 1 inch wide by 1/2 inch thick, then dimensions
b and h are also defined. The undamped natural frequency of the system c
should be equal to the measured excitation frequency of 1,780 CPM. Performing
these numerical substitutions into the last expression yields:
W mass
Lb
6 Lb
In
3
386.1 -------- 30 10 ------2 1 In ( 0.5 In ) 3 1 In 0.5 In 0.283 ------3 L
2
Sec
In
In
= --------------------------------------------------------------------------------------------------------2- -------------------------------------------------------------------------
8
Min
Rad
4 L 3 1, 780 Cycle
------------
----------------------
2
Cycle 60 Sec
Min
Performing these calculations, the following simplified result is obtained:

21.84 3
Pounds
3
W mass = ------------- Pound-Inch ( 0.053 L ) -------------- L
Inch
(2-55)
Equation (2-55) correlates the weight of the overhung mass to the overhung
length for the defined conditions. The graph shown in Fig. 2-10 is a plot of equation (2-55). It describes this specific relationship between the length of the spring
and the magnitude of the overhung mass. From this plot it is obvious that the
longer the spring, the less mass required. Conversely, as the spring is shortened,
the overhung weight must be increased. For this particular piping problem, a
spring length of 12 inches was selected with an overhung weight calculated from
equation (2-55) of 5.4 pounds. This same weight value could also be extracted
from the curve plotted in Fig. 2-10 for a spring length of 12 inches.
To allow fine tuning of the absorber resonant frequency, the spring was fabricated to be 15 inches long. This additional length does slightly violate the
developed equation array, but the error is small. In addition, it must be recognized that the developed equations do not constitute a rigorous solution, but they
do provide an acceptable solution. Thus, the extra spring length allows the ability to perform a final adjustment of the natural frequency to correct for variations in the calculations, the fabrication process, or the field attachment.
Normally, it is desirable to bench test the vibration absorber in the shop
before installation, and perform most adjustments before installing the device in
the field. In most instances, a simple hammer test with an accelerometer and
36
Chapter-2
Fig. 210 Spring Length

Versus Overhung Weight
For Piping System Lateral
Vibration Absorber
Overhung Weight (Pounds)
20
Vibration Absorber for

1,780 CPM
1" x 1/2" Flat Bar Spring
18
16
14
12
Selected Configuration of
12" Long with 5.4 Pound Weight
10
8
6
4
2
0
8
10
12
14
16
18
20
Length of Spring (Inches)
spectrum analyzer will identify the natural frequency of the absorber. If the natural frequency is low, then the overhung weight should be moved in towards the
support. The opposite is also true. That is, if the measured natural frequency of
the absorber in the shop test is on the high side, then the overhung weight
should be moved away from the support.
In this case, the 5.4 pound weight was finally positioned at 12.5 inches from
the base during the shop frequency response test. Another minor adjustment
was made after the absorber was bolted into place on the discharge line. This
device proved to be successful, and piping vibration was reduced from levels in
excess of 25 Mils,p-p, to a final condition at the pipe midspan of 1.5 to 2.0 Mils,p-p.
More significantly, the vibration amplitudes on the two transfer pumps returned
to previous historical levels, and the failures ceased.
The article by Raynesford also offers the following two important rules
regarding the attachment and fabrication of absorbers:
1. Try to attach the absorber at the point of maximum vibration and in such
a way as to vibrate in the same plane. That is, if the bearing housing vibrates in
the horizontal plane, mount the absorber vertically so it can also vibrate in the
horizontal plane. Adjust the weight in and out until minimum vibration on the
unit (maximum on the absorber) is achieved.
2. A rigid attachment is essential-the wand must flex, not the attachment. Be
careful when using welds. They are prone to failure in the heat effected zone. Make
generous use of large radii at the juncture of the wand and the base or attachment
To this pair of recommendations, it would also be advisable to suggest that
the absorber be shop tuned to the desired natural frequency. This is always easier to perform in the machine shop versus the field. In addition, the vibration
absorber should be installed with a permanent safety chain loosely connecting
the assembly with some adjacent rigid structure. If the support saddle or the
attachment welds fail, this safety chain would restrain the spring mass assembly, and significantly minimize the potential for any personnel injury.
Overall, it must be recognized that a vibration absorber provides a cost-
Free Vibration with Damping
37
effective solution to some difficult problems, and it demonstrates a practical

application of an undamped mechanical system. On the other hand, a vibration
absorber is certainly not a universal solution for all machinery vibration problems. It is often ineffective when used to correct a rotor resonance, or an acoustic
resonance problem. In most instances, this type of vibration absorber is useful
for addressing certain types of structural resonance problems and it should
always be applied with good engineering judgment and common sense.
FREE VIBRATION WITH DAMPING
Spring with
Stiffness=K
Coil Spring
M
O
N
R
O
E
Damper with
Coeff.=C
Damper
Mass
Mass = M
Equilibrium
D=0
Damping Force =
- Coeff. x Velocity
Stationary I-Beam
Spring Force =
- Stiffness x Displ.
Now consider an expansion upon the concepts of the undamped system by

including another type of vibration absorber. At this point, consider a single concrete block suspended from a rigid I-Beam by a spring, plus an automotive type
shock absorber as shown in Fig. 2-11. Again, the weight is allowed to move in a
vertical direction, and the equivalent spring mass damper system is depicted in
Fig. 2-12. Since damping is now involved, this is a damped mechanical system.
As before, there is no external force applied, and the behavior of this system
must be classified as free vibration when it is allowed to oscillate.
Displacement = +D
Gravity Force =
Mass x Accel.
Fig. 211 Spring Mass

Damper System
Fig. 212 Equivalent Spring Mass Damper Mechanical System And Associated Free Body Diagram
The shock absorber is a viscous damper that displays the fundamental

property of a damping force that is proportional to velocity times the damping
coefficient. This relationship is quite clear when the analogy of an old versus new
shock absorber is considered. Specifically, an old worn out shock absorber will be
quite loose, and the inner rod will move easily in and out of the main body. However, a new shock absorber will be tight, and extension or compression of the
inner rod requires the application of a slow steady force. If an individual
attempts to rapidly move the inner rod, they will find that this higher speed
motion is resisted by a significantly larger force.
38
Chapter-2
Functionally, the shock absorber or damper removes energy from the system. To state it another way, the damper provides the fundamental means of
energy dissipation for the mechanical system. If this physical representation is
converted into a traditional physics free body diagram, Fig. 2-12 evolves. Once
more, the vertical motion must be initiated by an initial disturbance, and the
system now reveals a spring force, a damping force, plus the necessary gravitational term. From this free body diagram, the force balance yields the following
equation of motion for this damped mechanical system:
( Mass Accel ) = ( Stiffness Displ ) + ( Damping Coeff Velocity )
Moving all terms to the left side of the equation, the expression becomes:
( Mass Accel ) + ( Damping Coeff Velocity ) + ( Stiffness Displ ) = 0
Substituting a simpler alpha identification for the six physical variables,
the equation of motion may be stated as:
( M A) + (C V ) + ( K D) = 0
(2-56)
Again, in the manner used by W. T. Thomson, the periodic displacement of

this damped spring mass system may be defined with an exponential function
similar to equation (2-23). If the displacement is identified by D, and the time is
specified by t, and S is a constant that has to be determined, an appropriate
exponential equation would have the following form:
D = e
St
(2-57)
As demonstrated earlier in this chapter, acceleration, velocity, and displacement are integrally related, and equation (2-56) may be rewritten in terms of displacement by the substitution of equation (2-57). Certainly the displacement
term may be inserted directly. The velocity and acceleration terms are obtained
by taking the first and second time derivatives of equation (2-57) to yield the following equation of motion for this damped single degree of freedom system:
2
St
St
St
( M S e ) + (C S e ) + ( K e ) = 0
This expression may be simplified by factoring out the common exponential
term, and dividing by the mass M to yield the next form of the motion equation:
St
2
C
K
S + ------ S + ------ = 0
M
M
(2-58)
As discovered in the undamped case previously discussed, equation (2-58)

may be satisfied for all values of time t, when the following occurs:
2 C
K
- S + ------ = 0
S + ----
M
M
39
This last expression takes the distinctive form of a quadratic equation.

From basic algebra, it is known that this expression may be solved for the constant S in the following traditional manner:
C
C 2
K
S 1, 2 = --------- --------- ------
2M
2M
M
(2-59)
Two solutions are produced ( radical), and the general equation must be
expanded to correspond with this dual root. Hence, the periodic displacement
described by equation (2-57) is redefined in the following manner:
D = Ae
S1 t
+Be
S2 t
(2-60)
Constants A and B depend on how the oscillation was started. The behavior
of a damped system is dependent on whether the radical from equation (2-59) is
real (+), imaginary (-), or zero (0). The simplest case is the zero value for the radical, and this term is defined as critical damping Cc, as follows:
Cc 2
K
-------- ------ = 0
2M
M
(2-61)
By rearranging terms, the following intermediate result is obtained:

Cc 2
K
-------- = ------
2M
M
By taking the square root of both sides of the equation, and substituting
equation (2-44), the following is obtained:
Cc
--------- =
2M
K
------ = c
M
(2-62)
For convenience, a damping ratio of will be defined as the actual damping

C divided by the critical damping Cc as follows:
C
= -----Cc
(2-63)
Combining the damping ratio from equation (2-63), and equation (2-62), the
term C/2M may be reconfigured as:
Cc
Cc
C
C
C
--------- = --------- ------ = --------- ------ = c
2M
Cc
2M
Cc
2M
(2-64)
Based on these derived expressions, the solution to the quadratic equation

(2-59) may now be rewritten as:
40
Chapter-2
2
S 1, 2 = c ( c ) ( c )
If the undamped natural frequency c is factored out of this expression, the

following result is obtained:
2
S 1, 2 = c 1
(2-65)
Interestingly enough, the solution for constants S1 and S2 reveals a relationship between the undamped natural frequency c, and the damping ratio .
The transition between oscillatory and non-oscillatory motion is referred to as
critical damping. For this case, C=Cc, =1, and equation (2-65) simplifies to:
S 1, 2 = c
(2-66)
Substituting this critical damping solution for S1 and S2 back into the general equation of motion, equation (2-60) produces the following result:
D= ( A + B) e
c t
(2-67)
This function contains only one constant (A+B), and the solution lacks the
required number of independent constants to properly represent the general
c t
solution. In this case, an expression in the form of t e

will satisfy the equation. Upon substitution of this new form, the general solution of equation (2-67)
can be correctly written in the following manner:
D= ( A + B t) e
c t
(2-68)
The significance of a critically damped system is depicted in Fig. 2-13. In

this diagram, the displacement of the mass is plotted against time. For numerical simplicity, the constants A and B were assigned values of +10 and -5 respec10
Critically Damped, =1.0
Displacement
Fig. 213 Time Domain

Amplitude Response Of
Critically Damped, And
Overdamped Mechanical
Systems
Constants:
A = 10, B = -5, = 5.0
c
Overdamped, =2.0
2
0
0.0
0.5
1.0
1.5
Time
2.0
2.5
41
tively, and the natural frequency c was set equal to 5.0. The solid line
represents a critically damped system. The resultant motion is aperiodic, and
this critically damped system returns to rest in the shortest time without oscillation (vibration) of the mass. Stated in another way, a critically damped system
contains the minimum amount of damping necessary for aperiodic motion.
If the system damping is greater than critical damping, the system is considered to be overdamped. Conversely, if the mechanical system has less than
critical damping, the system is underdamped, and it will oscillate or vibrate with
time. It should be noted that most process machines are underdamped, and sustained motion of the rotating or reciprocating elements is normal behavior.
For a better understanding of damping, consider an overdamped system
(equivalent to a new shock absorber). In this case, the damping ratio would be
greater than one (>1), and the S terms in the quadratic solution equation (2-65)
may be specified as:
2
S1 = c + 1
2
S2 = c 1
Combining these expressions with the general equation produces the following equation for the motion of an overdamped mechanical system:
D= A e
c t +
+Be
c t
(2-69)
This function describing an overdamped mechanical system is plotted as

the dotted line in Fig. 2-13. The displacement change with time is the sum of two
decaying exponential functions, and system vibration is not maintained. Motion
is aperiodic, and the body returns to rest without oscillation. It is also clear from
this composite diagram that the overdamped system of equation (2-69) does not
return to rest as rapidly as the previously discussed critically damped case.
Finally, consider the situation of a mechanical system with small damping
(equivalent to a worn out shock absorber). This is generally referred to as an
underdamped system where <1, and the radical of equation. (2-65) is imaginary.
The constants S1 and S2 for this condition can be written as follows:
2
S 1, 2 = c 1
(2-70)
Using equation (2-70) to recompute the S terms, and then including these
new expressions into the general equation, the following solution for the equation of motion for a under damped system was presented by W.T. Thomson as:
42
Chapter-2
D= e
c t
Ae
2
c t i 1
+Be
2
c t i 1
This expression may be simplified to a more understandable format as:

D = Y e
c t
2
sin ( c t 1 + )
(2-71)
By inspection, equation (2-71) consists of the superposition of an oscillating

sine wave plus an exponential term. In most cases, the amplitude of the sine
wave is decreased by the exponential function with increasing time. The variable
Y in this equation represents the peak intersection between the exponential
function and zero time. The term is the timing lag between the oscillatory
curve and a zero time starting point. For demonstration purposes, a response
curve for an under damped system is plotted in Fig. 2-14. A displacement value
of 10.0 was assigned to Y, the timing offset was set equal to zero, and a constant value of 5.0 was used for the undamped natural frequency c. This system
exhibits an oscillatory motion with respect to time, and this is referred to as free
vibration of the under damped mechanical system.
10
Underdamped, =+0.1
Constant: c= 5.0
Y2=4.560
-4

An Under Damped
Mechanical System
Y4=1.289
0
-2
Y3=2.425
Y =8.581
Displacement
-6
-8
0
Time
Another interesting point from examining (2-71) is that the term

2
c 1 is multiplied by the time t to determine the number of radians. This
suggests that the undamped natural frequency c is altered by the damping

ratio to produce a new frequency. In fact, this is commonly identified as the
damped natural frequency or damped critical for the mechanical system, and it is
defined in the following manner:
damped
critical
= c 1
(2-72)
43
This is a very important equation because it directly influences the impact

of damping upon a resonance. As shown in the forthcoming Fig. 2-18 for forced
vibration variations in damping produce major changes in amplitude and
phase through a resonance. However, there is also a subtle shift in the resonant
frequency as the damping is varied. Thus, the machinery diagnostician must be
fully aware of the fact that changes in system damping will alter the behavior
through a resonance in the following three ways:
Significant changes in the peak amplitude at the resonance.
Significant variations in the phase angle change across the resonance.
Subtle change in the damped natural frequency (i.e., damped critical speed).
Fig. 2-14 for an under damped system shows that the oscillatory motion
decays with time. Examination of a longer time record would reveal that the
amplitude decrease is actually an exponential decay. The rate of this exponential
decay may be quantified by the log decrement which is defined as follows:
Y1
Y1
1
Log Decrement = = ln ------- = ----- ln ----------------
Y 2
Y N + 1
N
(2-73)
In equation (2-73), Y1 and Y2 represent any two successive amplitudes in

the decaying dynamic signal. The natural logarithm of this ratio defines the
damping as the log decrement . In some cases, particularly with lightly damped
or short duration experimental data, it is necessary to examine multiple cycles of
the decaying signal to determine the log decrement. For these situations, the
right hand side of equation (2-73) may be used. Within this part of the expression, the initial peak amplitude is still specified by Y1, and the amplitude following N number of cycles is identified as YN+1. The validity of this relationship may
be checked by calculating the log decrement for different combinations. For
instance, in Fig. 2-14, the first amplitude peak Y1 has a magnitude of 8.581, and
the second peak Y2 is equal to 4.560. Using the first part of equation (2-73), the
log decrement may be computed from these values in the following manner:
Y1
8.581
Log Decrement = = ln ------- = ln ------------- = ln ( 1.882 ) = 0.632
Y 2
4.560
The same calculation may be performed using the first three cycles in Fig.
2-14. For this case the third peak Y4 has an amplitude of 1.289, and the log decrement may be computed with the right side of equation (2-73) as follows:
Y1
Y1
1
1
Log Decrement = = ----- ln ---------------- = --- ln --------------
Y N + 1
Y 3 + 1
N
3
Y1
1
1
8.581
ln ( 6.657 )
1.896
= --- ln ------- = --- ln ------------- = ------------------------ = ------------- = 0.632
Y 4
1.289
3
3
3
3
The same result of =0.632 has been reached using a single cycle and multi-
44
Chapter-2
ple cycles. Certainly this concept may be extended to the examination of various
decaying dynamic data sets. It should also be mentioned that the log decrement
may also be expressed in terms of the critical damping ratio . It can be shown
that the log decrement is accurately expressed as:
2
Log Decrement = = -----------------2
1
(2-74)
The decaying signal plotted in Fig. 2-14 was produced with a damping ratio
of =0.1. To check the validity of equation (2-74), this damping ratio may be used
to calculate the log decrement as follows:
2
2 0.1
Log Decrement = = ------------------ = ----------------------- = 0.632
2
2
1 0.1
1
Once more the same value for the log decrement has been obtained. This
provides confidence that equations (2-73) and (2-74) are compatible, and consistent. Depending on the available data, one expression may be easier to apply versus the other. Another usable format for these expressions is obtained by solving
equation (2-74) for the damping ratio to produce the following:
Damping Ratio = = ------------------------2

2
4 +
(2-75)
Equation (2-75) is useful for determining the damping ratio based upon
experimental or analytical values of the log decrement. Further examination of
equations (2-74) and (2-75) reveals that the damping ratio and the log decrement are closely related. For instance, the polarity of the log decrement and the
damping ratio must be the same. If the damping ratio is positive, the log decrement must also be positive. Similarly, if the log decrement is negative, then the
damping ratio must be negative. The physical significance of negative damping is
depicted in Fig. 2-15. This diagram is based upon equation (2-71) where Y was
assigned a value of 1.0, the timing offset was equated to zero, a value of 5.0 was
8
Underdamped, =-0.1
Displacement
4
2
0
-2
-4
-6
Constant: c= 5.0
-8

An Unstable Mechanical
System
-10
-12
0
Time
Forced Vibration
45
used for c, and the damping ratio was minus 0.1. It is clear that the oscillatory
motion of this system increases with time, and this is representative of an unstable mechanical system.
In essence, a large positive value for the log decrement is synonymous with
a well damped system. Conversely, a small positive log decrement is indicative of
a stable system with lower damping. The inclusion of the plus sign signifies a
stable mechanical system that will exhibit decreasing amplitudes as a function
of time. The appearance of a negative value for the log decrement defines a system that displays increasing amplitudes with time. This is descriptive of an
inherently unstable system. In this type of mechanical system, the motion will
continue to increase until physical damage, or eventual destruction occurs.
FORCED VIBRATION
The previous sections have addressed free vibration where the motion originates with an initial disturbance, and the energy of the resultant motion is dissipated by damping or friction. When the available energy is completely
removed, the system returns to a rest condition of zero motion.
However, real mechanical systems are influenced by both external and
internal forcing functions. These forces are often periodic, and they provide a
continuous energy input into the system. In this situation, the mechanical system continues to vibrate, and does not return to a state of zero motion. A simple
example of a periodic forcing function can be defined with an expression such as:
Force = F sin ( t )
(2-76)
Equation (2-76) defines a maximum force F that varies periodically (in this
case sinusoidally) as a function of frequency , and time t. This force could be
equated to the summation of damped spring mass forces described by W.T. Thomson in equation (2-56) to provide the following force balance:
( M A ) + ( C V ) + ( K D ) = F sin ( t )
(2-77)
Equation (2-77) states that the weight, damping force, and spring force are
equal to the applied forcing function. It is presumed that the body oscillates at
the frequency of the forcing function, and that the resultant motion is identical
to the forcing function. For this case of forced vibration, assume that the displacement is represented by the following circular function that is similar to the
previous equation (2-11):
Displacement = D = Y sin ( t )
(2-78)
In this equation, Y represents the peak displacement, and is the phase

lag between the applied force and the motion. The first and second time derivatives of equation (2-78) will yield the following two expressions for velocity and
acceleration of the mechanical system:
46
Chapter-2
Velocity = V = Y sin ( t + 2 )
(2-79)
Acceleration = A = Y sin ( t + )
(2-80)
Inserting equations (2-78), (2-79), and (2-80) back into the equation (2-77),
yields the next expression for forced vibration of a spring mass damper system:
2
{ M Y sin ( t + ) } + { C Y sin ( t + 2 ) } +
{ K Y sin ( t ) } = F sin ( t )
(2-81)
This expression defines four vector quantities that remain in a fixed relationship with respect to each other as depicted in Fig. 2-16. These four vectors
rotate together at a constant frequency . The graphical representation of these
four vectors reveals that the stiffness term includes displacement, the velocity
term considers damping, and the acceleration vector incorporates the inertia
force. As shown, the three vibration vectors are mutually perpendicular.
Kx
(Kt
)xY
Cx
xY
Yx
Yx
Cx
2
x
2
x
Fig. 216 Forced Vibration - Primary Vectors
Fig. 217 Simplified Vector Triangle
This is consistent with the vector diagram previously shown in Fig. 2-2. In
both cases, the 90 shifts between displacement, velocity, and acceleration are
clearly evident. Since the displacement and acceleration terms are opposite in
direction, the vector diagram of Fig. 2-16 may be redrawn into the simplified
sketch shown in Fig. 2-17.
The simplified vector diagram contains the same elements as Fig. 2-16.
However, the differential vector between displacement and acceleration terms is
used instead of the separate and opposite vectors. This simplification provides a
right triangle with the applied forcing function as the hypotenuse, the damping
term as one leg of the triangle, and the differential stiffness-inertia vector as the
third leg of the triangle. The maximum displacement Y may be determined from
trigonometry by setting the square of the hypotenuse equal to the sum of the
squares of the other two sides of this right triangle:
2
F = {( K M ) Y } + {C Y }
2 2
F = Y ( K M ) + (C )
Forced Vibration
47
Solving for the displacement Y produces the following equation:

F
Y = ---------------------------------------------------------------------2 2
( K M ) + (C )
(2-82)
The maximum angle between the forcing function and the displacement
may be determined from a trigonometric relationship in the following manner:
C
tan = -----------------------------2
K M
(2-83)
If the numerator and denominator of equations (2-82) and (2-83) are

divided by the stiffness K, the following intermediate results are obtained. Once
again, these equations are consistent with the approach used by W.T. Thomson.
F
--- K
Y = ---------------------------------------------------------------------2 2
C 2
M
1 ------------------ + --------------
K
K
(2-84)
C
--------------
K
tan = --------------------------------2
M
1 -----------------
K
(2-85)
Equations (2-84) and (2-85) may be further simplified by converting these

expressions into a non-dimensional format. This can be accomplished by incorporating the following previously defined terms:
From equation (2-44) the system natural frequency = c = K M
From equation (2-63) the damping ratio = = C/Cc
From equation (2-62) the critical damping coefficient = Cc = 2 Mc
In addition, the following two new non-dimensional terms will be defined,
and included in this conversion:
Critical Speed Frequency Ratio = = -----c
(2-86)
Y
Amplitude Ratio = -----Yc
(2-87)
Where Yo is the zero frequency deflection of the spring mass damper system
due to the application of the periodic forcing function F. In this context, Yo is
equal to F/K in equation (2-84). By inserting the previously defined non-dimen-
48
Chapter-2
sional expressions into equations (2-84) and (2-85) the following are obtained:
Yo
Y = ---------------------------------------------------------------------2 2
1
-------- + 2 ------
2
c
2 ----
c
tan = -----------------------------2

1 --------2-
Finally, by incorporating equations (2-86), and (2-87) into the above expressions, the non-dimensional Amplitude Ratio, and Phase Angle are described by
the resulting equations (2-88) and (2-89).
1
Amplitude Ratio = ---------------------------------------------------------------2 2
(1 ) + (2 )
(2 )
tan = ---------------------------2
(1 )
(2-88)
(2-89)
From these two expressions, the amplitude ratio and phase angle are
reduced to functions of the damping ratio , and the critical speed ratio . This is
an expected result since it is common knowledge that vibration amplitudes in
the vicinity of a resonance are directly related to the frequency offset from the
center frequency of the resonance. Furthermore, the damping ratio should also
participate in determining the response characteristics since it is the fundamental indicator of energy dissipation.
This relationship is easier to understand in the graphic format of Fig. 2-18.
Within this diagram, a family of five curves are plotted over a critical speed frequency ratio between 0 and 3.0, and a damping ratio ranging from an underdamped system at =0.1 to an overdamped condition of =2.0. This general type
of data display is referred to as a response plot or a Bode plot. In most cases, the
synchronous 1X vibration amplitude and phase are plotted against rotative
speed instead of the non-dimensional values used for Fig. 2-18. However, the concepts of tracking synchronous amplitude and phase as a function of speed are the
same for a dimensional or a non-dimensional system.
In either case, an amplitude increase occurs at the natural or critical frequency (=1) of the system. This amplitude response is coincident with a substantial phase shift at the critical speed. It is also apparent that the magnitude
of the amplitude response, and the amount of the phase shift are both directly
related to the system damping.
Forced Vibration
49
Phase Lag (Degrees)
Low Damping, =0.1

30
60
90
High Damping,
=2.0
=2.0
120
=1.0
150
=0.2
=0.5
=0.1
180
5
Low Damping, =0.1
Amplitude Ratio
4
=0.2
3
2
=0.5
1
Fig. 218 Calculated

Bode Plot Of Forced
Response For A Simple
Mechanical System
=1.0
=2.0
0
0.0
0.5
1.0
1.5
2.0
2.5
Critical Speed Frequency Ratio ()
3.0
Specifically, an under damped mechanical system (e.g., =0.1) exhibits a

large amplitude change through the resonant frequency, combined with a phase
shift that approaches 180. Conversely, a heavily damped system (e.g., =2.0) displays no amplitude change through the critical speed region, and only a minor
variation in phase angle to reveal the presence of the natural frequency.
It is meaningful to recall that the forcing function for equation (2-88) was
based upon a constant driving force. That is, the same force is applied to the system at every frequency. In actuality, forced vibration is often excited by a force
that varies with machine speed. For instance, centrifugal force due to unbalance
varies as the square of the speed. Hence, it is desirable to consider revising the
amplitude ratio of equation (2-88) to include a forcing function that incorporates
an 2 term. This would simulate the relationship between the amplitude
response, and the mass unbalance forcing function. In the simplest format, it can
be shown that the amplitude ratio for an unbalance response is described by the
next equation that includes the speed squared 2 term in the numerator:
2
Amplitude Ratio = ---------------------------------------------------------------2 2
(1 ) + (2 )
(2-90)
For this forced unbalance response, the phase relationship remains identi-
50
Chapter-2
cal to the previously presented equation (2-89). If the amplitude ratio and phase
for a forced unbalance condition are plotted as a function of the critical speed
ratio, Fig. 2-19 emerges. Note that the damping relationship remains consistent
with the previous discussion, and the amplitude exhibits the most significant
change. In this forced unbalance case, the amplitude ratio at low speeds
approaches zero due to a small driving force (i.e., low unbalance force). At the
critical speed of =1, the magnitude of the peak is governed by the damping. At
frequencies above the critical speed, the amplitude and phase remain fairly constant for each value of damping ratio. Hence, above a resonance, it is normal to
encounter a plateau region where synchronous amplitude and phase remain reasonably constant with increasing machine speed. This behavior will be demonstrated with actual examples of machinery vibration data throughout this text.
0
Phase Lag (Degrees)
Low Damping, =0.1

30
60
High Damping,
=2.0
=2.0
90
120
=1.0
150
=0.2
=0.5
=0.1
180
5
Low Damping, =0.1
Amplitude Ratio
4
=0.2
3
2
=0.5
Fig. 219 Calculated

Bode Plot Of Unbalance
Mechanical System
0
0.0
=1.0
0.5
1.0
1.5
=2.0
2.0
2.5
3.0
A calculated or a measured Bode plot is excellent for observing rotational

speed vibration amplitude and phase as a function of speed (frequency ratio).
However, there are conditions when a detailed examination of phase changes is
required. Fortunately, the same vector information that is used to construct a
Bode plot may also be viewed in terms of polar coordinates. In this type of data
presentation, the vectors (amplitudes and angles) at each speed (or speed ratio)
are plotted in a polar format. The heads of the vectors are then connected to form
a continuous line. This type of diagram is generally referred to as a polar plot,
Forced Vibration
51
and it has many applications within the domain of machinery malfunction analysis. A typical example of this type of data is shown in Fig. 2-20. In this diagram,
the conditions of low damping (i.e., =0.1 and 0.2) were extracted from Fig. 2-19,
and these curves were replotted in a polar coordinate format. The peak of the
resonance occurs at =1, which is coincident with a 90 phase angle shift. As in
the Bode plot, the high speed condition is identified as =3.0.
0
5.0
Amplitude Ratio
330
300
30
4.0
3.0
60
2.0
=0.2
1.0
0.0
270
90
=1.0
=3.0
240
Fig. 220 Calculated Polar

Plot Of Unbalance Response
For A Simple Mechanical
System
120
Low Damping
=0.1
210
150
180
Vector angles on the polar plot are always plotted against rotation. This
phase lag logic is directly associated with the vibration measurement systems
used to analyze the machinery behavior. In virtually all cases, it is highly desirable (if not mandatory) to generate data plots that are physically representative
of the machinery geometry. This topic will be discussed in much greater detail in
the subsequent chapters 3, 5, 6, 7 and 11.
Before leaving the calculated Bode and polar plots for this forced response
of a spring mass damper system, the magnitude of the vibration at the natural
frequency should be examined in greater detail. From the previous definitions, it
is clear that the critical speed frequency ratio is equal to unity (/c=1) at the
natural frequency. Substituting this value of =1 into either equations (2-88) or
(2-90) yields identical results. Specifically, when a value of =1 is placed into
equation (2-88), the following result is obtained:
1
Amplitude Ratio = ---------------------------------------------------------------2 2
(1 ) + (2 )
1
1
Amplitude Ratio = ------------------------------------------------------------- = ---------------------------2
2
2
2
0+4
(1 1 ) + (2 1)
52
Chapter-2
Thus, at the natural resonance of =1, the Amplitude Ratio is reduced to:
1
Amplitude Ratio at Resonance = -----------2
(2-91)
The amplitude ratio for this simple system is a function of the critical
damping ratio . This relationship is easily tested by running some trial values.
For instance, if the damping ratio =0.1, then the amplitude ratio is equal to 5.0.
Similarly, if the damping ratio =0.5, then the amplitude ratio is equal to 1.0.
These values are consistent with the plots presented on both Figs. 2-18 and 2-19.
The amplitude ratio computed with equation (2-91) is a useful quantity
that is often referred to as the Amplification Factor for the mechanical system.
This amplitude ratio, or amplification factor is also called the Q for the resonance. This dimensionless quantity provides a way to describe the severity of a
particular resonance, or the magnitude of the damping ratio at a resonance. In
all cases, a high Q is indicative of a system with minimal damping that exhibits
large amplitudes at the peak of the resonance. Systems with low available damping may be easily excited, and may be susceptible to stability problems due to a
lack of available system damping.
Conversely, systems with a small value for Q must be well damped by definition, and this type of system will exhibit low vibration amplitudes at the resonant frequency. Systems with higher damping will be more difficult to excite, and
will be less susceptible to a variety of stability problems.
The amplification factor for a rotating machine passing through a specific
resonance (critical speed) may be evaluated analytically from the damped critical
speed calculations discussed in chapter 5. Based upon the real and imaginary
portions of the complex Eigenvalue, the log decrement may be computed. The
amplification factor for the resonance is determined by dividing by the calculated log decrement.
From a measurement standpoint, various methods are used to determine
the amplification factor based upon the vibration response data of a machine
passing through the rotor resonance. A comparison of the three traditional methods are reviewed in chapter 3.
The theoretical model of mechanical system behavior is closely matched by
the motion of actual rotating machines. For example, Fig. 2-21 depicts a Bode
plot of a high speed compressor rotor mounted between bearings. In this diagram, the rotational speed in RPM is plotted on the horizontal axis, with synchronous 1X amplitude and phase lag presented on the dual vertical axes.
This field vibration data was measured with a shaft sensing proximity
probe mounted close to the coupling end journal bearing. From this plot it is
apparent that a resonance occurs at a speed of 6,100 RPM. This response is the
first critical speed of the rotor. Amplitude response through this resonance is
moderate, and the overall phase roll through the critical is approximately 110.
This data indicates that the mechanical system is underdamped, with a damping
ratio in the vicinity of =0.2. This type of transient speed behavior is normal and
customary for many types of machines within the process industries.
Forced Vibration
Fig. 221 Measured Bode Plot Of Actual

Industrial Centrifugal Compressor
53
Fig. 222 Measured Polar Plot Of Actual

Industrial Centrifugal Compressor
The same synchronous 1X vector vibration data may be plotted in the polar
coordinate format of Fig. 2-22. This data is identical to the Bode plot, but the
polar presentation provides improved resolution of phase changes. As viewed
from the driver end of the train, this machine rotates in a counterclockwise
direction. The angular reference system begins with 0 at the probe, and the
phase angles increase in a direction that is counter to shaft rotation. This is the
correct angular convention, and the specific logic for this phase convention will
be discussed in succeeding chapters. Note that the origin of the polar plot is coincident with low speed, and the plot evolves in a clockwise direction as speed
increases. Again, this is normal behavior for a rotor resonance where the phase
continues to increasingly lag as the unit passes through the critical speed (balance resonance) region.
Generally, both Bode and the polar plots are required to accurately define
resonances. This is applicable to rotor, structural, and secondary resonances.
Although a Bode plot will allow accurate frequency identification of the critical,
the type of resonance is often identified by the polar plots along the length of the
rotor. A proper understanding of these transient speed plots is vital to a full comprehension of the transient vibration behavior of the machinery.
It should also be mentioned that this data is sensitive to any type of vector
offset. This is particularly true for shaft measurements made with proximity
probes. These displacement transducers are susceptible to shaft surface conditions such as scratches, surface imperfections, metallurgical variations, magnetized segments, and eccentricity of the observed shaft surface. These types of
conditions produce erroneous signals that often appear as a substantial 1X vec-
54
Chapter-2
tor at low speeds. For demonstration purposes, the transient data previously
shown in Fig. 2-21 is replotted in Fig. 2-23 with the inclusion of a constant 1X
runout vector of 0.83 Mils,p-p at 168.
Fig. 223 Measured Bode

Plot Of Actual Industrial
Centrifugal Compressor
With Shaft Runout
The solid lines on this Bode plot are identical to the data in Fig. 2-21, and
the dotted lines for amplitude and phase show the influence of the 0.83 Mil,p-p
runout vector. Clearly this slow roll vector influences the vibration signal
throughout the entire speed domain. The uncompensated critical speed peak
appears at 5,900 RPM instead of the actual resonance speed of 6,100 RPM.
Vibration amplitudes at the operating speed of 9,500 RPM appear as 0.4 Mils,p-p
with runout, versus the true magnitude of 0.95 Mils,p-p. Finally, the phase shift
through the resonance with runout included is about 50, whereas the properly
compensated vibration signal displays a more realistic 110 phase change
through the resonance region.
It is apparent that the inclusion of a slow speed runout vector can result in
serious data interpretation problems. Due to the potential implications of shaft
runout, the origin and various corrective measures for shaft runout will be discussed in greater detail in subsequent chapters.
Overall, the relationship between the physical parameters of mass, stiffness, damping and the motion of a body including the displacement, velocity,
acceleration, and frequency have been established in this chapter. When these
fundamentals are clearly understood, complex mechanical vibration problems
may be addressed, and successfully solved.
The previously discussed resonant response is quite typical for a piece of
rotating machinery. In this common behavior, a rotor resonance is excited by a
synchronous unbalance force during transient startup, and coastdown conditions. The coincidence of the excitation frequency (rotor speed) with the natural
Forced Vibration
55
frequency (critical speed) results in a blossoming or amplification of the vibration

amplitudes. It is also apparent that during normal operating conditions, the
rotor resonance is normally dormant (inactive) due to the lack of any appreciable
exciting force within the bandwidth of the resonant frequency.
In actuality, a variety of rotor and structural resonances exist on every
machinery train. In most instances, the various major and minor resonances are
inactive due to the lack of an appropriate stimulation. However, when a machine
emits a discrete excitation that is coincident with a natural frequency, that particular natural frequency will become active. Similarly, when a process machine
produces, or is subjected to a wide-band excitation, the resonant frequencies
within the excitation bandwidth may likewise become active.
Case History 2: Steam Turbine End Cover Resonance
As an example of an excitation of a normally inactive resonance, consider a
high pressure steam turbine driving a series of tandem centrifugal compressors.
The operating speed for this train normally varied between 3,500 and 3,750
RPM. The eleven stage turbine is rated at 49,000 HP, and it had a history of low
vibration amplitudes with minimal evidence of any abnormalities.
Prior to a maintenance turnaround, plant rates were increased to maximize
production going to storage. Under this operating condition, the machine speed
was temporarily increased to 3,800 RPM. Since the rated speed for this train was
3,930 RPM, the moderate speed increase was considered to be well within the
performance envelope for this unit. Unfortunately, continuous operation at this
higher speed resulted in an objectionable governor housing vibration. This axial
casing vibration occurred at a frequency of 15,200 CPM.
With field measurements, it was readily determined that the high vibration
amplitudes were confined to the bolted end cover of the governor housing. The
maximum response was measured at the middle of the cover, and minimal
amplitudes were evident within the bolt circle of the cover. It was clear that the
measured frequency occurred at four times machine speed (4X), and it was discovered that a slight speed reduction resulted in the virtual elimination of the
end cover vibratory motion. Naturally, the operations personnel were against
any corrections that might reduce the plant production rates.
Hence, a solution other than slowing down the turbine was required. In an
effort to determine the mode shape of the cover plate, a series of additional measurements were obtained. A grid pattern was established on the cover plate face
with approximately 5 inches between grid lines. Vibration readings filtered at
four times running speed were obtained at each grid intersection. In addition, a
reference accelerometer was located at the middle of the cover plate, and the timing relationship to each intersection measurement point was visually determined on a digital oscilloscope.
In all cases, the readings were directly in phase, and the cover was moving
back and forth in unison. The accelerometer field measurements were manually
converted to casing displacement at the average frequency of 15,200 CPM. Plotting the amplitude at each grid intersection allows the construction of the three-
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0.0
0.0
0
Fig. 224 Three-Dimensional Mode Shape Of

Vertical Steam Turbine
End Cover at 15,200 CPM
Amplitude (Mils,p-p)
Chapter-2
56
10
Cov
1
er W 5 20
idth
(Inch 25
es)
30
2
s)
nche
I
(
h
t
5
g
n
e
er L
Cov
10
300
25
15
dimensional plate mode shape shown in Fig. 2-24.

This diagram suggests a simple drum mode, with maximum displacement
at the center of the plate, and minimum motion at the plate perimeter. Since this
appears to be a resonant plate, and since it behaves like a structural resonance;
it would be desirable to confirm this hypothesis based upon the physical characteristics of the plate. An affirmation of this resonance may be obtained by performing a Finite Element Analysis (FEA). Alternatively, since this cover plate is
a simple geometric structure, a set of manual calculations may be performed.
For instance, Roarks Formulas for Stress & Strain10, includes Table 36 of
Natural Frequencies of Vibration for Continuous Members. Case number 15, on
page 717, within this table describes the equation for a rectangular flat plate
with uniform thickness, and all four edges fixed. The natural frequency for this
geometric configuration is presented as:
K
DG
F = ------ -----------------4
2
Wa
where:
F
a
K
G
W
D
=
=
=
=
=
=
(2-92)
Plate Natural Frequency (Cycles / Second)

Plate Width and Length (Inches)
Constant based upon ratio of Length over Width. For a square plate, K = 36.
Acceleration of Gravity (= 386.1 Inches / Second2)
Plate Unit Weight (Pounds / Inch2)
Plate Flexural Rigidity (Pounds-Inch)
The turbine end cover was physically measured to be 30 inches square, with
a thickness of 0.625 inches. The cover was attached to the turbine casing by a
series of bolts that were located on a centerline of approximately 28.5 inches
10 Warren C. Young, Roarks Formulas for Stress & Strain, 6th edition, (New York: McGrawHill Book Co., 1989), pp. 714-717.
Forced Vibration
57
square. Since the bolt pattern represented the point of attachment, and the zero
motion perimeter, the cover dimension of 28.5 inches was used for the plate calculations. The plate unit weight W is computed by multiplying the plate density
times the plate thickness t as follows:
W = t
where:
(2-93)
= Plate Density = 0.283 Pounds / Inch3

t = Plate thickness = 0.625 Inches
The plate unit weight may now be computed from equation (2-93) as:
3
W = t = 0.283 Pounds/Inch 0.625 Inches = 0.177 Pounds/Inch
From Roark, page 714, the plate flexural rigidity D is determined from the
modulus of elasticity E, Poissons ratio , and the plate thickness t as presented
in equation (2-94):
3
Et
D = ------------------------------2
12 ( 1 )
where:
(2-94)
E = Modulus of Elasticity (= 29.5 x 106 Pounds / Inch2)

= Poissons Ratio = 0.28 (Dimensionless)
Based on the plate thickness t, and the material constants (E and ), the
Flexural Rigidity D may now be determined with equation (2-94) as:
3
Et
( 29.5 10 Pounds/Inch ) ( 0.625 Inches )
D = ------------------------------- = -----------------------------------------------------------------------------------------------------2
2
12 ( 1 )
12 ( 1 0.28 )
6
7.202 10
D = -------------------------- = 651, 200 Pounds-Inch
11.059
From these computations of the unit weight W, and the flexural rigidity D,
it is now possible to compute the natural frequency of the turbine cover plate. As
previously noted, the constant K in the natural frequency equation is equal to 36
due to the equal length of the sides. Combining the various values, the natural
frequency is computed from equation (2-92) in the following manner:
2
K
36
DG
( 651, 200 Pounds-Inch ) ( 386.1 Inches/Second )
F = ------ ------------------ = ------ ------------------------------------------------------------------------------------------------------------------4
4
2
2
2
Wa
( 0.177 Pounds/Inch ) ( 28.5 Inches )
18
251, 428, 000
Cycles
Seconds
Cycles
F = ------ ----------------------------------------- = 265.8 ----------------- 60 -------------------- = 15, 950 ----------------2
Second
Minute
Minute
116, 776 Second

This calculated value of 15,950 CPM is higher than the measured frequency
of 15,200. However, the 750 CPM difference represents only a 5% deviation. Considering that the material physical properties were estimated, and the plate
58
Chapter-2
dimensions were not precise, the 5% variation in frequencies is understandable.

Thus, the measured and calculated data point towards a plate resonance as
the culprit. This was verified during operation by bolting a beam with a jackscrew across the plate. Application of the screw to the center of the cover plate
immediately suppressed the plate resonant response. During the next scheduled
turnaround, the temporary beam and jackscrew were removed, and two angle
beams were welded across the inside of the cover plate. Subsequent operation at
speeds up to 3,950 RPM revealed no reoccurrence of this problem.
In retrospect, this was not a serious mechanical problem, but it was an irritant to the Operations personnel since the vibratory response was audible on the
compressor deck. Hence, sometimes problems must be corrected to satisfy the
mechanics as well as the politics of the situation.
TORSIONAL VIBRATION
The previous sections of this chapter have considered the vibration of a
mechanical system in general, and the lateral vibration of a rotating machine in
particular. These same concepts are also applicable to the torsional behavior of a
mechanical system. In fact, an analytical simulation of a new machine usually
considers both the lateral and the torsional response of the proposed rotor system. Fortunately, the same concepts and equations developed for longitudinal
vibration may be directly converted to torsional motion. For example, the four
basic torsional vibration parameters are defined as follows:
Dtor = Torsional Displacement Degrees,peak to peak = Deg,p-p
Vtor = Torsional Velocity Degrees/Second,zero to peak = Deg/Sec,o-p
Ator = Torsional Acceleration Degrees/Second2,zero to peak = Deg/Sec2,o-p
F
= Frequency Cycles/Second (Hz)
The torsional displacement Dtor is the angular twist due to a natural

oscillation, or an applied torque. The first time derivative of torsional displacement yields the torsional velocity Vtor. Similarly, the second time derivative of
torsional displacement provides torsional acceleration Ator. These quantities are
identical to their lateral counterparts, and a 90 timing shift exists between torsional vectors. As before, measurement of one torsional vibration amplitude and
frequency will allow computation of the other two torsional vibration parameters. For instance, replacing 2F for the frequency , and considering the peak
values of the terms, equation (2-12) may be restated for a torsional system as:
D tor
D tor
Radians
Cycles
V tor = ------------ = ------------ Degrees 2 ------------------- F -----------------
2

Cycle
Second
2
Torsional Vibration
59
Simplification produces the following equation:

V tor = D tor F
(2-95)
Next, consider the relationship between torsional acceleration and displacement as originally described by equation (2-13), and modified as follows:
2
D tor
D tor
2
Radians
Cycles
A tor = ------------ = ------------ Degrees 2 ------------------- F -----------------
2

Cycle
Second
2
Which simplifies to the following equation:

2
A tor = 2 D tor F
(2-96)
A similar relationship exists between torsional acceleration and velocity as:

Degrees
Radians
Cycles
A tor = V tor = V tor ------------------- 2 ------------------- F -----------------
Second
Cycle
Second
Which is equal to the following expression:

A tor = 2 V tor F
(2-97)
These last three equations allow conversion between torsional displacement, velocity, and acceleration at a fixed frequency in Cycles per Second. For
those unaccustomed to torsional vibration units, these expressions, and the
resultant amplitudes may appear to be unusual. It should be noted that torsional
displacement amplitudes are typically much smaller than one degree.
For demonstration purposes, consider the conversion of torsional vibration
units on vibration data acquired on a 28,000 horsepower speed increasing gear
box. In this case, a proximity probe was used to measure torsional displacement
from the bull gear teeth. At the same time, a direct coupled torsiograph (velocity
based torsional pickup) was attached to the outboard blind end of the gear. Operating at 100% load with a speed of 5,100 RPM (85 Hz), the proximity probe system revealed a torsional displacement of 0.0021 Degrees,p-p. Simultaneously, the
torsiograph system indicated a torsional velocity of 0.567 Degrees/Second,o-p.
The installation of both types of torsional sensors was directed at providing
a verification of the data by using two different and independent measurement
systems. Hence, this steady state data can be directly compared if the units of
one measurement are converted to the units of the other measurement. Using
the previous equation (2-95), the measured torsional displacement may be converted to torsional velocity as follows:
Cycles
Second
V tor = D tor F = 0.0021 Deg o-p 85 --------------- = 0.561 Deg/Sec o-p

The measured torsional displacement of 0.0021 Degrees,p-p at 5,100 RPM
(85 Hz) is thereby equal to 0.561 Degrees/Second,o-p. This converted torsional
60
Chapter-2
displacement amplitude is in excellent agreement with the directly measured

torsional velocity of 0.567 Degrees/Second,o-p.
Torsional behavior of a mechanical system parallels the equations developed for a lateral system. Torsional vibration measurements are complex signals
that usually require some type of post processing. The synchronous 1X rotational
speed data is generally viewed as a vector quantity, and the concepts of springs,
dampers, masses, and forcing functions are fully applicable. The names and
units have changed somewhat, but the structure of the equations of motion
remain the same.
The lateral stiffness K is equivalent to the torsional stiffness Ktor. The traditional units used for torsional stiffness are Inch-Pounds/Radian. Multiplication
of the torsional displacement Dtor in Degrees, times the torsional stiffness Ktor,
and a conversion of units yields the torsional stiffness term for the general equation of motion. This term will carry the final torque units of Inch-Pounds.
Lateral damping C is analogous to torsional damping Ctor. The customary
units used for torsional damping are Pound-Inch-Second/Radian. Multiplication
of the torsional velocity Vtor in Degrees/Second, times the torsional damping Ctor,
and a conversion of units yields the torsional damping term for the general equation of motion. This term will carry the final torque units of Inch-Pounds.
Mass M in a lateral system is the comparable of the mass polar moment of
inertia Jmass in a torsional system. The correct engineering units for polar inertia are Pound-Inch-Second2/Radian. Multiplication of the torsional acceleration
Ator in Degrees/Second2, times the inertia Jmass, and a conversion of units produces the torsional inertial term for the general equation of motion. This inertial
term will also carry the torque units of Inch-Pounds.
A periodic forcing function Fsin(t) in a lateral system is the equivalent of
an oscillating torque Tsin(t) in a torsional system. The traditional units used
for torque are either Inch-Pounds, or Foot-Pounds. For the sake of completeness,
it should be recognized that shaft torque in a process machine is directly associated with the shaft speed, and the transmitted power. The classical definition of
power is the time rate of doing work or delivering energy. The inclusion of time
into this definition is very important. For instance, a small motor can provide
any amount of energy given a sufficient amount of time. However, producing a
large amount of energy within a short period of time requires a large motor.
In a linear system, power is the product of force and velocity. Within a rotating system, power is determined by multiplying torque times speed. In either
case, the expected engineering units for power would be in the format of FootPounds/Minute or Inch-Pounds/Minute. Although these units are consistent with
the previous discussion, these units are generally not used. In most instances,
power is identified in units of horsepower for nonelectrical equipment, and units
of kilowatts or megawatts for electric machinery. The traditional conversion
expressions for computing steady state torque from both types of power units are
presented in the following equations (2-98) and (2-99):
Torsional Vibration
61
5, 252 Horsepower
Torque ( Foot-Pounds ) = --------------------------------------------------------RPM
(2-98)
7, 043 Kilowatts
Torque ( Foot-Pounds ) = -------------------------------------------------RPM
(2-99)
The conversion factor between kilowatts and horsepower is 1.341, and that
value provides the difference between the constants shown in equations (2-98)
and (2-99). These expressions also define the relationship between speed, power,
and torque. Thus, a machine at constant speed will exhibit periodic torque variations if the load or power oscillates. Conversely, if the power level remains constant, variations in operating speed (e.g., due to governor malfunction) will
manifest as variations in torque.
The generalized equation of motion for a lateral single degree of freedom
system subjected to a periodic forcing function was presented in equation (2-77).
This expression equated the periodic forcing function to the summation of the
spring, damping, and mass forces. This previous equation is also appropriate for
a torsional system, and the following modification of equation (2-77) is presented
to describe a forced torsional system:
{ J mass A tor } + { C tor V tor } + { K tor D tor } = T sin ( t )
(2-100)
Equation (2-100) will respond in exactly the same way in a lateral or a torsional system. The same characteristic family of curves may be developed for
amplitude ratio as a function of frequency ratio for various damping ratios.
Although the physical system description has changed, the mathematical structure remains constant. Another proof of the mathematical similarity between
lateral and torsional mechanical systems resides in a comparison of the engineering units applied on both types of systems. For example, Table 2-1 summarizes several mechanical parameters, and typical engineering units that may be
used for a lateral, and a torsional system. From Table 2-1, it is noted that the lateral vibratory measurements are based on distance, versus an equivalent angular twist in the torsional analysis. The stiffness and damping coefficients are
formulated in a similar manner with lateral displacement replaced by angular
twist and the applied lateral force is replaced by a torsional torque. Momentum or Impulse in a torsional system includes the moment arm, and work or
energy carries the same units in either system.
Although the lateral and torsional systems display identical equation structures, the diagnostician must be aware of the physical differences between the
two types of motion. For instance, torsional vibration problems are often transient excitations that drive one or more torsional resonances (e.g., during synchronous motor startup). It is unusual to encounter self-excited steady state
torsional behavior. Although lateral vibration produces shaft deflection, torsional
vibration results in shaft stress reversals. Hence, when a torsional problem is
encountered, it is often characterized by mechanical failures (e.g. broken shaft).
To gain a better understanding of this type of twisting behavior, consider a
simplification of the complex torsional system into an undamped system sub-
62
Chapter-2
Table 21 Comparison Of Units For Lateral Versus Torsional Mechanical Systems

Parameter
Lateral
Torsional
Displacement
Mils p p
Degrees p p
Velocity
Inches
----------------- Second o p
Degrees
---------------------
Second o p
Inches
------------------- Second 2
Degrees
---------------------
Second 2
Acceleration
op
op
Stiffness
Pounds
------------------Inch
Inch Pounds
------------------------------------Radian
Damping
Pound Seconds
-------------------------------------------Inch
Pound Inch Seconds

-------------------------------------------------------------Radian
Mass Inertia
Pound Second
-------------------------------------------Inch
Pound Inch Second

-------------------------------------------------------------Radian
Force Torque
Pounds
Inch Pounds
Momentum or Impulse
Pound Seconds
Pound Inch Seconds
Work or Energy
Pound Inches
Pound Inches
jected to free vibration. The diagram in Fig. 2-25, describes such a mechanical
system consisting of a disk supported by a circular rod. The top of the rod is rigidly attached to the stationary I-Beam, and the rod has a torsional stiffness of
Ktor. The disk connected to the bottom of the rod has a mass polar moment of
inertia identified as Jmass. Clearly, the stiffness and mass properties exhibited by
the lateral spring mass system described in the earlier Fig. 2-7 are analogous to
this simple undamped torsional system. It is intuitive that the torsional system
shown in Fig. 2-25 will remain at rest until an initial disturbance is applied. In
this case, if a torque, or twist is applied to the disk, the system will oscillate in
accordance with the following simplification of equation. (2-100):
{ J mass A tor } + { K tor D tor } = 0
Since the system has no damping, the torsional velocity term must be equal
to zero. Also, the system is excited with an initial angular displacement, and a
continuous forcing function does not exist. Thus, the forced Tsin(t) term is also
equated to zero. The resultant expression may be rearranged as follows:
K tor
A tor + ---------------- D tor = 0
J mass
(2-101)
Torsional Vibration
63
Stationary I-Beam
Angular
Displacement
Torsional
Stiffness = Ktor
Fig. 225 Undamped

Simple Torsional System
Mass Polar Moment of Inertia = Jmass
This torsional expression is identical to the lateral equivalent presented in

equation (2-42). Performing the same manipulation on equation (2-101) for the
torsional system, the following expected solution evolves:
c
tor
K tor
---------------J mass
(2-102)
Another common form of this expression is obtained by converting the rotational frequency c units of radians per second to cycles per second in accordance
with equation (2-2) to yield the following:
Fc
tor
K tor
1
= ------ ---------------2
J mass
(2-103)
This solution is the torsional natural frequency for the undamped disk
hanging by a torsional spring (i.e., rod). As discussed in the lateral system, this
natural frequency is a function of the torsional stiffness, and the mass polar
moment of inertia. It is clear that following an initial angular disturbance, the
disk will oscillate at this natural frequency, and the angular motion will gradually decay as a function of time.
From the previous discussion, it is apparent that the equation structure for
lateral and torsional systems are virtually identical. However, the issue of measured torsional vibratory behavior should also be addressed. For machines such
as reciprocating engines, pumps, or compressors, a synchronous torsional component will generally be present under all operating conditions. These types of
machines often exhibit strong torsional resonances, and the clear identification
of resonant frequencies is generally easily achieved.
However, on centrifugal machines, the torsional characteristics are generally quite small, and many of the torsional resonances are well damped by the
64
Chapter-2
process fluid. In addition, the torsional signals are often cluttered with extraneous excitations. Due to the sensitivity of the measurement, the influence from
adjacent machines, electrical interference at 60, 120 and 180 Hz, transducer support resonances, and machining imperfections on the observed elements all contribute to noise in the torsional signal. Hence, the machinery diagnostician must
be very careful during data acquisition and reduction. Special attention should
be paid to the signal to noise ratio of the entire measurement system and the
use of low-pass and/or high-pass filters, plus a synchronous tracking filter may
be necessary. These signal processing considerations will be discussed in much
greater detail in the following chapters 6, 7, and 8 of this text.
Torsional vibration data may be steady state information acquired at a constant speed, and load. Data may also be obtained during transient conditions of
machinery train startups or coastdowns. In most cases, the data processing techniques used for lateral vibration measurements may be successfully used for torsional data. However, there are significant differences in the vibration
transducers used for lateral and torsional measurement. In most instances, the
torsional measurements are delicate, and they must be handled carefully. For
additional information on the characteristics of common transducers, and their
signal conditioning requirements, chapters 6 and 7 should be reviewed.
An example of transient torsional vibration data is shown in the Bode plot
of Fig. 2-26. This data depicts the transient response of a large, single helical,
bull gear mounted in a speed increasing gear box. Note the clearly defined resonant response at 2,050 RPM, and the associated 160 phase shift. This data is
analogous to the lateral response Bode plot previously presented in Fig. 2-21.
Fig. 226 Measured Torsional Bode Plot Of A Bull

Gear In A Speed Increasing Gear Box
In the overview, lateral and torsional equations of simple systems are

essentially identical. It is also clear that strong similarities exist between lateral
and torsional vibration response characteristics. However, it must be understood
that lateral and torsional measurements represent two entirely different modes
of machinery behavior. The mechanical implications of torsional versus lateral
vibration are quite different. For instance, torsional criticals are usually gov-
Torsional Vibration
65
erned by coupling stiffness, and the major mass polar moment of inertia as
described by equation (2-102). Lateral criticals on the other hand are generally
controlled by the support and/or shaft stiffness combined with the rotor mass as
described by equation (2-44). In most cases, the stiffness of the coupling torque
tube (spacer) is significant to torsional behavior, whereas it is often inconsequential to the lateral response.
Furthermore, torsional vibration may not be detected by lateral vibration
sensors, but lateral vibration may adversely influence the accuracy of torsional
vibration measurements. It has been documented that a torsional failure may be
in progress, and the machine might continue to exhibit low radial vibration
amplitudes until a major component failure (e.g., broken shaft). For this type of
machinery situation, the cross-coupling between torsional and lateral vibration
may be virtually nonexistent. However, in other machinery trains, lateral and
torsional modes are closely coupled so that vibration of a lateral resonance may
excite a coupled torsional resonance, or vice versa. When in doubt, both the lateral and torsional characteristics of the machinery should be carefully examined
as a routine part of the problem solving investigation.
66
Chapter-2
BIBLIOGRAPHY
1. Avallone, Eugene A. and Theodore Baumeister III, Marks Standard Handbook for
Mechanical Engineers, Tenth Edition, pp. 2-36, 3-15, New York: McGraw-Hill, 1996.
2. Den Hartog, J.P., Mechanical Vibrations, 4th edition, p. 87, New York: McGraw-Hill
Book Company, 1956.
3. Raynesford, John D., Use Dynamic Absorbers to Reduce Vibration, Hydrocarbon
Processing, Vol. 54, No. 4 (April 1975), pp. 167-171.
4. Shigley, Joseph E. and Charles R. Mischke, Standard Handbook of Machine Design,
pp 11.5-11.6, New York: McGraw-Hill Book Company, 1986.
5. Thomson, William T., Theory of Vibration with Applications, 4th Edition, pp. 1-91,
and 150-159, Englewood Cliffs, New Jersey: Prentice Hall, 1993.
6. Thomson, William Tyrell, Mechanical Vibrations, 2nd Edition, 9th Printing, pp.1-75,
Englewood Cliffs, New Jersey: Prentice Hall, Inc., 1962.
7. Young, Warren C., Roarks Formulas for Stress & Strain, 6th edition, pp. 100-102,
and 714-717, New York: McGraw-Hill Book Co., 1989.
Rotor Mode Shapes
he analysis of machinery vibration characteristics must be based upon a solid working knowledge of the dynamics associated with a rotating mechanical system. Items that influence rotor motion
include the shaft construction, support locations, and the distribution of masses
such as impellers, spacers, and couplings. Certainly the stiffness and damping
characteristics of the bearings and machine support structure will play an
important role in influencing rotor behavior. In addition, the relationship
between the operating speed range, and the system criticals are very important.
These factors are parameters for establishing the response characteristics
of the rotor system, and the associated dynamic rotor mode shapes. It is meaningful to recognize that many machine characteristics can be adequately
described by applying the basic concepts of mass and support distribution, system viscous damping, plus stiffness of the major elements and support structure.
It must also be recognized that mode shapes may be forced deflections based
upon the active system forces. They may also be natural modes of vibration (resonances), plus a combination of the two to yield a forced resonant response.
MASS AND SUPPORT DISTRIBUTION

The distribution of weight and supports along the length of a rotor establishes the static deflections, plus the static bearing loads. For example, consider a
constant diameter shaft that is simply supported between two points as shown in
Fig. 3-1. This shaft will have a maximum deflection at the midspan, and each
support location will carry one half of the shaft weight.
If one support is moved to the rotor midspan, the condition described in Fig.
3-2 will occur. In this case, the maximum deflection occurs at the unsupported
end of the shaft. The load applied to each support will now be dependent upon
the support characteristics. Specifically, if the left support is a free support, it
will not have a static load. Under this condition, the center support will carry the
entire weight, and the shaft will be balanced on this center pivot. On the other
hand, if the left support is connected to the rotor (e.g., a bearing), it will produce
a vertical reaction if the middle support is not perfectly centered. For this reason,
67
68
Chapter-3
Uniform Shaft
Uniform Shaft
1/2 Weight
1/2 Weight
Maximum
Deflection
Maximum Deflection
20
Fig. 31 Simply Supported Shaft With

Static Deflection Due To Beam Weight
20
Fig. 32 Overhung Shaft With Static

Deflection Due To Beam Weight
overhung machines such as power turbine rotors for dual shaft gas turbines, or
overhung blowers must be carefully examined to determine the static and
dynamic bearing loads, and directions.
Next consider the addition of a concentrated load (e.g., an impeller) at the
middle of the rotor. The diagram presented in Fig. 3-3 represents the deflection
associated with the additional force applied at the midspan of the simply supported shaft. Clearly the center deflection must increase when the additional
load is applied. In addition, with the supports located at the shaft ends, it is reasonable to conclude that the total weight (shaft plus midspan load) will be
equally shared between the two supports.
Fig. 3-4 illustrates the condition of an overhung rotor with the addition of a
center weight. In this configuration, one support is located directly below the
concentrated midspan load. In this case, the mode shape, and the maximum
deflection are identical to Fig. 3-2 with zero external load. However, the force balance has been altered, and the center support must now carry the shaft weight
plus the center load.
Concentrated
Load
Concentrated
Load
Maximum Deflection
35
Fig. 33 Simply Supported Shaft Deflected By

Beam Weight And Center Load
Maximum
Deflection
20
Fig. 34 Overhung Shaft Deflected By

Beam Weight And Center Load
Mass and Support Distribution
69
Next, consider the static mode shapes displayed in Figs. 3-5 and 3-6 that
describe the influence of moving the concentrated load from the midspan to the
end of the rotor. For the shaft simply supported between bearings (Fig. 3-5), the
mode shape returns back to the initial condition (Fig. 3-1). The additional load is
directly transmitted to the right hand support. Under this configuration, the
deflected mode shape is dependent only on the shaft weight, but the support
loads are clearly different.
Finally, for the overhung case of Fig. 3-6, the cantilevered load at the end of
the shaft results in an increase in the maximum deflection. This type of behavior
certainly makes intuitive sense, and it is representative of real overhung
machines. It should also be recognized that the application of this load to the free
end of the rotor will result in a downward vertical restraining force at the left
end support. Again, this is consistent with the forces and moments encountered
in machines such as power turbines and overhung blowers.
Concentrated
Load
Maximum Deflection
Concentrated
Load
Maximum
Deflection
20
50
Fig. 35 Simply Supported Shaft

Deflected By Beam Weight And End Load
Fig. 36 Overhung Shaft Deflected By

Beam Weight And End Load
Overall, it is recognized that the static support forces (i.e., at the bearings),
plus the location and magnitude of the maximum deflection are dependent upon
the support characteristics. It is apparent that the addition of elements such as
impellers, couplings, balance pistons, and spacers will directly influence the
resultant support forces, and the associated maximum static deflection.
Simple mechanical systems can often be modeled as a uniform weight distribution for the shaft, combined with concentrated loads for the impellers. The
static deflections can be calculated with beam theory, and the static bearing
loads determined by summation of moments. Reference books such as Roarks
Formulas for Stress and Strain1 provide characteristic equations for many typical mechanical systems without resorting to detailed beam calculations.
For more complex rotors, it is necessary to divide the rotor into discrete and
1 Warren C. Young, Roarks Formulas for Stress & Strain, Sixth Edition, (New York: McGrawHill Book Company, 1989).
70
Chapter-3
Fig. 37 Hollow Circular Cylinder

With Physical Dimensions
Inner
Diam.(Di)
Outer
Diameter (Do)
definable segments. Based upon the dimensions and material density, it is possible to calculate the weight for each station. From this weight distribution, the
static loads at both bearings can be computed. This approach identifies the mass
distribution along the rotor, plus the resultant bearing forces in a static position.
D o = 2 Ro
D i = 2 Ri
Length (L)
The weight of each section or portion of a rotor is dependent upon the physical dimensions, plus the density of the shaft material. For example, the hollow
circular cylinder depicted in Fig. 3-7 is dimensionally specified by an outer diameter Do, an inner diameter Di, and an overall length L. The shaft radius to the
outer diameter is Ro, and the internal bore radius is identified as Ri. Based on
these dimensions, a variety of necessary calculations may be performed. For
instance, from plane geometry, the cross sectional area of this annulus is computed by subtracting circular areas in the following manner:
2
2
2
2
A annulus = R o R i = --- ( D o D i )
4
(3-1)
The volume of the annulus may be determined by multiplying the cylinder

length L times the cross sectional area Aannulus as stated in the next expression:
L
2
2
V annulus = L A annulus = ------------- ( D o D i )
4
(3-2)
The weight of the annulus may be calculated by multiplying the cylinder

material density times the total volume Vannulus as shown in equation (3-3):
L
2
2
W annulus = V annulus = ----------------------- ( D o D i )
4
(3-3)
For reference purposes, the densities of many common metals have been
summarized in Appendix B of this text. This tabulation also includes the modulus of elasticity, the shear modulus, and the coefficient of thermal expansion for
each metal. These fundamental properties are referred to throughout this book,
and it is convenient to have typical values readily available. It should be mentioned that some materials do not have totally unique properties. For critical calculations the precise physical properties should be obtained from a metallurgical
reference source, or specific tests of the metal.
71
For some calculations, it is necessary to know the mass of the mechanical

part. For the annulus under discussion, the mass is easily determined by dividing the weight from equation (3-4) by the acceleration of gravity G as follows:
W annulus
L
2
2
M annulus = -------------------------- = ----------------------- ( D o D i )
4G
G
(3-4)
In many instances, machinery shafts are not hollow, and they are fabricated
of solid metal. If the inner diameter Di is set equal to zero in equations (3-1)
through (3-4), the following expressions for solid machine shafts with an outer
diameter of D are easily developed:
2
D
A solid = ----------------4
(3-5)
2
LD
V solid = --------------------------4
(3-6)
2
LD
W solid = ------------------------------------4
(3-7)
LD
M solid = ------------------------------------4G
(3-8)
In order to be perfectly clear on the dimensional aspects of equations (3-1)

to (3-8), the defined variables and their respective English engineering units are
summarized as follows:
L
D
Do
Di
Ro
Ri
A
V
W
M
G
=
=
=
=
=
=
=
=
=
=
=
=
Cylinder Length (Inches)

Solid Cylinder Outer Diameter (Inches)
Hollow Cylinder Outer Diameter (Inches)
Hollow Cylinder Inner Diameter (Inches)
Hollow Cylinder Outer Radius (Inches)
Hollow Cylinder Inner Radius (Inches)
Cylinder Material Density (Pounds / Inch3)
Cylinder Cross Sectional Area (Inches2)
Cylinder Volume (Inches3)
Cylinder Weight (Pounds)
Cylinder Mass (Pounds-Second2 / Inch)
These variables and associated units will be used in the next sections on
inertia, plus throughout the remainder of this text.
72
Chapter-3
Case History 3: Two Stage Compressor Rotor Weight Distribution

Consider the compressor rotor depicted in Fig. 3-8. This high speed rotating
assembly consists of a shaft with various diameters, plus two midspan impellers,
and a thrust collar. Since the impellers are mounted back to back, this particular
design does not include a balance piston. As indicated on the diagram, the rotor
weights 282 pounds, and there is a moderate overhang on the thrust end of the
rotor. Normal operating speed varies between 15,000 and 17,500 RPM. This compressor is a drive through unit with a stream turbine coupled to one end, and
another compressor coupled to the opposite (thrust) end. Historically, the bearing
on the drive end of the rotor seldom exhibited any damage, but the journal bearing located at the thrust end was often found to be in a distressed or damaged
condition. On other occasions, the bearings would fail during operation, and the
rotor would be severely damaged. Depending on the severity of the failure, the
shaft was often chrome plated. After severe failures, the entire rotating assembly
was totally replaced.
This machine operates in a very dirty and corrosive environment. The rotor
was constantly subjected to large unbalance forces due to the accumulation of
Bearing
Journal
3.62"
Seal
Impellers
14.5"
Seal
Thrust
Collar
40.68"
Wa
Wb
7.49"
21.17"
24.03"
27.24"
33.43"
Fig. 38 Two Stage Centrifugal Compressor Rotor

With Weight And Moment
Diagram
40.68"
43.66"
45.92"
49.25"
53.06"
2.09#
13.84#
9.13#
17.51#
10.40#
32.67#
Seal
17.37#
37.18#
14.51"
17.96"
Thrust End
Journal
Thrust Collar
4.40"
Impeller
2.62"
Impeller
19.21#
36.41#
Seal
1.11#
4.83#
10.58#
Drive End
Journal
37.18#
32.40#
Total Rotor Weight

282 Pounds
73
material in the impellers, plus random corrosion of the impeller vanes. During a
typical run, the compressor would startup smoothly after an overhaul. However,
synchronous running speed vibration response would always deteriorate with
the passage of time.
An initial step in the analysis of this problem required the determination of
static bearing loads. To achieve this goal, the rotor was divided into fifteen sections, and the weight of each section was calculated. Since this was a solid shaft,
equation (3-7) was used to compute the weight of each shaft section. The weight
of the thrust collar was determined with (3-3), and the two impellers were
weighed separately. Next, the thrust collar and wheel weights were combined
with their respective shaft section weights, and the weight distribution summarized in Fig. 3-8.
This information was then combined with the distance from the center of
the drive end bearing to the centroid of each rotor segment. The complete array
of weights and distances are shown in Fig. 3-8. From this sketch, it is possible to
perform a summation of moments around the center or transverse axis of the
drive end journal in the following manner:
Momentsccw = Momentscw
(3-9)
17.54 Inch-Pounds + W b 40.68 Inches = 6, 768.74 Inch-Pounds

W b 40.68 Inches = 6, 751.20 Inch-Pounds
W b = 166 Pounds
Since the total rotor weight is 282 pounds, the next force balance applies:
W a + W b = 282 Pounds
W a + 166 Pounds = 282 Pounds
W a = 116 Pounds
Thus, the drive end bearing has a 116 pound static load, and the thrust end
journal carries 166 pounds. Although the differential force is only 50 pounds, it is
an appreciable percentage difference. Ultimately, it was determined that the
bearings were only marginally sized to accommodate the rotor weight. However,
they were considerably undersized when the additional unbalance forces due to
foreign objects were included. It was also determined that the available load
capacity for the drive end was barely acceptable, whereas the load carrying capability for the thrust end journal bearing was unacceptable. Based upon these
conclusions, both bearings were increased in size and load capacity.
The larger bearings reduced the number of machine failures per year, and
overall reliability was substantially improved. Further improvements in
machine longevity would require changes in the chemical plant process. Unfortunately, the required alterations to the processing scheme could not be economically justified. Hence, the compressor rotor was occasionally sacrificed to meet
74
Chapter-3
production quotas.
From this example, it is clear that even a simple analysis of rotor weight
distribution and bearing static loads may be beneficial. In some cases this may
solve a problem, or it may provide insight into prospective solutions. It must also
be recognized that the dynamics of the rotating system must be considered. This
includes the effects of mass unbalance that serves to deform the mode shape,
plus the effects of inertia, stiffness, and damping of machine elements.
INERTIA CONSIDERATIONS AND CALCULATIONS

As shown in the previous example, the calculation of rotor weights is easily
achieved for rotors that can be segmented into a series of cylinders. For a homogeneous material, the volume and weight of each section may be computed. The
summation of individual cylindrical section weights yields the total rotor weight.
In most instances, this is an acceptable and achievable computational procedure.
However, problems occur with machines that have non-symmetric shafts, or are
composed of multiple materials. For instance, the rotor weight of electric
machines, such as motors and generators, may be indeterminate due to an
unknown combination of iron, insulation, and copper within the rotor windings.
In these situations, the rotor assembly can be weighed on a scale, and the distribution of weight between bearing journals estimated. Any additional elements
such as a cooling fan, flywheel, or overhung exciter may then be added to determine the total rotating system weight. Hence, when accurate calculations cannot
be performed, the machinery diagnostician can revert to the traditional experimental technique of weighing the rotor.
In addition to weight or mass, the inertia properties of a rotating system
must also be considered during any dynamic analysis. Due to the complexity of
inertia calculations, and the strong potential for confusion between area and
mass moment of inertias, significant errors are possible. The physics definition of
inertia is the property of a body to resist acceleration. This includes the tendency
of a body at rest to remain at rest, or the tendency of a body in motion to stay in
motion. This reference to a physical body implies mass, which suggests the presence of a three-dimensional object. Hence, an area moment of inertia is somewhat of a misnomer since a plane area has no depth, and therefore no mass, nor
the potential for inertia. Nevertheless, area moments of inertia abound in the literature, and for the sake of completeness, they will be discussed.
It should also be mentioned that confusion regarding inertia is further compounded by some of the technical literature that transposes polar with the transverse inertias. In other cases, the authors fail to identify the specific type of
inertia calculations, much less the reference inertia axis. In an effort to circumvent these potential sources of confusion, the following discussion of inertia is
offered for consideration.
It is desirable to begin with a review of the area moment of inertia. In Fig.
3-9, Sketch A shows an irregular plane surface with axes x-x and y-y passing
through the area designated as A. These rectangular axes may be in any location
Inertia Considerations and Calculations
Area (A)
75
Incremental
Area (dA)
At Coordinates
(X and Y)
y
Incremental
Area (dA)
At Radius (R)
from Axis z-z
y0
z
x
z
y
Sketch A
Sketch B
XOffset
y0
Sketch C
Fig. 39 Area Moment Of Inertia For An Irregular Plane Surface
with respect to the area. If the area was symmetric (e.g., circle or rectangle), and
if the axes were axes of symmetry, they would be called the principal inertia
axes. However, the general solution for inertia about each axis is determined
from strength of materials integrals as:
I xx =
dA
(3-10)
I yy =
dA
(3-11)
The X and Y terms in (3-10) and (3-11) are distances from each respective
axis to the incremental area identified as dA. This definition of area moment of
inertia is consistent with references such as Roark2, or Shigley3. In both equations, inertia is calculated by multiplying each area by the square of the distance
to the respective reference axis. The area is measured in inches squared, and this
is multiplied by the square of the distance to the axis in inches. This product
yields an area moment of inertia with English engineering units of Inches4.
In Sketch B of Fig. 3-9, a third axis z-z has been added. This new axis is perpendicular to the plane of the area A, and it passes through the intersection of
axes x-x and y-y. The distance from the intersection of axis z-z with the area A, to
any incremental area dA is identified by the radius R. The inertia about this axis
is termed the area polar moment of inertia J that is given by:
J zz =
dA
(3-12)
From plane geometry it is known that the radius R, may be expressed in

terms of X and Y coordinates as:
2 Warren C. Young, Roarks Formulas for Stress & Strain, Sixth Edition, (New York: McGrawHill Book Company, 1989), p. 59.
York: McGraw-Hill Book Company, 1986), p. 9.13.
76
Chapter-3
2
R = X +Y
(3-13)
Equation (3-13) may be inserted into (3-12), and expanded as follows:

J zz =
J zz =
( X
+ Y ) dA
dA + Y dA
Substituting (3-10) and (3-11) into the above expression, the polar inertia
Jzz is equated to the summation of the transverse area inertias Ixx and Iyy as:
J zz = I xx + I yy
(3-14)
For a symmetrical area such as a circle, the transverse area inertias Ixx and
Iyy are equal. Thus, setting inertias Ixx=Iyy=I, equation (3-14) may be restated in
the following simplified format:
J zz = I + I = 2 I
(3-15)
Since most rotating shafts are circular, equation (3-15) is common within
the machinery business. The applicability of this geometric simplification will be
better appreciated during the forthcoming discussion of mass moment of inertia.
However, before addressing that topic, the translation of the inertia axis should
be reviewed. Specifically, Sketch C in Fig. 3-9 identifies a new axis yo-yo that is
parallel to the previously defined vertical axis y-y. The constant distance between
the two axes is identified as Xoffset. It can be shown that the area moment of inertia about this new axis is given by the expression:
Iy
o yo
= I yy + A ( X offset )
(3-16)
The same argument may be made in the perpendicular plane for the area
polar moment of inertia. Consider a new polar axis zo-zo that is parallel to the
previously defined z-z axis. If the distance between the two axes is identified as
Roffset, the area moment of inertia about this new polar axis is given by:
Jz
o zo
= J zz + A ( R offset )
(3-17)
The moment of inertia of an area about a particular axis may be converted

to a mass moment of inertia by including the thickness, and the density of the
body. This converts a two-dimensional into a three-dimensional problem, and the
complexity of the associated equation structure increases proportionally. For
explanation purposes, assume that the plane area shown in Fig. 3-9 is expanded
by adding a finite depth. If a material density is applied to the resulting volume,
the mass shown in Fig. 3-10 evolves. Sketch A in this diagram reveals that a twodimensional x-y coordinate system is inadequate for locating a point within the
77
Incremental
Mass (dM)
At Coordinates
(X, Y and Z)
Mass (M)
z
x
XOffset
z
y
Sketch A
Sketch B
z1
z0
z0
y
YOffset
z1
Sketch C
Fig. 310 Mass Moment Of Inertia For An Irregular Solid Object
mass M. Hence, the three-dimensional x-y-z coordinate system in Sketch B of

Fig. 3-10 is necessary to locate an incremental mass dM within the boundaries of
the object. The equations for mass moment of inertia around the three axes are
defined by sources such as Marks Handbook4 as follows:
2
I xx =
(Y
+ Z ) dM
I yy =
( X
+ Z ) dM
I zz =
( X
+ Y ) dM
(3-18)
(3-19)
(3-20)
The translation of one axis to another parallel axis also applies to mass
inertia as well as to the previously discussed area moment of inertia. For example, the polar mass moment of inertia around axis z-z may be translated to the
parallel axis zo-zo with the following common expression:
Jz
o zo
= J zz + M ( X offset )
(3-21)
Within this equation, the distance Xoffset represents the parallel offset along
the x-x axis as shown in Sketch C of Fig. 3-10. If the offset of the new axis was
further displaced from both the x-x and y-y axes, the polar moment of inertia
along the new axis identified as z1-z1 would be given by the following:
Jz
1 z1
= J zz + M ( R offset )
(3-22)
The distance between the z-z axis and the z1-z1 axis is defined by the radial
offset Roffset. From (3-13) it is known that Roffset2 = Xoffset2 + Yoffset2. The validity
4 Eugene A. Avallone and Theodore Baumeister III, Marks Standard Handbook for Mechanical Engineers, Tenth Edition, (New York: McGraw-Hill, 1996), p. 3-9.
78
Chapter-3
of this is demonstrated by allowing Yoffset to equal zero, and noting that (3-22)
then reverts back to equation (3-21). In all cases, (3-21) and (3-22) are analogous
to the parallel axis equation (3-17) for a plane area. In fact, this similarity in
mathematical formats is the source for some of the confusion associated with the
topic of inertia. In order to maintain a proper distinction between area and mass
moment of inertia, the diagnostician should always pay close attention to the
engineering units. Recall that the area moment of inertia carries English engineering units of Inches4, and the mass inertia has units of Pound-Inch-Second2.
The integrals shown on the last few pages are interesting calculus topics,
but they do not help the field diagnostician until they are solved for specific geometric shapes. For instance, it is meaningful to develop the equations for a circular shaft cross section. In order to maintain continuity, the hollow circular
Diameter
Centerline
Do = 2 Ro
Inner
Diam.(Di)
Outer
Diameter (Do)
Di = 2 Ri
o
Axial
Centerline
Fig. 311 Hollow Circular

Cylinder With Principal
Axes Identified
b
Length (L)
cylinder from Fig. 3-7 will be reused. For the calculation of inertia, the principal
axes passing through the axial centerline o-o, and the midspan diametral centerline b-b have been shown in Fig. 3-11. For this geometric figure, the cross sectional area moment of inertia Iarea along a diameter of the hollow circle is given
by Spotts5, or Harris6 in the following common format:
4
I area
annulus
( Do Di )
= ----------------------------------64
(3-23)
The area polar moment of inertia Jarea is computed about an axis that is
perpendicular to the circular cross section. On Fig. 3-11 this would be the centerline axis o-o of the cylinder. The polar inertia of this circular area is equal to
twice the inertia along a diameter (i.e., J=2xI) as evident from (3-15). Multiply-
5 M.F. Spotts, Design of Machine Elements, 6th Edition, (Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1985), p. 18.
6 Cyril M. Harris, Shock and Vibration Handbook, Fourth edition, (New York: McGraw-Hill,
1996), p. 1.12.
79
ing equation (2-23) by two, the next expression for the area polar inertia evolves:
4
J area
annulus
( Do Di )
= ----------------------------------32
(3-24)
The validity of equation (3-24) is supported by Spotts, page 150, and others.
Note that the last two expressions are plane area moments of inertia (i.e., no
depth, no density, and no weight). However, in many machinery calculations it is
mandatory to consider the mass of the element. When mass is included, the calculations become more complex. Specifically, if the hollow cylinder in Fig. 3-11 is
considered to be a rotating machinery shaft, the mass polar moment of inertia
along the axis of rotation (axis o-o) is defined by Gieck7 and others. If these equations are placed into the nomenclature used in Fig. 3-11, the general expression
for the mass polar moment of inertia of the cylinder may be stated as:
J mass
annulus
M annulus
2
2
= -------------------------- ( R o + R i )
2
(3-25)
Equation (3-25) is the mass inertia term of a rotating shaft that is often
described as the WR2 of the rotor. In actuality, this quantity should be identified
as the mass polar moment of inertia, but common nomenclature sometimes
supersedes technical accuracy. Nevertheless, if the diameters are used instead of
the radii, and if (3-4) for the annulus mass is substituted into (3-25); the following manipulation may be performed to reach a common equation used for calculation of the mass polar moment of inertia for a hollow cylinder.
2
J mass
J mass
annulus
annulus
M annulus
M annulus D o D i
2
2
= -------------------------- ( R o + R i ) = -------------------------- ------- + -------
2
2
4
4
M annulus
L
2
2
2
2
2
2
= -------------------------- ( D o + D i ) = ----------------------- ( D o D i ) ( D o + D i )
8
4G8
Simplification of the last expression yields the following equation:

J mass
annulus
L
4
4
= ----------------------- ( D o D i )
32 G
(3-26)
This mass polar inertia is as important to a torsional analysis as the mass

is necessary for a lateral analysis. Specifically, equation (2-102) identifies the utilization of the polar inertia to compute a torsional natural frequency in the same
way that the mass is used in (2-44) to compute a lateral natural frequency. As
shown in Table 2-1, the inertial term in the general equation of motion is governed by mass in a lateral system, and polar inertia in a torsional system. During the analysis or modeling of real machinery both the polar inertia and the
7 Kurt Gieck and Reiner Gieck, Engineering Formulas, 6th edition, (New York: McGraw-Hill
Inc., 1990), p. M3.
80
Chapter-3
transverse inertia are utilized. Since (3-26) defines the mass polar moment of
inertia Jmass through the axial centerline o-o, it is now reasonable to define the
transverse mass Inertia Imass of the hollow cylinder. Although any defined axis
may be used for the calculations, the customary midspan diameter axis b-b
depicted in Fig. 3-11 will be used for the following exercise. Again, extracting a
standard mass inertia equation from Gieck, and modifying the terms to be consistent with Fig. 3-11, the general expression for the mass moment of inertia on
the cylinder diameter axis b-b passing through the center of gravity is as follows:
I mass
annulus
M annulus
2
2
2
= -------------------------- ( 3 R o + 3 R i + L )
12
(3-27)
If diameters are used instead of radii, and if the mass equation (3-4) is
included, the transverse inertia equation (3-27), may be modified as shown:
2
I mass
annulus
I mass
annulus
I mass
annulus
I mass
annulus
I mass
annulus
M annulus 3D o 3D i
M annulus
2
2
2
2
= -------------------------- ( 3 R o + 3 R i + L ) = -------------------------- ---------- + ---------- + L
12
12
4
4
2
M annulus
2
2 4L
= -------------------------- D o + D i + ----------
16
3
M annulus
M annulus
2
2
2
= -------------------------- ( D o + D i ) + -------------------------- L
16
12
L
L
2
2
2
2
2
2
2
= -------------------------- ( D o D i ) ( D o + D i ) + -------------------------- ( D o D i ) L
12 4 G
16 4 G
3
L
4
4 L
2
2
= ----------------------- ( D o D i ) + -------------------------- ( D o D i )
64
G
48
If the mass polar moment of inertia from (3-26) is substituted into the last
expression, the transverse inertia equation may be simplified as follows:
I mass
annulus
J mass
L3
2
2
annulus
= ------------------------------- + -------------------------- ( D o D i )
48
G
2
(3-28)
If the length L of the annulus is small compared to the outer diameter, the
influence of the far right hand term in equation (3-28) is significantly diminished, and the following approximation is often used:
I mass
J mass
annulus
------------------------------annulus
2
(3-29)
This result is consistent with the previous relationship described by equation (3-15) between the ratio of inertias for a plane circular cross section. In
actual practice during the analytical modeling of a rotor system, the shaft station lengths are normally kept fairly short, and equation (3-29) may be a good
81
approximation. For longer shaft sections, the more complex equation (3-28) must
be used. Equation (3-29) is particularly useful for estimating the transverse inertia of wheels that have a large diameter, and a comparatively short length. For
example, a centrifugal compressor impeller may be 24 inches in diameter, with a
disk and a cover thickness of only 0.25 inches each. In this situation, one half of
the mass polar moment of inertia Jmass will be very close to the detailed transverse mass moment of inertia Imass.
As mentioned earlier in this chapter, the diagnostician must always be
aware of the potential dilemma in the application of inertia within technical documents, and computer programs. For instance, one set of software uses weight
inertia instead of mass inertia. The output inertia values from these programs
carry inertia units of Pounds-Inches2 instead of Pound-Inch-Second2. Although
the difference between the two inertia values is only the acceleration of gravity,
the results can be confusing to the unprepared. In all cases, it is mandatory to be
completely knowledgeable of all aspects of any inertia calculations. Within this
text, inertias and their respective English engineering units are as follows:
Iarea
Jarea
Imass
Jmass
=
=
=
=
Area Moment of Inertia on Diameter (Inches4)

Area Polar Moment of Inertia (Inches4)
Mass Transverse Moment of Inertia (Pound-Inch-Second2)
Mass Polar Moment of Inertia (Pound-Inch-Second2)
In many instances, machinery shafts are not hollow, and they are fabricated
of solid metal. If the inner diameter Di is set equal to zero in equations (3-23), (324), (3-26), and (3-28), the following inertia expressions for solid shafts with a
diameter of D are easily developed:
4
I area
solid
D
= ----------------64
(3-30)
J area
solid
D
= ----------------32
(3-31)
4
J mass
I mass
solid
LD
= ------------------------------------solid
32 G
J mass
L3 D2
solid
= ------------------------+ ---------------------------------------
48 G
2
(3-32)
(3-33)
Once again these expressions may be verified from various sources such as
Spotts, Marks, or Shigley. Please recall that the developed equations are based
upon a circular cross section. If the cross sectional area is not circular, then the
equations must be modified based upon the original integrals used to define inertia. Often this type of calculation is not practical due to the complexity of the
mechanical part. In these cases, the diagnostician must resort to other tech-
82
Chapter-3
niques to determine the inertia properties of the machine element.

In the same way that a scale may be used to determine the weight of a
rotor, there are experimental techniques that may be applied to determine inertia properties. For example, consider the machine part shown in Fig. 3-12. This
could be a blank for a bull gear or a flywheel, or any other machine element for
which a mass polar moment of inertia may be required. Due to the complexity of
the part it may not be feasible to calculate the inertia directly, but it is possible to
experimentally determine the inertia.
Fixed and
Rigid
Support
X0ffset
Element
Weight (W)
Fig. 312 Mechanical

Arrangement For Rocking
Test To Determine Mass
Polar Moment of Inertia
Rocking
Oscillation
40 to 60
One technique consists of suspending the part from a horizontal support as

shown in Fig. 3-12. Ideally, this support member should be a knife edge, but more
realistically, it will probably be a solid circular rod as depicted. If the mass polar
moment of inertia through the axial centerline of the element is desired, then the
distance between that centerline and the supporting pivot point Xoffset must be
accurately measured. In addition, the part should be weighed to determine the
total weight W in pounds. The machine element may now be rocked back and
forth as a compound pendulum. By measuring the period of the motion, the inertia about the pivot point may be determined. This offset inertia may now be
translated back to the centerline of the machine part by applying the parallel
axis equations previously developed.
In actual practice, the Inertia of this compound pendulum may be determined from Marks Handbook8. Extracting the appropriate equation, and placing
it in terms used within this text, the following equation for overall inertia about
the pivot point may be stated:
2
Jz
o zo
W X offset Period
= ----------------------------------------------------------2
4
(3-34)
8 Eugene A. Avallone and Theodore Baumeister III, Marks Standard Handbook for Mechanical Engineers, Tenth Edition, (New York: McGraw-Hill, 1996), p. 3-16.
83
If equation (3-34) is inserted into the axis translation equation (3-21), the
following is obtained:
Jz
o zo
M ( X offset )
J zz = J z
o zo
J zz
W X offset Period
2
= ----------------------------------------------------------- M ( X offset )
2
4
2
W X offset Period W
2
J zz = ----------------------------------------------------------- ----- ( X offset )
2
G
4
Factoring out the common terms, the above may be simplified to:
X offset
Period 2
J zz = W X offset ------------------- ------------------
G
2
(3-35)
This is a useful expression since all of the variables are easily determined.
Specifically, the weight W of the machine part can be measured on a scale, and
the distance between the pivoting point and the geometric center of the element
Xoffset is easily measured. The acceleration of gravity G is constant, and the
Period of the swinging motion is determined with a stopwatch in seconds. Normally, a series of runs are made to determine the average period of oscillation.
X0ffset = 5.07"
Fixed and
Rigid
Support
Length
(L =3.735")
Diameter
(D=6.312")
Fig. 313 Rocking Test To

Determine Mass Polar
Moment Of Inertia Of A
Solid Brass Cylinder
Brass
Cylinder
Weight
(W=36 Lbs.)
Rocking
Oscillation
40 to 60
Experimental techniques are always more credible if the validity of the

equations, and the experimental procedures can be verified with a controlled
test. With an inertia experiment, the object to be tested should be of simple
geometry such as the solid cylinder shown in Fig. 3-13. As indicated on the diagram, the average cylinder diameter D is equal to 6.312 Inches, and the average
length L is 3.735 Inches. A shop scale indicated that the weight W of the speci-
84
Chapter-3
men was 36.0 0.1 Pounds. Since the cylinder was solid brass, the density was
found in Appendix B of this text to be 0.308 Pounds/Inch3. Two holes were drilled
and tapped into the side of the cylinder, and two screw eyes were inserted to provide the pivot arm. Since the mass polar moment of inertia along the axial centerline is desired, the distance between the pivot point and the cylinder center
line Xoffset was measured to be 5.07 Inches.
Before starting the experiment, it is desirable to check the physical properties of the brass cylinder. For example, equation (3-7) may be used to compute
the weight of the brass cylinder based upon the average dimensions, and the
density of the material as follows:
2
LD
W solid = ------------------------------------4
2
3.735 Inches 0.308 Pounds/Inch ( 6.312 Inches )

W solid = ------------------------------------------------------------------------------------------------------------------------------------ = 35.997 Pounds
4
3
The calculated value shows excellent agreement with the measured weight
of the cylinder, and 36.0 pounds may be confidently used for the ensuing inertia
measurements. Some individuals might argue that this type of weight check is
unnecessary. However, if the material contained inclusions, or if the density was
wrong, or if one or more of the dimensions were incorrect the weights would
not match, and the experimental accuracy would be in jeopardy. So a simple calculation such as this weight check is desirable to insure that the physical parameters are in unison. The other calculation that should be made at this stage is the
mass polar moment of inertia of the brass test specimen. From the previously
developed equation (3-32) the polar moment of inertia of this brass cylinder is
computed in the following manner:
4
J mass
solid
LD
= ------------------------------------32 G
4

= -----------------------------------------------------------------------------------------------------------------------------------2
solid
32 386.1 Inches/Second
3.735 0.308 1, 573.33
2
J mass
= ----------------------------------------------------------------------- = 0.464 Pound-Inch-Second
solid
32 386.1
3
J mass
At this point the test calibration setup is established, the final answer is
known, and the only remaining variable to be defined for equation (3-35) is the
period of the oscillatory motion. Several preliminary swings of the brass cylinder
revealed that the time for one complete cycle was less than a second. In addition,
the friction between the two eye bolts and the support rod caused the oscillating
mass to grind to a stop after only a few cycles. This problem was partially remedied by putting a tight plastic sleeve on the support rod, and then covering this
surface with lithium grease. This friction reduction effort was rewarded by a substantial increase in the number of possible oscillatory cycles. However, it was
then discovered that the horizontal support rod was not quite level, and the
85
brass cylinder had a tendency to walk down the rod. This problem was corrected
by re-leveling the support rod and the cylinder.
Following these test setup modifications, the actual test was conducted. The
brass cylinder was displaced about 30 from the vertical centerline and released.
The peak of the motion at one extremity was visually sighted, and a stopwatch
was used to measure the time required for multiple back and forth cycles. The
final measured test data is summarized as follows:
Rocking Run #1
Rocking Run #2
Rocking Run #3
Rocking Run #4
Rocking Run #5
Rocking Run #6
Rocking Run #7
Rocking Run #8
Rocking Run #9
Rocking Run #10
Total Time =
7.83 Seconds for 10 cycles

Average Time for 1 Cycle = 0.7919 Seconds = Period
It is easy to lose track of the cycle count, or miss a timing point, and negate
the accuracy of a data set. These types of errors are evident during the data collection work, and erroneous times are identified and discarded. For instance,
approximately twenty runs were made to collect the data in the above tabular
summary. Ten of the timing runs were not used due to obvious errors in the data
accumulation. The ten acceptable test runs reveal an average period of 0.7919
Seconds. This is considered to be a consistent value, and the experimental mass
polar moment of inertia may now be computed from equation (3-35).
X offset
Period 2
J zz = W X offset ------------------- ------------------
2
G
0.7919 Seconds 2
5.07 Inches
J zz = 36.0 Pounds 5.07 Inches ------------------------------------ ------------------------------------------------2
386.1 Inches/Second
J zz = 182.52 Pound-Inches ( 0.01588 0.01313 ) Second = 0.502 Pound-Inch-Second

2
The experimental value for the polar inertia is 0.502 versus the calculated
value for this brass cylinder of 0.464 Pound-Inch-Seconds2. The difference of
0.038 represents an 8% error of the experimental versus the computed value. In
some instances this level of deviation is perfectly acceptable. For example, if the
part under test was a coupling hub that will be mounted on a power turbine with
an inertia of 40.0 Pound-Inch-Seconds2, the small differential of 0.038 PoundInch-Seconds2 would be insignificant. However, if the part under test was one of
86
Chapter-3
eight impellers to be mounted on a slender shaft, the cumulative error may be

unacceptable. In order to explain this error, it is necessary to re-examine the
component equations used for the inertia test calculation. Specifically, (3-34) for
the total inertia about the pivot point may be solved as follows:
2
Jz
o zo
W X offset Period
= ----------------------------------------------------------2
4
2
Jz
o zo
36.0 Pounds 5.07 Inches ( 0.7919 Seconds )

2
= -------------------------------------------------------------------------------------------------------------- = 2.899 Pound-Inch-Second
2
4
Substituting the test inertia from the above calculation back into equation
(3-21), and performing the axis translation, the following result is obtained:
Jz
J zz
o zo
2
W
2
2.899 Pound-Inch-Second = J zz + ----- ( X offset )
G
36.0 Pounds
2
2
= 2.899 Pound-Inch-Second ------------------------------------------------2 ( 5.07 Inches )
386.1 Inches/Second
J zz = ( 2.899 2.397 ) Pound-Inch-Second = 0.504 Pound-Inch-Second

2
As expected, this result is identical to the previous answer obtained by

using the composite equation (3-35). The interesting point of the above calculations is that inertia due to the axis translation is equal to 2.397, versus the overall test inertia of 2.899 Pound-Inch-Seconds2. Hence, the cylinder inertia is only
about 20% of the inertia due to the axis translation. This is not a desirable condition since the axis translation is the dominant term. The geometrical configuration displayed in Fig. 3-12 would not be as error prone since the XOffset distance
resides within the body of the element, and the axis translation term would not
dominate the test. Hence, the diagnostician should always be concerned about
trusting this type of experimental inertia test with a long XOffset distance between
the pivot axis and the desired principal polar moment of inertia axis. The other
lesson to be learned is that simplified expressions such as (3-35) may not provide
full visibility concerning the potential accuracy of the final results. In some cases
it is necessary to revert back to the basic equations, and reexamine the entire
calculation and/or experimental test procedure.
With respect to the brass cylinder, it is concluded that improvement of the
test accuracy will require a reduction or elimination of the axis translation term.
This could be accomplished with a test that consisted of suspending the mass
from cables, and then measuring the period as the mass oscillated in a twisting
manner (Fig. 3-14). Since the axis of oscillation is the axial geometric centerline
of the element, there is no axis translation involved, and test accuracy should be
improved. This type of inertia test is ideal for machine parts such as compressor
impellers or turbine disks that contain complex geometrical cross sections. For
87
Suspension Cable
Length (Ls-c)
Suspension
Cable
Radius
(R s-c)
Element
Weight (W)
Fig. 314 Mechanical

Arrangement For Twisting
Test To Determine Mass
Polar Moment of Inertia
Twisting
Oscillation
30 to 40
Element Axial
Centerline
instance, the internal star pattern shown in Fig. 3-14 might be difficult to model
with an equivalent inner diameter for the machine part.
For this procedure, the machine element is suspended from three thin
cables (3 points determine a plane), spaced at 120 apart. The test piece must be
leveled as precisely as possible. If it is not level, then any induced twisting oscillations will cause the machine element to wobble during the test. This wobble
not only negates the test accuracy, it can prove to be dangerous for parts with
any appreciable physical size and weight. After leveling, the average suspension
cable length Ls-c and the cable radius Rs-c are accurately measured and recorded.
For best results, each of the suspension cable lengths should be equal, and the
radius for all three cables should be identical. As before, the machine element to
be tested is weighed on an accurate scale in English units of Pounds.
During execution of this test, the machine element is manually displaced in
a twisting manner, and released. The machine part will torsionally twist back
and forth, and the period of the twisting oscillations will be measured with a
stopwatch. Since friction should not be major problem, the part will oscillate
back and forth for many cycles. It is not unusual to observe thirty or more cycles
resulting from one initial displacement. Based upon these measured parameters
the mass polar moment of inertia may be computed as follows:
2
W R s c Period
J zz = ------------------------------------------------------2
4 L s c
(3-36)
The general form of (3-36) was extracted from the Shock & Vibration Handit was converted to the nomenclature used in this text. As previously
book9, and
1996), p. 38.5.
88
Chapter-3
Fig. 315 Twisting Test To

Determine Mass Polar
Moment Of Inertia Of A
Solid Brass Cylinder
Axial Centerline
(Ls-c=33.73")
Suspension Cable Length
discussed on the rocking inertia test, it is mandatory to validate the test procedure with an actual test on a known geometric shape. For comparative purposes,
the solid brass cylinder used for the rocking test will be used for this twisting
inertia test as shown in Fig. 3-15. From this diagram it is noted that the average
suspension cable radius Rs-c was 3.00 Inches, and the average cable length Ls-c
was 33.73 Inches. As before, the total cylinder weight was 36.0 Pounds.
Suspension
Cable
Radius
(R =3.00")
s-c
Twisting
Oscillation
30 to 40
Brass
Cylinder
Weight
(W=36.0 Lbs.)
The first test configuration used 48 Inch long suspension cables in an effort
to increase the period of the oscillation, and improve the time measurement
accuracy. Conceptually this was a good idea, but it turned out to be impractical
since the long cables had a tendency to wrap around each other. This proved to be
an unmanageable situation, and the support cable lengths were reduced to 33.73
Inches. During the acquisition of test data, the cylinder was twisted about 20
circumstantially from rest, and released. The peak of the motion at one extremity was visually sighted, and a stopwatch was used to measure the time required
for complete back and forth cycles. The test data is summarized as follows:
Twisting Run #1
Twisting Run #2
Twisting Run #3
Twisting Run #4
Twisting Run #5
Twisting Run #6
Twisting Run #7
Twisting Run #8
Twisting Run #9
Twisting Run #10
Total Time =

Average Time for 1 Cycle = 1.4018 Seconds = Period
89
This is a much smoother test than the rocking inertia previously discussed.
The number of miscounts and aborted runs were substantially reduced, and
approximately fifteen runs were made to collect the data shown in the above tabular summary. Five of the timing runs were not used due to obvious errors in
data accumulation. The ten consistent test runs reveal an average period of
1.4018 Seconds. This was considered to be a consistent value for this experimental procedure, and the mass polar moment of inertia may be computed from
equation (3-36) as follows:
2
J zz
W R s c Period
= ------------------------------------------------------2
4 L s c
2
36.0 Pounds ( 3.00 Inches ) ( 1.4018 Seconds )

2
J zz = ---------------------------------------------------------------------------------------------------------------------- = 0.478 Pound-Inch-Second
2
4 33.73 Inches
The experimental polar inertia from this twisting procedure of 0.478 is
quite close to the previously calculated value of 0.464 Pound-Inch-Seconds2. The
3% deviation is quite acceptable for most rotor dynamics calculations. This is
particularity true for smaller components that are stacked on a shaft to achieve
a final rotor assembly. It should be recognized that both the rocking and the
twisting inertia tests have their own domain of application that is dependent on
the size and geometry of the machine element.
Just as the weights of individual components are summed up to determine
a total rotor weight, the inertia of the component pieces may be added to determine the overall rotor polar inertia. The origin of the inertia values may be from
calculations of defined geometries, or from experientially determined inertia values. In any case, as long as the engineering units and the inertia axis are common, the numeric inertia values may be summed up to determine the mass polar
moment of inertia for the entire rotating assembly.
In some instances, there is minimal opportunity to determine the inertia of
rotor components since the unit cannot be disassembled or unstacked. In these
situations, the general inertia characteristics may be estimated based upon
available dimensions and probable materials of construction. In other cases, the
complexity of the rotor may not allow a reasonable segmentation and estimation
of inertia properties. This is particularly true for rotors that are constructed of
multiple materials, plus rotors that contain complicated geometric configurations. In these instances, another experimental technique may be employed to
determine the overall mass polar moment of inertia of the rotor.
This technique is based upon the familiar college physics experiment
depicted in Fig. 3-16. In this diagram, a cylinder or drum is mounted in rigid
bearings that allow rotation of the cylinder, but restrict any lateral or translation
movement of the cylinder. A cord is wrapped around the cylinder at a shaft
radius of Rs. It is assumed that this cord is of insignificant weight and diameter,
and that it will not stretch with the application of axial tension. Next, a known
weight (mass M) is attached to the end of the cord, and allowed to free fall. The
90
Chapter-3
time T required to fall a distance D is measured with a stop watch. The experiment normally consists of determining the cylinder mass polar moment of inertia Jmass based on the four known quantities of radius Rs, mass M, fall distance
s)
(R
R
ad
iu
s
D, and the average fall time T.
Angular
Velocity ()
Cylinder with Mass Polar

Moment of Inertia (Jmass)
Fig. 316 Traditional

Mechanical Arrangement
For Polar Moment Of Inertia Experiment Based On
Falling Mass Attached To
A Rotating Cylinder
Fall Distance (D)

during Fall Time (T)
Mass
(M)
This basic physics problem may be solved by constructing free body diagrams of the cylinder and the falling mass, developing equations of motion, and
then solving for the polar inertia term. Another way to achieve the same result is
by performing an energy balance on the system shown in Fig. 3-16. Conservation
of energy requires that the change in potential energy is equal to the change in
kinetic energy. In this case, the change in potential energy is simply the elevation change in the mass (M x G x D). The overall change in kinetic energy is composed of a change in translational energy of the falling mass (M x V2/2), plus the
change in the rotational kinetic energy of the rotor (Jmass 2/2). These traditional
physical concepts may be represented mathematically in the following manner:
2
2 J
MV
mass
M G D = ------------------- + ----------------------------2
2
(3-37)
The velocity V in equation (3-37) represents the average velocity of the falling mass M. In all cases, the falling weight W is equal to the mass M times the
acceleration of gravity G. Substituting the quantity W/G for the mass M produces the following:
91
2
2 J
WV
mass
W D = ------------------- + ----------------------------2
2G
2
2
WV
2W D = ------------------- + J mass
G
2
2
WV
J mass = 2W D ------------------G
This may now be solved for the mass polar moment of inertia Jmass as:
2
W
V
J mass = -----2- 2D -------
G
(3-38)
The average velocity V of the mass M during free fall may be determined
from the fundamental equations of motion for rectilinear motion with constant
acceleration. More specifically, all physics books agree that the average velocity
multiplied times the drop time T will determine the fall distance D as follows:
Vo + V
D = ------------------ T
2
(3-39)
If the initial starting velocity Vo is equal to zero, then equation (3-39) may
be simplified to represent the velocity V in terms of the drop distance D, and the
total elapsed drop time T as shown in the next equation:
2D
V = -------T
(3-40)
The last conversion to be performed consists of an expression for the angular velocity of the cylinder in terms of the known experimental parameters.
Since tangential velocity V divided by the radius Rs is equal to the angular velocity , equation (3-40) may be used to determine in the following manner:
V
2D
= ------- = ----------------Rs
Rs T
(3-41)
Equation (3-38) for the polar inertia may now be clarified into the known
experimental data by substituting equations (3-40) and (3-41) back into equation
(3-38), and performing the following simplification of terms:
92
Chapter-3
2
W
V
J mass = -----2- 2D -------
G
W
(2 D T )
J mass = ------------------------------------------2- 2D ------------------------------
G
( 2 D Rs T )
2
J mass
2
W Rs T
4D
2D
= -------------------------------
----------------
2
2
4D
GT
2
2
2
T
4D
T
2
J mass = W R s 2D ----------------2 -----------------2 ----------------2
4D GT
4D
The cancellation of common terms results in equation (3-42).

2
2 T
1
J mass = W R s -------- ----
2D
G
where:
Jmass
W
Rs
D
T
G
=
=
=
=
=
=
(3-42)
Mass Polar Moment of Inertia (Pounds-Inches-Seconds2)

Falling Weight (Pounds)
Shaft Radius (Inches)
Drop Distance (Inches)
Drop Time (Seconds)
With a few minor modifications, this traditional physics experiment may be

used to determine the polar inertia of a complete rotor assembly. One implementation of this concept is the technique by Michael Calistrat10 that is illustrated
in Fig. 3-17. This diagram depicts a rotor resting in the rollers of a two bearing
shop balancing machine. To minimize errors, the rotor assembly should be properly balanced, and the rollers in the balancing machine should be in good condition. As shown in Fig. 3-17, an overhead pulley is supported by an external
structure, or the overhead crane that is used to move rotors in and out of the balancing machine. The diameter of the overhead pulley does not matter, but this
pulley should be in good condition, and it should turn easily. The pulley redirects
the gravitational force to pull upward on the rotor, but it does not diminish the
effect of the falling mass. A plastic covered braided steel cable is attached to the
shaft with Duct tape at a radius of Rshaft, and wrapped around the shaft for ten
or fifteen turns. Its probably a good idea to keep the cable away from any of the
rotor bearing journals, or the rotor sections observed by proximity probes. The
thickness of this cable is generally small compared to the diameter of the shaft.
Again, as shown in Fig. 3-17, the steel cable passes overhead, through the pulley,
and then is secured with a pair of vertical weights. The first weight is a tare
10 Michael M. Calistrat, Flexible Couplings, their design selection and use, (Houston: Caroline
Publishing, 1994), p 464.
93
Overhead Pulley
Fig. 317 Mechanical Test

Arrangement For Polar
Moment Of Inertia Measurement Of Full Rotor Assembly Mounted In A Shop
Balancing Machine, Technique by M. M. Calistrat
Shaft With
Radius (Rshaft)
Cable With
Radius (Rcable)
Axial Shaft
Centerline
ted in e
Moun
hin
Rotor ncing Mac
e
in
Mach aring Bala
e
Two B
Weight (W)
Tare Weight
weight that is used to counterbalance the rotor, and keep the cable taunt. This
tare weight should just barely allow the rotor to rotate, and overcome system
friction.
The tare weight is determined by trial and error, and small changes in this
weight will have a big influence on the rotor. Since this work is performed in a
balancing machine, it makes sense to use balancing clay to get a good tare
weight established. The second weight W is the experimental weight that actually turns the rotor. It may be compromised of balancing clay, a stack of washers,
or any reasonable combination thereof. In all cases, the diagnostician must have
the ability to accurately weigh this mass W at the end of the test.
As with the two previous inertia measurements, the majority of the time is
spent in setting up the test. The test execution requires minimal time, and this
rotor inertia procedure is no exception. Following the setup per Fig. 3-17, the
cable is wound up on the shaft, and the weight is naturally raised in elevation.
At this time the shaft is released, and the time T for the weight to fall through a
predetermined distance D is measured with a stopwatch. The inertia calculation
described by equation (3-42) may be used directly. However, improved accuracy
will result by including the cable radius Rcable with the shaft radius Rshaft as recommended by Calistrat. This addition is not required for very thin cables, but it
does improve accuracy for thicker cables, since the moment arm is really the sum
of the shaft plus the cable radius. Hence, (3-42) may be rewritten as follows:
2
2
Time
1
J mass = Weight ( R shaft + R cable ) ----------------------------------- ----
2
Dis
tan
ce
G
(3-43)
The nomenclature in equation (3-43) has been expanded to be more understandable, and yet maintain consistency with the general terms used within this
text. Again, the experimental weight W is in pounds, the shaft radius Rshaft, the
94
Chapter-3
cable radius Rcable, and the fall Distance are all in inches. The measured Time for
the free fall through the predetermined distance is defined as the time in seconds, and G is the gravitational constant.
As with any experiment, the validity should be checked with a test mandrel
of known characteristics. For this test, a section of 4140 steel was selected with a
density of 0.283 Pound/Inch3. The length L of the specimen was 59.81 inches, and
the ends were squared up in a lathe. In addition, the outer diameter D of the test
piece was cleaned up to nominally 5.980 inches across the length of the element.
Once more equation (3-7) is used to compute the mandrel weight:
2
LD
W solid = ------------------------------------4
2

W solid = --------------------------------------------------------------------------------------------------------------------------------- = 475 Pounds
4
This calculated weight of 475 pounds was lower than the uncalibrated shop
scale reading of 490 pounds. Although a better agreement in weights would have
been comforting, the 15 pound deviation was considered to be within the measurement accuracy of the shop scale. Next, the mass polar moment of inertia of
this steel shaft may be computed with equation (3-32) in the following manner:
3
J mass
solid
LD
= ------------------------------------32 G
solid
59.81 In. 0.283 Pounds/In. ( 5.98 In. )

2
= ------------------------------------------------------------------------------------------------------------ = 5.504 Pound-Inch-Second
2
32 386.1 Inches/Second
3
J mass
Prior to testing, the runout (eccentricity) along the length of the mandrel
was confirmed to be less than 1.0 Mil. In addition, a check balance was performed, and the residual unbalance was minor. For the actual test runs, a fall
distance of 60 Inches was established on a vertical reference stand. The shaft
radius Rshaft was 2.99 Inches, and the cable diameter was 1/8, for a Rcable of 0.06
inches. The tare weight was found to be 0.696 Pounds (315.7 grams). The experimental weight W was adjusted until the free fall time was about 5 seconds. This
required a total of 3.43 Pounds (1,556 grams). The timed test data was obtained
by releasing the shaft and measuring the time required for the weight to fall 60
inches. A stopwatch was used to measure the time, and the test data is follows:
Falling Weight Run #1
Total Time =
4.52 Seconds
4.54 Seconds
4.65 Seconds
4.54 Seconds
4.56 Seconds
22.81 Seconds for 5 Drops
Average Time for 1 Drop of 60 Inches = 4.562 Seconds
95
Eight different test runs were made, and the obviously incorrect times were
discarded. The above array was quite consistent, and there was good confidence
in both the validity and the accuracy of this experimental data. Hence, equation
(3-43) was now applied to compute the measured mass polar moment of inertia:
2
2
Time
1
J mass = Weight ( R shaft + R cable ) ----------------------------------- ----
2
Dis
tan
ce
G
2 ( 4.562 Seconds )
1
J mass = 3.43 Pounds ( 2.99 + 0.06 Inches ) ---------------------------------------- -----------------------------------------
2
2
60
Inches
386.1 Inches/Sec
J mass = 3.43 9.302 ( 0.1734 0.0026 ) = 5.450 Pound-Inch-Second
Note that there is excellent agreement between the experimental inertia of

5.450 and the calculated polar inertia of this steel shaft of 5.504 Pound-InchSecond2. The actual deviation of 1% is considered to be quite acceptable accuracy
for this type of measurement. It is often desirable to perform this test with different weights, and different fall distances to verify the consistency of the procedure. In all cases, the test weight should be substantially larger than the tare
weight, and please recall that the initial starting velocity is zero at the start
time. For bigger and heavier rotors that have a tendency to sag within the balancing machine, consideration should be given to rolling the rotors for several
hours before attempting this type of inertia test. In addition, dial indicator measurements should be made at the rotor midspan before beginning the inertia
tests, and after the conclusion of the last run. Any appreciable sag to the shaft of
the rotor under test might negate the test results, and require a repeat of the
slow roll and the test procedure.
Finally, it should be recognized that other experimental procedures exist for
determination of mass moment of inertia. In addition, many computation programs provide the capability for three-dimensional calculations of the inertia
properties of complex bodies. As always, the machinery diagnostician should be
aware of the engineering units, and test cases of known geometries should
always be run to verify the technique.
96
Chapter-3
DAMPING INFLUENCE
Three basic types of damping occur in a machinery system. These damping
types are commonly referred to as viscous damping, coulomb damping, and solid
or structural damping.
Viscous damping is encountered by solid bodies moving through a viscous
fluid. In this type of damping, the resistance force is proportional to the velocity
of the moving object. As an example of viscous damping, consider the situation of
a cook stirring a pot of soup versus a pot of molasses. It is self-evident that stirring the molasses is considerably more difficult due to the thickness and higher
viscosity of the molasses as compared to the thin soup. The required force is
directly proportional to the velocity of the stirring spoon. In most cases the cook
would stir the molasses at a much slower rate than he would the soup, simply
because it would take too much strength or energy to stir it rapidly.
The same type of physical property, i.e., viscous damping, is encountered in
the bearings and oil seals of large rotating machines. In this case, the damping is
provided by the lubricating oil, and the rotating shaft is the rigid body moving
through the viscous fluid. The process fluids handled by the machine also provide
damping to the rotor system. For liquid handling machines such as pumps and
hydraulic turbines, this is significant. However, for gas handling machines such
as turbines or centrifugal compressors this is a minor consideration.
0
Phase Lag (Degrees)
Low Damping, =0.1

30
60
90
High Damping,
=2.0
=2.0
120
=1.0
150
=0.2
=0.5
=0.1
180
Low Damping, =0.1

Amplitude Ratio
Fig. 318 Calculated

Bode Plot Of Unbalance
Mechanical System With
Variations In Damping
=0.2
3
2
=0.5
0
0.0
=1.0
0.5
1.0
1.5
=2.0
2.0
2.5
3.0
Damping Influence
97
The next type of damping is coulomb damping, which arises from the
sliding of one dry surface upon another (rub condition). The coulomb friction
force is nearly constant, and it depends on the nature of the sliding surfaces, and
the perpendicular pressure between the surfaces. This type of force is generally
dominant in damped systems during the final stages of motion when other types
of damping become negligible.
The third category of damping is often referred to as solid or structural
damping. This is due to internal friction within the material, and it differs from
viscous damping in that it is independent of frequency, and proportional to the
maximum stress of the vibratory cycle. Since stress and strain are proportional
in the elastic range, it can be stated that solid damping force is proportional to
deflection. Structural damping in rotating machinery is small when compared
with viscous damping, but it does exist.
The major contribution that positive damping makes to a rotating machinery system is the dissipation of energy. This influence is most dramatically illustrated when a mechanical system passes through a resonance as in Fig. 3-18.
This calculated Bode diagram was duplicated from chapter 2. The unbalance
response plot of frequency ratio versus amplitude ratio and phase of a damped
system provides a good perspective of the actual influence of damping. The family of curves in this diagram are plotted with a damping ratio (or damping factor)
extending from =0.1 to 2.0. Recall that this ratio is the actual damping divided
by the critical damping. Note, that with a lightly damped system of =0.1, the
response at the resonance is quite high. This translates to the fact that there is
little energy dissipation under this condition. The system is under damped, and
it is susceptible to instability due to a lack of an energy dissipation. Conversely,
when the damping factor is large, =2.0, the system is over damped, response
through the resonance is restrained, and overall stability of the system is high.
In many instances of free vibration, rotor instability can be related to a lack
of damping. The system damping may be assessed by examining the critical
speed response on Bode plots (synchronous 1X vectors versus speed). Typically,
the amplification factor Q through the resonance is used to quantify the severity
of the resonance, plus the damping ratio. A large amplification factor is associated with a poorly damped, high amplitude resonance. Conversely, a low amplification factor is generally associated with a well damped resonance, that displays
small amplitudes at the peak of the resonance.
Extraction of the amplification factor from the Bode plots may be performed
in several different ways. Unfortunately, there is disagreement within the technical community as to the best manner to obtain this information from the measured vibration data. For the sake of completeness, three separate approaches
for determination of this dimensionless amplification factor will be presented.
The first technique consists of visually comparing the measured vibration
response data with a set of calculated curves, and estimating a damping ratio .
The rotor amplification factor for the specific resonance is then computed directly
from the damping ratio. As a practical example of an industrial machine, consider the compressor Bode plot shown in Fig. 3-19 (same as Fig. 2-21). This variable speed vibration data is runout compensated, and it exhibits a clean
98
Chapter-3
transition through a single critical speed. If this data is viewed in conjunction

with the computed family of curves shown in Fig. 3-18, a suitable damping ratio
may be selected. Specifically, the shape of the amplitude versus speed, and the
Fig. 319 Measured Bode

Plot Of A Centrifugal Compressor Startup
phase versus speed are compared between the measured (Fig. 3-19) and the calculated plots (Fig. 3-18). It is reasonable to conclude that the calculated plot with
a damping ratio of 0.2 is the closest match to the measured machine response
data. From the previously developed equation (2-91), the amplification factor
(i.e., amplitude ratio) Q may be determined from the damping ratio as follows:
Q Curve
Fit
1
= Amplification Factor = Amplitude Ratio = -----------2
(3-44)
Substituting the previously identified value for the damping ratio of =0.2,
the following result is obtained:
Q Curve
Fit
1
1
1
= ------------ = ---------------- = ------- = 2.5
2
2 0.2
0.4
This visual comparison between curve shapes provides any easy way to
estimate Q for a cleanly defined resonance. However, for more complex response
characteristics, other techniques are available. For instance, the second
approach for computation of the amplification factor is derived from electrical
engineering terminology. This technique has also been adopted by various
mechanical standards organizations such as the American Petroleum Institute.
In this procedure, the center frequency of the resonance is divided by the resonance bandwidth at the Half Power Point in accordance with the following:
Q Half
Power
Center Frequency
= --------------------------------------------------------------------------------------Frequency Bandwidth @-3dB
(3-45)
Damping Influence
99
The half power point is equivalent to the amplitude that is -3 dB down from
the peak of the resonance. In terms of linear scales, an amplitude change of -3 dB
is equal to 0.707 times the peak vibration amplitude. This level defines the specific point on the resonance response curve where the frequency bandwidth is
measured. Although this calculation procedure may sound complicated, the
actual execution is fairly straightforward.
If the Bode plot from the last example is reconsidered, the amplification factor using this half power approach may be computed as shown in Fig. 3-20.
Within this rendition of the Bode, various features have been identified to
Fig. 320 Amplification

Factor From A Bode Plot
Using the Half Power Point
Calculation Technique
enhance this discussion. Specifically, the center of the resonant peak (critical
speed) has been identified at 6,100 RPM. The vibration amplitude at the peak of
this resonance is 1.95 Mils,p-p. The half power point is determined by multiplying
0.707 times the maximum of 1.95 Mils,p-p to obtain a -3 dB point of 1.38 Mils,p-p.
The response curve width at this amplitude is 900 RPM. Stated in another way,
the frequency equal to an up slope amplitude of 1.38 Mils,p-p is 5,800 RPM. The
frequency equal to a down slope amplitude of 1.38 Mils,p-p is 6,700 RPM. The
bandwidth at this -3 dB amplitude is 6,700 minus 5,800, which is equal to 900
RPM. Substituting these values for center frequency and bandwidth into the previous equation (3-45) the next result is obtained:
Q Half
Power
6, 100 Rpm
Center Frequency
= --------------------------------------------------------------------------------------- = --------------------------- = 6.8
Frequency Bandwidth @-3dB
900 Rpm
The calculated amplification factor using this second approach (6.8) is considerably higher than the first method that produced a Q of 2.5. However, this
second method may be used for complex resonant response conditions where
multiple critical speeds appear close together. For instance, a machine that has a
split critical (e.g., horizontal followed by a vertical mode) would be difficult to
100
Chapter-3
handle with the first approach. Whereas, the second scheme would allow a better
quantification of the Q for each response peak.
One objection to the second approach is that Q varies with changes to the
center frequency of the resonance. For instance, if the shape of the response
curve is maintained, and the center frequency is reduced to 3,000 RPM, the Q
drops to 3.3 (=3,000/900). By the same token, if the critical speed occurs at
12,000 RPM, the Q now increases to 13.3 (=12,000/900). This change in amplification factors with a constant shape to the response peak may be quite confusing, as well as contradictory to the desired definition.
The third approach for calculation of the amplification factor consists of
dividing the amplitude at the resonance by the amplitude at a speed far above
the resonance. Again referring to the same Bode plot example, this approach is
illustrated in Fig. 3-21. As before, the magnitude of 1.95 Mils,p-p at the transla-
Fig. 321 Amplification

Factor from a Bode plot
Using A Simple Amplitude
Ratio Between the Resonance and a Speed Far
Above the Resonance
tional critical speed of 6,100 RPM is identified. Above this resonance, the phase
and amplitude enter a plateau region where the 1X vector remains reasonably
constant as speed increases. This is normal behavior, and the synchronous
response above a critical is typically flat until some external force, or a higher
order resonance influences the motion. Within the context of this example, the
region at 8,000 RPM is sufficiently removed from the critical speed, and the rotational speed vibration amplitude in this region is 0.85 Mils,p-p. The amplification
factor determined with this third technique is computed by simply dividing the
maximum vibration amplitude at the critical speed by the amplitude measured
well above the resonance in accordance with the following expression:
Amplitude At Resonance
Q Resonance/Above = --------------------------------------------------------------------------Amplitude Above Resonance
(3-46)
Damping Influence
101
Using the vibration amplitudes identified on Fig. 3-21, the amplification

factor determined by this method is easily computed as:
Amplitude At Resonance
1.95 Mils p p
Q Resonance/Above = --------------------------------------------------------------------------- = ----------------------------- = 2.3
Amplitude Above Resonance
0.85 Mils p p
This value of 2.3 is indicative of an adequately damped machine, and a stable mechanical system. Due to the differences between machines, it is difficult to
precisely categorize the variations between good and bad Q factors. However, it
is reasonable to identify several categories of amplification factors versus damping and stability characteristics. Summary Table 3-1 attempts to provide some
realistic guidelines for these interrelated parameters:
Table 31 Shaft Amplification Factor Versus Damping And Anticipated Rotor Stability
Amplification Factor
Damping
Stability
Q=<2
Well Damped
Extremely Stable
Q = 2 to 8
Adequately Damped
Normal Stability
Q = 8 to 15
Poorly Damped
Marginal Stability
Q = > 15
Insufficient Damping
Inherently Unstable
Machines that fall into the last category may be potentially dangerous, and
may not survive a maiden startup. Machines within this group often require
extensive modifications to bearings and/or seals to increase system damping. In
some instances, modifications such as the installation of squeeze film damper
bearings may be required to provide adequate damping for the system.
It should also be noted that the amplification factors referred to herein are
associated with shaft vibration measurements where the oil film viscous damping is dominant (i.e., between the journal and bearing). Casing vibration measurements would typically be more receptive to structural damping, and not so
sensitive to viscous damping in the bearings. Thus, casing measurements generally display Qs that are much higher than shaft measurements, simply due to
the lack of damping within the casing and the support structure.
Another consideration that must be applied to any evaluation of variable
speed data is the acceleration rate, or more specifically the rate of speed change
of the rotating system. During machinery train coastdowns, there is usually minimal, if any, control of the deceleration. However, during startup, the rate of rotor
acceleration is often controllable on variable speed drivers such as steam turbines. On machines with older control systems, the startup rate is often dependent upon the skill and knowledge of the operator handling the trip & throttle
valve. On newer speed control systems, the startup rate is usually controlled by
an electronic governor with predetermined startup speed ramps. Unfortunately,
some electronic governors suppliers are not well versed in the acceptable startup
rates for various types of machinery trains. These vendors often set abnormally
102
Chapter-3
fast startup rates in an effort to snap the rotor(s) through a critical speed range.
This tendency has resulted in machinery damage, and more than one rotor failure. Hence, the end user should always verify that the proposed startup acceleration rates are reasonable for the machinery in question.
Intuitively, the passage of a rotor system through a critical speed region
should be performed in a direct and knowledgeable manner. If the startup rate is
inordinately slow, the machine may hang up in a resonance, and cause mechanical damage due to the high vibration levels. An example of this type of occurrence is briefly discussed in the turbine generator case history 39 in chapter 11.
On the other hand, if the speed acceleration rate through a resonance is excessive, the machine may self-destruct after it reaches operating speed and
attempts to rebalance itself about the mass center. Although this type of occurrence is rare, it is certainly avoidable, and totally unnecessary.
The speed transition rate through a resonance will alter the characteristics
of the vibration response data. A slow startup will show a higher peak at the resonance, combined with a broader bandwidth. Conversely, a rapid startup will
produce a lower peak amplitude at the critical speed, plus a smaller resonant
bandwidth. This attenuated response characteristic has erroneously led many
individuals into a false sense of security by ramping through critical speeds at a
high rate. Hence, the diagnostician must be aware of the transition rate through
the system critical speed(s), and any evaluation of the resonant characteristics
(e.g., Q) should be weighed by this speed change rate.
It is impossible to fully quantify proper startup acceleration rates for all
classes of machinery operating with various types of drives, and control systems.
However, to provide some guidance in this area, Table 3-2 of acceptable startup
rates is offered for consideration. This table summarizes field measurements on
a variety of variable speed trains. In all cases, the general machine type is indicated, the rotor weight is shown, and the maximum machine speed is listed. The
last column provides typical peak startup acceleration rates (generally through
Table 32 Various Machine Types Versus Acceptable Startup Acceleration Rates
Rotor Weight
Maximum Speed
Acceleration Rate
(Pounds)
(RPM)
(RPM/Second)
Expander/Pinion
60
34,000
550 to 750
Pinion/Compressor - Small
300
14,000
350 to 450
Pinion/Compressor - Medium
1,500
11,000
80 to 100
Steam Turbine/Compressor
7,500
8,000
70 to 80
Gas Turbine/Compressor
25,000
5,400
20 to 30
29,000
3,800
40 to 50
54,000
4,000
35 to 40
Hydro Turbine/Generator
180,000
420
5 to 10
Machine Type
Damping Influence
103
resonances). It should be recognized that other portions of the startup sequence

often occur at much lower rates. Thus, the compressor that exhibits a 400 RPM
per Second ramp through a critical may be preceded by a region with a rate of
only 20 to 30 RPM per Second.
Table 3-2 represents conservative speed acceleration rates. The observed
trend of light weight, high speed rotors that accelerate rapidly is considered to be
reasonable. In addition, large rotors that run at slow speeds exhibit understandably lower rates. This table is only provided for reference purposes based upon a
finite set of measurements. For specific recommendations of acceleration rates
for defined machinery trains, the Original Equipment Manufacturer (OEM)
must always be contacted.
Throughout this section on damping influence, the emphasis has been
placed on the effects of damping upon a mechanical system. The changes of rotor
amplification factor, and the changes in stability due to variations in damping
are generally understood. For simple cases, it has been demonstrated that the
theory and the actual machinery behavior are quite compatible. In all cases, the
machinery diagnostician is encouraged to simplify problems to the greatest
extent possible, and to strive towards straightforward explanations and solutions. Unfortunately, some mechanical systems may not cooperate with this
approach, and the complexity of the problem plus the potential solution may very
well exceed the explanations offered by simple systems.
When faced with difficult technical problems, the traditional engineering
mentality often attempts to fully quantify and define all variables associated
with the problem. With respect to damping characteristics, this is difficult to
accomplish. For example, John Vance11 states that Accurate values for the damping coefficients are usually difficult to obtain. The author has found from experiments that an all-steel rotor/shaft assembly with tight fits and no gear backlash
will have about 1.52.0 percent of critical damping, not including friction between
the rotor and ground. This statement reinforces the difficulty in determination
of actual damping coefficients. In many instances, the damping characteristics
are either ignored or determined experimentally.
There is experimental information published on specific damping materials
used for vibration and shock isolation. Technical references such as the Shock
and Vibration Handbook12 devote several chapters to isolators, and the characteristics and utilization of damping materials. This information is applicable to
machines such as air blowers and refrigeration compressors that are installed on
large buildings. For the comfort of the inhabitants of these buildings, the
machines are normally isolated from their respective structures. However, on
large industrial machines, the foundations and supporting structures are normally rigid, and external damping materials are not used. On large gas handling
process machines, most of the damping is provided by the bearing lubricating oil,
11 John M. Vance, Rotordynamics of Turbomachinery, (New York: Wiley-Interscience Publication, 1988), p 61.
1996), chapters 32, 33, 36 and 37.
104
Chapter-3
with only a minor contribution from the shaft oil seals. The viscous damping
associated with the bearings normally dominates any coulomb and structural
damping in the system. The engineering values for bearing damping are usually
calculated with analytical programs, as discussed in chapter 5. Structural or
support damping is often determined experimentally, and this topic is addressed
in the following chapter 4.
Although quantification of discrete damping values is necessary from an
initial design standpoint, it may not be particularly useful to the diagnostician
attempting to solve a field problem. In most cases, a working knowledge of the
influence of damping upon the machinery, plus an understanding of the contributing parameters that are involved in determination of the damping are the
most important items. To this extent, it is mandatory for the diagnostician to
understand that bearing damping varies with speed, applied load, bearing geometry, plus the lubricant characteristics. This knowledge may allow the development of a logical thought process during the analysis of a machinery problem.
For instance, if a compressor exhibits symptoms of decreased damping, such as
higher vibration amplitudes passing through a critical speed, one mechanism
that might considered would be the influence of expanded bearing clearances
upon the damping. Of course, increased bearing clearances might also produce
significant changes in the bearing stiffness, which would probably change the
center frequency of the measured critical speed. Hence, the influence of stiffness
upon the machinery behavior must also be addressed, and the following section
addresses the fundamentals of the stiffness aspects of machinery supports.
Stiffness Influence
105
STIFFNESS INFLUENCE
The stiffness of a mechanical system is the spring like quality of mechanical elements to elastically deform under load. In essence, the application of a
force measured in pounds will produce a deflection measured in inches. Thus, the
stiffness of a mechanical element carries the traditional engineering units of
Pounds per Inch. Mechanical properties such as the dimensions and weight of an
object have a physical meaning for most individuals. However, a characteristic
such as stiffness is commonly used, but it is seldom associated with anything of
real physical significance. To provide a sense of magnitude to stiffness values,
consider the following derivation of axial, radial, and torsional stiffness for a
series of identical cylinders.
Length (L)
Deflection (L)
Axial Force
(Fax )
Fig. 322 Solid Cylinder

Mounted On Infinitely
Rigid Base And Subjected
To Axial Compression
Diameter (D)
The diagram presented in Fig. 3-22 describes a solid metal cylinder with a
diameter D, and a length L. Assume that the bottom of this cylinder is rigidly
supported, and consider the application of a pre axial compressive force Fax. Further assume that the applied force is sufficient to cause an elastic deformation of
the cylinder, and that the amount of this deformation is described by the deflection L. From fundamental strength of materials it is known that:
L
Deflection
Strain = ------------------------------ = -------Length
L
(3-47)
F ax
4 F ax
Force
Stress solid = -------------------------- = ------------------------ = ------------------2
2
Area solid
D 4
D
(3-48)
It is also known that within the elastic region for the material, stress and
strain are related by the modulus of elasticity E as follows:
106
Chapter-3
Stress solid
Modulus of Elasticity = E = -----------------------------Strain
(3-49)
Substituting equations (3-47) and (3-48) into (3-49) produces the following:
4 F ax
-----------------2
4 F ax L
D
E = -------------------------- = ------------------------------2
D L
L
-
------ L
(3-50)
In accordance with the initial definition for stiffness, the axial stiffness of
this metal cylinder may be determined by dividing the applied force Fax by the
deflection L. Hence, it may be properly stated that:
Kax
solid
F ax
= ---------L
(3-51)
By combining (3-50) and (3-51), the axial stiffness may be included as:
4 L F ax
4L
E = ----------------2- ---------- = ----------------2- K a x
solid
L
D
D
By rearranging terms, (3-52) for axial stiffness of a solid cylinder evolves:

2
Kax
solid
ED
= --------------------------4L
(3-52)
Equation (3-52) makes sense from the standpoint of the contribution of the
terms. Specifically, a large diameter cylinder will be stiffer than a smaller diameter cylinder, and a short cylinder should be stiffer than a tall one. To provide a
physical representation of the meaning of various stiffness values, consider solving the last expression for the diameter as follows:
L Kax
solid
D = 2 -----------------------------E
(3-53)
If the cylinder material is steel, the modulus of elasticity E would be in the

vicinity of 30,000,000 Pounds/Inch2. If the total length of the cylinder L is
assumed to be 30 inches, then (3-53) may be simplified to the following format:
30 Inches K a x
Kax
solid
solid
D = 2 --------------------------------------------------------------= ----------------------6
2
886.2
30 10 Pounds/Inches
With this expression, the equivalent cylinder diameter may be computed
for various stiffness values. For instance, if the axial stiffness of a solid cylinder
Stiffness Influence
107
is 1,000,000 Pounds/Inch, the required diameter is calculated as follows:

6
Kax
1 10
solid
D = ----------------------- = ------------------- = 1.1284 Inches
886.2
886.2
If this calculation is repeated for a series of stiffness values, the resultant

diameters may be plotted against stiffness. Another way to view this information
is shown in Fig. 3-23. Within this diagram, a series of consecutive cylinders are
drawn that represent stiffness values ranging from 1,000,000,000 Pounds/Inch
(i.e., 109) with an associated diameter of 35.682 inches to a minimum stiffness
of 100,000 Pounds/Inch (i.e., 105) with a computed diameter of 0.3568 inches.
Diameter = 35.682"
11.284"
3.568"
1.128"
Axial Stiffness
= 109 Pounds
Inch
108 Lbs.
In.
106 Lbs.
In.
107 Lbs.
0.85 Pounds
8.5 Pounds
85 Pounds
849 Pounds
Cylinder Weight =
8,489 Pounds
Height = 30.00"
0.357"
105 Lbs.
In.
In.
Fig. 323 Axial Stiffness Of 30 Inch Tall Steel Cylinders Of Various Diameters
Fig. 3-23 is a rendition of a similar explanation of stiffness presented by

F.L. Weaver13, in his paper entitled Rotor Design and Vibration Response.
Although Weaver did not present the method for computing the stiffness of the
cylinders his diagram of spring gradients of 30 inch tall solid steel cylinders
on an infinitely stiff support are equivalent to the drawing in Fig. 3-23.
If the cylinders subjected to axial compression are hollow instead of solid,
the previously developed equation structure may be easily modified by incorporating an inner diameter into equation (3-48) as follows:
4 F ax
F ax
Force
Stress annulus = ----------------------------------- = ------------------------------------ = ----------------------------------2
2
Area annulus
2
2
( Do Di )
Do Di
------- -------
4
4
(3-54)
13 F.L. Weaver, Rotor Design and Vibration Response, Proceedings of the First Turbomachinery Symposium, Gas Turbine Laboratories, Texas A&M University, College Station, Texas, (1972),
pp. 142-147.
108
Chapter-3
In equation (3-54), the cylinder outer diameter is designated by Do, and the
hollow inner diameter is once more identified as Di. The strain equation (3-47)
may now be combined with equation (3-54) in the traditional stressstrain relationship described by equation (3-49) to produce the following result:
4 F ax
---------------------------------2
2
( Do Di )
Sress annulus
4 F ax L
E = ------------------------------------ = ------------------------------------------ = ------------------------------------------------2
2
Strain
( D o D i ) L
L
------ L
(3-55)
As before, the general expression for the spring constant (Fax/L) may be
inserted into (3-55) to establish an expression that includes the physical dimensions, the modulus of elasticity, plus the axial stiffness of a hollow cylinder:
F ax
4L
4L
- ---------- = ----------------------------------- K a x
E = ---------------------------------2
2 L
2
2
annulus
( Do Di )
( Do Di )
This equation may now be solved for the axial stiffness of a hollow cylinder:
2
Kax
annulus
E ( Do Di )
= ---------------------------------------------4L
(3-56)
Equation (3-56) may be used to compute the axial stiffness of a hollow cylinder. It is clear that if the internal diameter is set equal to zero, then (3-56)
reverts back to (3-52) for a solid cylinder. From this pair of stiffness equations, it
is noted that a hollow cylinder will always have an axial stiffness that is less
than an equivalent solid cylinder of equal outer diameter.
This array of cylinders shown in Fig. 3-23 helps to visualize stiffness values
for various physical dimensions. However, it should be recalled that these values
are predicated upon a cylinder placed in axial compression. It is also meaningful
to examine the radial or lateral stiffness for this same group of steel cylinders.
For instance, assume that a cylinder is positioned horizontally as shown in Fig.
3-24. Further assume that the cylinder is simply supported at each end, and that
a radial midspan force Frad is applied. As before, the dimensions of the cylinder
are the diameter D, and the length L. Furthermore, the maximum midspan
radial deflection R of the cylinder is identified on Fig. 3-24.
The radial cylinder stiffness is the applied force Frad divided by the deflection R. Even though Fig. 3-24 depicts a radial instead of an axial deflection, the
same physical laws apply. Hence, it may be properly restated that the radial
stiffness is equal to the applied force Frad divided by the deflection R as shown:
F rad
K rad = ------------R
(3-57)
Stiffness Influence
109
Diameter (D)
Radial Force
(Frad)
Fig. 324 Simply Supported

Solid Cylinder Subjected To
A Midspan Radial Force
Radial Deflection (R)

Length (L)
The lateral deflection of this cylinder may be determined from simply supported beam formulas. From Roark14, or Spotts15, the maximum midspan deflection may be computed from the following common expression:
3
F rad L
R = ------------------------------------48 E I area
(3-58)
Combining the previous equations (3-57) and (3-58), and solving for the lateral (radial) cylinder stiffness Krad, the next expression is formed:
F rad
F rad
K rad = ------------- = -----------------------------------------3
R
F rad L
------------------------------------
48 E I area
This is simplified to the following common format:
48 E I area
K rad = ------------------------------------3
L
(3-59)
Equation (3-59) is often used for the calculation of the shaft stiffness. Since
the element under load has a circular cross section, the area moment of inertia
Iarea for a solid may be determined from (3-30), and incorporated into the stiffness equation (3-59) to obtain the following result:
K rad
solid
4
48 E I area
48 E D
solid
----------------= --------------------------------------------
- = ---------------3 64
3
L
L
110
Chapter-3
This simplifies to the next common equation for lateral shaft stiffness:
4
K rad
solid
3ED
= ------------------------------------3
4L
(3-60)
If the length L remains at 30 inches, and the modulus of elasticity for steel
is 30,000,000 Pounds/Inch2, equation (3-60) may be solved for the stiffness of
various cylinder diameters. For consistency, the previously calculated cylinder
that had an axial stiffness of 106, was computed to have a diameter of 1.1284
inches. The lateral stiffness for this same cylindrical element is as follows:
6
3 30 10 Pounds/Inch ( 1.1284 Inches )

= ------------------------------------------------------------------------------------------------------------------ = 4, 240 Pounds/Inch
3
solid
4 ( 30 Inches )
Repeating this calculation for cylinder diameters of 35.682, 11.284, 3.5682,
and 0.3568 inches respectively, the summary diagram of Fig. 3-25 may be constructed. In this drawing, the cylinder dimensions are identical to the values previously determined from the axial load case. However, it is apparent that the
radial stiffness magnitudes have changed significantly from the axial stiffness
values shown in Fig. 3-23.
2
K rad
0.357"
1.128"
0.85 Pounds
8.5 Pounds
Inch
4,240 Pounds
Inch
4.24 x 105 Pounds
3.568"
85 Pounds
11.284"
42.4 Pounds
849 Pounds
Inch
4.24 x 107 Pounds

Inch
Diameter = 35.682"
Length = 30.00"
Cylinder Weight=
8,489 Pounds
Radial Stiffness =
4.24 x 109 Pounds
Inch
Fig. 325 Radial Stiffness Of 30 Long Steel Cylinders Of Various Diameters
For the case of a hollow cylinder, the radial stiffness is determined by

installing the area moment of inertia for an annulus, equation (3-23), into the
general radial stiffness equation (3-59) in the following manner:
Stiffness Influence
K rad
annulus
111
4
4
48 E I area
48 E ( D o D i )
annulus
----------------------------------
= --------------------------------------------------
- = ---------------
3
3
64
L
L
Simplifying and combining terms, the expression (3-61) for radial stiffness
of a hollow cylinder may be easily produced:
4
K rad
annulus
3 E ( Do Di )
= ------------------------------------------------------3
4L
(3-61)
This is similar to equation (3-60) for a solid shaft. If the inner diameter of
equation (3-61) is set equal to zero, the above expression will revert back to equation (3-60). Again, it is apparent that a solid shaft is always laterally stiffer than
a comparable hollow shaft of equal outer diameter.
In addition to the radial and axial directions, machinery shafts are also
subjected to twisting forces. For example, the normal movement of reciprocating
machines produces synchronous torsional excitations in rotating elements such
as crankshafts. Although centrifugal machines are less susceptible to torsional
motion failures do occur, and the torsional aspects of centrifugal machinery
must be considered. For either class of machine, the torque on one end of a shaft
Angular Deflection
Torque
Diameter (D)
Fig. 326 Solid Cylinder,

Fixed At One End, And
Subjected To Torque
Length (L)
may result in an angular deflection along the shaft length. The amount of twist
is directly related to the applied torque, and the torsional stiffness.
In general, torsional stiffness Ktor is analogous to the axial or lateral stiffness. However, the force is replaced by torque, and the linear displacement is
replaced by an angular deflection . Performing this parameter substitution in
(3-51), the following general definition for torsional stiffness is obtained:
Torque
K tor = --------------------
(3-62)
A typical condition for a cylinder subjected to twisting deformation is

depicted in Fig. 3-26. From strength of materials (Roark16, or Spotts17), it is
112
Chapter-3
known that the angle of twist is related to the applied Torque, the member
length L, the area polar moment of inertia Jarea, and the material shear modulus
Gshear in the following manner:
Torque L
= --------------------------------------J area G shear
(3-63)
Note that the shear modulus and the acceleration of gravity are both traditionally described by the letter G. To distinguish between these two common
terms, the shear modulus will be identified by Gshear throughout this text. It
should also be mentioned that the shear modulus is also known as the modulus
of rigidity, the modulus of elasticity in shear, and the torsion modulus. In any
case, combining equations (3-62) and(3-64), the following general expression for
torsional stiffness is obtained:
J area G shear
Torque
Torque
K tor = --------------------- = ---------------------------------------- = --------------------------------------Torque L
L
--------------------------------------J area G shear
(3-64)
The resultant JG/L expression for torsional stiffness is common within the
literature. For instance, this same equation structure is used in the ASM Handbook of Engineering Mathematics18. Within (3-64), the area polar moment of
inertia Jarea is the value calculated through the center of the member (i.e., along
the rotational axis). If the cylinder cross section is an annulus (hollow shaft) with
an outer diameter of Do, and an inner diameter of Di, the area polar moment of
inertia Jarea was previously defined by equation (3-24). Substituting (3-24) into
equation (3-64), the following torsional stiffness expression is formed:
4
K tor
annulus
G shear ( D o D i )
= ------------------ -----------------------------------
32
L
This is combined into the following common equation for the torsional stiffness of a hollow shaft (cylinder):
4
K tor
annulus
G shear ( D o D i )
= -----------------------------------------------------------32 L
(3-65)
The validity of equation (3-65) is again supported by identical expressions

for torsional stiffness of hollow circular cylinders as shown in the Shock and
Vibration Handbook19. If the shaft cross section is solid, then the inner diameter
18 William G. Belding and others, ASM Handbook of Engineering Mathematics, 4th printing,
(Metals Park, Ohio: American Society for Metals, 1989), p. 319.
1996), p. 1.11.
Stiffness Influence
113
Di is equal to zero, and equation (3-65) may be simplified to the following expression describing the torsional stiffness of a solid circular shaft:
4
K tor
solid
G shear D
= ----------------------------------------32 L
(3-66)
Once more, this expression for torsional stiffness is common within the
technical literature previously cited. From (3-66), it is evident that the torsional
stiffness is a function of the cylinder diameter D, the cylinder length L, and the
shear modulus Gshear. If the physical dimensions, and Gshear are known, then the
torsional stiffness may be computed. However, in cases when the shear modulus
is unknown, it may be computed from the modulus of elasticity E, and Poissons
ratio . This relationship is defined by Shigley20 as:
E = 2 G shear ( 1 + )
(3-67)
If the steel cylinder dimensions previously used for computation of axial

and radial stiffness are used for the calculation of torsional stiffness the first
step is to determine the shear modulus. As previously noted, the modulus of elasticity E for these steel cylinders was assumed to be 30,000,000 Pounds/Inch2. If
Poissons ratio for this material is equal to 0.3, then Gshear may be determined
from equation (3-67):
6
6
E
30 10 Pounds/Inch
2
G shear = -------------------------- = -------------------------------------------------- = 11.5 10 Pounds/Inch
2 (1 + )
2 ( 1 + 0.3 )
For a solid steel cylinder length L of 30 inches, and a diameter D equal to

1.1284 inches, the computed shear modulus may now be used to calculate the
torsional stiffness from equation (3-66) in the following manner:
6
11.5 10 Pounds/Inch ( 1.1284 Inches )

= -------------------------------------------------------------------------------------------------------------solid
32 30 Inches
58, 573, 344
Inch-Pound
K tor
= ------------------------------ Inch-Pounds = 61, 010 ---------------------------Radian
solid
960
K tor
Note that the direct engineering units for this calculation are Inch-Pounds
but the non-dimensional angular units of Radians have been included. This is
inferred by the previous equation (3-62) of Torque/Angle and it is customary
nomenclature for torsional stiffness. For numerical comparison purposes, the torsional stiffness calculation is repeated for cylinder diameters of 35.682, 11.284,
3.5682, and 0.3568 inches respectively. The selected cylinder diameters are identical to the values previously determined from the axial load case. The axial,
radial, and torsional stiffness values for each of the five cylinder diameters are
now summarized in Table 3-3.
York: McGraw-Hill Book Company, 1986), pp. 10.6.
114
Chapter-3
Table 33 Comparison of Calculated Axial, Radial, and Torsional Stiffness of 30 Inch Long
Solid Steel Cylinders of Various Diameters
Cylinder Diameter
Axial Stiffness
Radial Stiffness
Torsional Stiffness
(Inches)
(Pounds/Inch)
(Pounds/Inch)
(Inch-Pounds/Radian)
35.682
1.00 x
109
109
6.10 x 1010
11.284
1.00 x 108
4.24 x 107
6.10 x 108
3.5682
1.00 x 107
4.24 x 105
6.10 x 106
1.1284
1.00 x 106
4.24 x 103
6.10 x 104
0.3568
1.00 x 105
4.24 x 101
6.10 x 102
4.24 x
Clearly, the steel cylinder stiffness varies considerably from the axial to the
radial direction. These values are independent of load, and they are a function of
cylinder dimensions, and the modulus of elasticity. The significant point of this
exercise is that individual machine elements seldom have singular elements that
exceed the range of stiffness values presented in these physical examples. That
is, machine parts will probably not have a stiffness much greater than 108 or 109
Pounds per Inch. These are enormously rigid elements, and stiffness values of
108 are seldom encountered in rotating machinery. At the other end of the scale,
the skinny cylinder with a diameter of 0.36 Inches, and a height of 30 inches, displays an axial stiffness of only 100,000 Pounds per Inch. Again, it is hard to
imagine many load carrying members within a rotating machine with this type
of aspect ratio. Thus, it is logical to conclude that this stiffness of 105 Pounds per
Inch is a realistic minimum value. For reference purposes, Table 3-4 of typical
stiffness ranges (in Pounds/Inch) for various machine elements is presented.
Table 34 Summary of Typical Machinery Element Stiffness Values
Mechanical Element
Typical Stiffness Range

(Pounds/Inch)
Oil Film Bearings
300,000 to 2,000,000
Rolling Element Bearings
1,000,000 to 4,000,000
Bearing Housing Support - Horizontal
300,000 to 4,000,000
Bearing Housing Support - Vertical
400,000 to 6,000,000
Shaft 1 to 4 Diameter
100,000 to 4,000,000
Shaft 6 to 15 Diameter
400,000 to 20,000,000
The concept of stiffness is applicable to shafts, bearings, cases, support

structures, and foundations. In a real machine all of these stiffness parameters
play a role in determination of the final vibration response characteristics. The
relationship between effective system stiffness, and each of the individual ele-
Stiffness Influence
115
ments is that the reciprocal of the overall system stiffness is equal to the sum of
the reciprocals of each individual element. Thus, the change in any element (e.g.,
soft bearing support) can influence the entire system. It must also be recognized
that the weakest (softest) member in the rotor support system will be the dominant element in establishing overall or effective system stiffness.
The previous discussion has centered around the variations in stiffness due
to changes in geometry and materials of construction. Although this is an academically interesting topic, it is not particularly useful until it is applied within
the realm of process machinery. One of the obvious influences of stiffness is the
effect upon the natural frequency of machine elements. For discussion purposes,
consider the axial, torsional, and lateral natural frequency of an undamped hollow cylinder based upon the previously developed equations. Specifically, the
axial natural frequency may be determined from equation (2-45). The required
mass of the cylinder may be specified by equation (3-4), and the axial stiffness of
this annulus may be determined from equation (3-56). Combining these three
expressions, the following result is obtained:
Fc
axial
E
2
2
- ( Do Di )
------------4
L
K
a x annulus
1 E
4G
1
1
= ------ ------------------------- = ------ -------------------------------------------------------------- = ------ -------------- ----------------------2
4
L
2
2 M annulus
L
2
2
---------------------
(
D
D
)
4G
o
i
This equation for the axial natural frequency may be further simplified to:
Fc
axial
1 EG
= ------ ---------------2 L 2
(3-68)
This is an interesting result. The diameters have canceled out, and the only
remaining physical dimension is the length L of the cylinder. Hence, the hollow
cylinder diameters do not influence the axial natural frequency. Based on the
material properties used for the steel cylinder examples, the axial natural frequency for a 30.0 Inch long member may be computed from equation (3-68) as:
6
Fc
axial
Fc
axial
1 30 10 Pounds/Inch 386.1 Inches/Second

Cycles
= ------ --------------------------------------------------------------------------------------------------------- = 1, 073 ---------------2
3
Second
2
0.283 Pounds/Inches ( 30 Inches )
Cycles
Second
Seconds
Minute
Cycles
Minute
= 1, 073 ----------------- 60 ------------------- = 64, 400 -----------------
The computed value of 64,400 CPM is a relatively high frequency, and

would probably fall outside of the operating speed range for most machines. Similarly, if the shaft length was extended to 60 Inches, the axial resonant frequency
would be 32,200 CPM, which is still quite a high frequency for a 5 foot long shaft.
Due to this general relationship, the axial natural frequency of a shaft is seldom
within the operating speed range. Hence, there is minimal opportunity to excite
116
Chapter-3
this resonance. Although axial vibration does occur in most machines, there is
little evidence to suggest that the shaft axial resonant frequencies are commonly
excited.
As a cautionary note, the diagnostician should not ignore other longitudinal
or axial resonances within the rotor system. For instance, flexible disc or metallic
membrane couplings will often have a lower frequency axial resonance that may
be in the vicinity of the operating speed range. This is generally referred to as
the Natural Axial Resonant Frequency (NARF) of the coupling assembly. It is
common to have NARF values between 5,000 and 15,000 CPM for many configurations of axially compliant (soft) couplings. The computation of these frequencies becomes quite difficult due to the complex diaphragm stiffness, and the
determination of the effective mass. Hence, the simplified format represented by
equation (3-70) will probably not be adequate for calculating a coupling NARF.
Furthermore, other machine elements such as wheels or thrust collars also have
axial resonant frequencies that may have to be considered during a machinery
analysis.
Moving on to the torsional behavior, the fundamental undamped torsional
resonance of a hollow cylinder may be determined based upon equation (2-103).
The mass polar moment of inertia may be specified by equation (3-26), and the
torsional stiffness of this annulus may be determined from equation (3-65). Combining these three expressions, the following result is obtained:
G shear 4
4
- ( Do Di )
--------------------------K
32
L
tor annulus
1
1
32 G
1 G shear
F c = ------ ------------------------------- = ------ ------------------------------------------------------------- = ------ ---------------------------- ----------------------tor
2
2 J mass
32
L
2
L 4
4
annulus
- ( Do Di )
--------------------- 32 G
The general expression for the cylinder torsional natural resonant frequency may be further simplified as follows:
Fc
tor
1 G shear G
= ------ ----------------------------2
2
L
(3-69)
Interestingly enough, the torsional resonance equation (3-69) has the same
format as the axial resonance equation (3-68). The only difference is that the torsional equation uses the shear modulus Gshear, and the axial equation uses the
modulus of elasticity E. Once more the cylinder diameter is canceled out, and
equation (3-69) reveals that the torsional natural resonant frequency of the hollow cylinder is a function of the length L and the material properties. For comparative purposes, the torsional resonant frequency for the 30.0 Inch long steel
cylinders may be computed as follows:
Stiffness Influence
117
6
Fc
tor
Fc
tor
1 11.5 10 Pounds/Inch 386.1 Inches/Second

Cycles
= ------ -------------------------------------------------------------------------------------------------------------- = 664.5 ---------------2
3
Second
2
Cycles
Second
Seconds
Minute
Cycles
Minute
= 664.5 ----------------- 60 ------------------- = 39, 870 -----------------
Since the shear modulus is smaller than the elastic modulus, the torsional
resonant frequency is lower than the previously computed axial resonant frequency for the 30.0 inch long cylinder. This torsional frequency is probably outside of the normal operating speed range, and the torsional natural resonance of
the shaft by itself is not the primary area of concern. For most machinery, the
torsional resonance(s) of the system are governed by the torsional stiffness of the
couplings, and the overall mass polar moment of inertia of the entire rotor
assembly (i.e., not just the shaft). When these items are combined, the actual torsional resonance(s) may fall within the operating speed range. An example of
this type of torsional resonance situation is presented in case history 11 of the
torsional analysis of a power turbine and pump.
The third type of undamped radial or lateral resonance of a hollow cylinder
may be determined with equation (2-45). The mass may be specified by equation
(3-4), and the stiffness of this annulus may be determined from equation (3-61).
Combining these three expressions, the following result is obtained:
3 E
4
4
- ( Do Di )
---------------------3
1 K rad annulus
1 4L
F c = ------ ---------------------------- = ------ -------------------------------------------------------------rad
2
2 M annulus
L
2
2
- ( Do Di )
---------------------4
3 E
2
2
- ( Do + Di )
---------------------3
1 3E
4G
1 4L
2
2
- ----------------------- ( D o + D i )
F c = ------ -------------------------------------------------------------- = ------ ---------------------3
rad
2
2
4L
L
-
---------------------4
The expression for the cylinder radial natural resonant frequency may be
further simplified as follows:
2
1 3 E G ( Do + Di )
F c = ------ --------------------------------------------------------4
rad 2
L
(3-70)
The lateral natural resonance equation (3-70) includes the complete array
of cylinder dimensions, plus the material constants. It is clear that the lateral
resonant frequency will change in accordance with diameter changes. Obviously,
this evolves as a problem of greater mechanical complexity than either the axial
or the torsional natural resonance of the hollow cylinder. For example, assume
118
Chapter-3
that the cylinder inner diameter Di is zero, and the outer diameter is Do is
3.5682 inches (compatible with previous calculations). The resultant radial resonant frequency for the 30.0 inch long steel cylinders may be computed as follows:
6
Fc
rad
Fc
rad
1 3 30 10 Pounds/Inch 386.1 Inches/Second ( 3.5682 Inches )

= ------ -----------------------------------------------------------------------------------------------------------------------------------------------------------------4
3
2
Cycles
Second
Seconds
Minute
Cycles
Minute
= 221.1 ----------------- 60 ------------------- = 13, 270 -----------------
This frequency of 13,270 CPM is low enough to be a potential threat to the

operating speed range of the machine. In actuality, the lateral critical speed of an
entire rotor with this type of shaft geometry (i.e., 30.0 inches long, and 3.5682
inches in diameter), would be even lower due to the influence of the attached
wheels, thrust collars, and couplings. Hence, based upon this simple example, it
is clear that a machine generally has a greater propensity towards lateral critical speed problems, rather than axial or torsional resonant problems. This conclusion is strongly supported by actual industrial malfunctions where resonant
problems are usually observed in the lateral directions, and infrequently encountered in the axial or torsional directions.
From the previously discussed static deflection curves in Figs. 3-1 through
3-6, the effect of mass distribution and support location should be appreciated.
Now when the magnitude of the restraint or stiffness is included into the discussion, the dynamic response of the machine can be subjected to further evaluation. One way to envision the influence of stiffness upon vibration amplitudes is
to consider the relationship of parameters with respect to an applied force. Specifically, the applied Force is equal to the Response (vibration) times the Restraint
(stiffness). This concept is often stated in the following manner:
Force
Response = ---------------------------Restraint
(3-71)
The larger the restraint, the smaller the response for a given unit force
input. Clearly, vibratory motion can be suppressed by increasing the stiffness,
but there are penalties to be paid. For example, a rolling element bearing is
stiffer than a fluid film bearing. However, the rolling element bearing may
adversely alter the shaft mode shape. In addition, high stiffness dictates low displacement, which generally means small viscous damping. Under this case, the
rotor may not make it through the critical speed range, and the stability will certainly suffer. Therefore, the machine that performed well for many years with
sleeve bearings might self destruct with the addition of stiff ball bearings. The
same argument often applies to tilting pad bearings. Over the years, many End
Users have suffered the consequences of an inappropriate change from sleeve
bearings to a tilt pad assembly with higher stiffness, and lower damping.
Another view of the stiffness influence is presented in Fig. 3-27 illustrating
rotor mode shapes at three different critical speeds. The mode shapes on the left
describe the rotor motion with compliant (soft) bearings. For comparison, the
Stiffness Influence
119
mode shapes on the right side of Fig. 3-27 depict the anticipated mode shapes
with rigid rotor supports. This rigid support condition implies that the bearing
clamps down the shaft motion, and produces a nodal point at each bearing. It is
important to recognize that a lateral nodal point is a location of negligible motion
along the shaft axis. This condition is not totally definitive of a node, and it necessary for the shaft motion on each side of a node to be nominally 180 out of
phase. This means that the shaft on one side of the node is moving in the opposite direction of the shaft on the other side of the node. In other words, a true
zero axis crossing nodal point is defined by a rocking motion.
Soft Supports
Hard Supports
1st
Mode
2nd
Mode
Fig. 327 Effect Of Radial

Bearing Stiffness On
Rotor Mode Shapes For
The First Three Criticals
3rd
Mode
Brg.
Brg.
Brg.
Brg.
Note that the existence of a nodal point within an oil film bearing may be
indicative of a serious situation. Specifically, a nodal point is a location of minimal motion. From previous discussions, low shaft vibration amplitudes within a
bearing may manifest as low viscous damping, and this may result in low stability margins. Hence, a good rotor and bearing design may be rendered inoperative
due to the coincidence of shaft nodal points and journal bearings.
Referring back to the rotor mode shapes shown in Fig. 3-27, it is noted that
the machine describes a cylindrical mode at the first critical speed with soft supports. Relative motion at the bearings are significant, and phase measurements
would indicate similar values between bearings. With a hard support, a nodal
point occurs at each bearing, and shaft vibration measurements at the bearings
would be nil. Furthermore, one can be fooled by taking simultaneous phase measurements inboard of one bearing, and outboard of the other bearing. This data
would indicate a nominal 180 phase difference, and the diagnostician might
believe that the unit was going through a second mode, rather than a first mode
with hard supports.
The second critical describes the shaft mode shapes at the pivotal balance
120
Chapter-3
resonance. This is the critical speed during which shaft motion pivots through
the centerline of the rotor producing a zero axis crossing nodal point within the
rotor span. With a soft support stiffness, a pure conical mode is observed. It is
evident that increasing stiffness will clamp down at the bearing locations, and
cause the formation of two more nodal points. Again, the vibration measurement
planes are extremely important, and a machine could be interpreted as passing
through a first critical, when it is really transcending through a well restrained
second mode. As with any type of modal analysis, the use of transducers at additional lateral locations along the shaft may be quite informative, and in some
cases absolutely necessary.
The third critical speed with soft bearings is similar in shape to a translational first mode with hard supports. Again, the diagnostician is cautioned about
premature mode shape speculations based upon partial data. The third mode
with hard bearings exhibits a predictable shape with four nodal points. For many
rotors this third critical speed is a rotor bending mode, and shaft alterations are
usually required to appreciably change this mode shape or the associated natural frequency. Conversely, the first two modes are normally bearing dependent.
The mode shapes and natural frequencies of the first and second criticals can
often be altered by changing the stiffness characteristics of the radial journal
bearings and/or the stiffness of the bearing housings.
CRITICAL SPEED TRANSITION

The nominal 180 phase change across a shaft nodal point is generally
understandable. The concept of a shaft reversing direction (i.e., pivoting) across a
nodal point makes physical sense, and extensive technical explanations are not
required. However, the amplitude change, and the ideal 180 phase shift associated with a speed transition through a critical speed (balance resonance) do not
lend themselves to direct intuitive logic. Certainly the analytical models exhibit
these characteristics (chapters 2 and 5), and the field vibration measurements
(chapters 7, 8, 11, etc.) also display vector changes across a rotor resonance.
However, an instinctive, and universal understanding of this behavior by
mechanically inclined individuals is not a common trait.
This is not a new topic, and references date back to the post civil war era.
For instance, the 1869 technical paper by W.A. Rankin21 initially addressed this
subject. By 1882, Carl Gustaf de Laval of Sweden introduced a flexible shaft, single stage steam turbine. The German teacher and researcher August Fppl demonstrated in 1895 that a rotor could operate above a critical speed. In 1919, the
English investigator H.H. Jeffcott22 published his classic paper The Lateral
Vibration of Loaded Shafts in the Neighborhood of a Whirling Speed: The Effect
of Want of Balance. Fifteen years later, McGraw Hill Book Company began pub21 W.A. Rankin, On the Centrifugal Force of Rotating Shafts, Engineer (London): 27, (1869).
22 H.H. Jeffcott, Lateral Vibration of Loaded Shafts in the Neighborhood of a Whirling Speed
The Effect of Want of Balance, Philosophical Magazine, Vol 37, (1919), pp. 304-314.
Critical Speed Transition
121
lishing the J.P. Den Hartog23 textbook Mechanical Vibrations. In 1948, PrenticeHall published the first edition of the W.T. Thomson24 textbook also titled
Mechanical Vibrations. The academic explanations continue with many papers
and textbooks such as the 1988 Rotordynamics of Turbomachinery by John
Vance25, followed by the 1993 Turbomachinery Rotordynamics, Phenomena,
Modeling, and Analysis by Dara Childs26. Obviously, these explanations of fundamental rotor behavior have gone on for well over a century.
Interestingly enough, Jeffcott corrected the earlier work of Rankin. Some
historians believe that the Jeffcott work should be credited to de Laval or Fppl.
It appears that Den Hartog, and Thomson both used the Jeffcott model, but they
apparently neglected to reference the earlier work of Jeffcott. The books by Vance
and Childs do recognize Jeffcott, and his contributions to the field of rotordynamics. However, many good machinery engineers have expressed the opinion to the
senior author that the physical understanding of the critical speed phenomena
still seems to be clouded by the lack of a direct physical explanation of the amplitude change and the phase shift through the resonance.
Since this is such an important concept to the field of rotating machinery,
another explanation will be attempted on the following pages. This interpretation of the phenomena is based upon the Jeffcott model, and comments by Den
Hartog and Thomson. This explanation will not be as mathematically rigorous or
as extensive as the discussions by either Vance or Childs. However, the derivation presented herein is simply directed at a physical explanation of rotor behavior through a critical speed region. It is hoped that the following discussion does
make intuitive sense to the machinery diagnostician.
To begin this discussion, it is appropriate to consider the traditional diagram of a Jeffcott rotor as depicted in Fig. 3-28. The bearings in this model are
mounted at the ends of the shaft, and they are considered to be rigid, and frictionless. Initially, the damping will intentionally be set to zero, and influences
from any fluids or other sources will be ignored. The shaft is uniform, massless,
and flexible (i.e., elastic), and the shaft will have a uniform radial stiffness
defined by K.
At the midspan of the shaft, a thin, flat disk is mounted. This disk will have
a residual, or concentrated mass unbalance that is identified as M. This single
mass unbalance is located on the mass centerline identified as G (not to be confused with the acceleration of gravity, or the modulus of rigidity). As shown in
Fig. 3-28, the bearing centerline B, and the shaft and disk centerline S are identified. When the rotor is at rest, the bearing, shaft, and disk centerlines are coincident. The distance between the shaft and disk centerline S, and the mass
centerline G, will be defined as the eccentricity e. This is not the journal offset
23 J.P. Den
24 William
Hartog, Mechanical Vibrations, (New York: McGraw-Hill Book Company, 1934).

Tyrrell Thomson, Mechanical Vibrations, (Englewood Cliffs, New Jersey: PrenticeHall, Inc., 1948).
25 John M. Vance, Rotordynamics of Turbomachinery (New York: Wiley-Interscience Publication, 1988).
26 Dara Childs, Turbomachinery Rotordynamics Phenomena, Modeling, and Analysis, (New
York: Wiley-Interscience Publication, 1993).
122
Chapter-3
Unbalance
Mass (M)
Uniform, Massless Shaft,

Flexible with Stiffness (K)
G
e
B
Rigid, Frictionless
Bearings, No Damping
Fig. 328 Jeffcott Rotor Configuration
within a bearing, or an eccentric surface. It is merely the distance between the

center of the shaft and disk S, and the unbalance mass M (and associated mass
centerline G). Throughout this discussion, the geometry of the disk and the mass
remains fixed, therefore the eccentricity distance e remains constant.
At very slow speeds, unbalance forces are negligible. The shaft turns
around the bearing centerline B, and all rotating elements are concentric. This
condition is depicted in the end view of Fig. 3-29. A dial indicator or a proximity
probe mounted anywhere along the length of the shaft will indicate zero motion.
Bearing
Center
Mass
Center
B S
Fig. 329 Jeffcott Rotor At Very

Low Shaft Rotational Speed
Shaft & Disk

Center
As rotor speed increases, the straight shaft will deflect into the predictable
mode shape shown in Fig. 3-30. The translational mode shapes within this discussion exhibit zero motion at the bearings, and a maximum deflection at the
shaft midspan. The only driving force in the system is the centrifugal force due to
the unbalance mass M. The following series of diagrams are consistently annotated with the previous centerline designations (B, S, and G), and the mass
eccentricity e (offset) from the deflected shaft center. In time honored tradition,
the maximum bending deflection of the shaft (distance between B and S) is identified by r. Furthermore, the rotational speed is indicated by in all of the diagrams and equations.
By inspection of these diagrams, it is clear that the shaft and disk are rotating at the operating speed . Simultaneously, the deflected shaft is whirling in
the bearings at this speed, and it carries the disk along with it as it moves. This
motion is often referred to as synchronous whirl. The mechanism driving this
whirl is the centrifugal force generated by the eccentric mass on the disk. As
123
G
G
e
S
e
S
r
B
Centrifugal
Force
Centrifugal
Force
Mass
Center
Mass
Center
Midspan Orbit
Bearing Center
Restoring
Force
Shaft & Disk

Center
Bearing
Center
B
Restoring
Force
Midspan Orbit
Shaft & Disk

Center
Fig. 330 Jeffcott Rotor At Moderate

Speed Well Below the Balance Resonance
Fig. 331 Jeffcott Rotor Approaching The

Critical Balance Resonance Speed
rotor speed increases, the outward force increases in accordance with the normal
centrifugal force equation:
Centrifugal Force = M ( r + e )
(3-72)
In this expression, the total radius of the mass unbalance M is composed of

the shaft bending r, plus the eccentricity of the mass with respect to the shaft
centerline e. If the disk was perfectly balanced (i.e., M=0) there would be no
deflection of the shaft, and no resultant whirl. With respect to real machines,
there is always some amount of shaft bending, and some level of residual unbalance in the rotor that produces synchronous whirl. The physical influence of the
centrifugal force defined in equation (3-72) may be graphically depicted in Figs.
3-30 and 3-31. The first drawing (Fig. 3-30) describes a moderate speed below the
balance resonance (critical) speed. The condition described in Fig. 3-31 depicts a
higher speed that is approaching the critical speed. Note that the two drawings
are identical except for the increased midspan deflection r caused by the higher
operating speed, and the associated larger centrifugal force.
The end views of the midspan disk in Figs. 3-30 and 3-31 show the bearing
center B, shaft center S, and mass G. Note that the centrifugal force is in the
same line as B, S, and G. These points are collinear. That is, they lie along a common straight line. The logic behind this statement is based on a force balance.
Specifically, the centrifugal force acts outward from the bearing center B through
the mass center G. The only opposition to this radial centrifugal force is the
124
Chapter-3
restoring force of the shaft. The restoring force is simply the shaft spring constant K multiplied by the shaft deflection r as described in the next expression:
Restoring Force = K r
(3-73)
The restoring force is the elastic pull of the shaft that attempts to
straighten the shaft, and resist any deflection. It acts from the shaft center S,
back to the bearing center B. In order for the centrifugal and the restoring forces
to be in equilibrium, they must be equal in magnitude, and opposite in direction.
Hence, for the simple case, points B, G, and S must be collinear. Based upon this
logic, it is reasonable to equate the two opposing forces as follows:
Restoring Force = Centrifugal Force
(3-74)
Substituting (3-72) and (3-73) into equation (3-74) the following is obtained:
K r = M (r + e)
(3-75)
From this expression, a mechanical dilemma is immediately apparent. The

left side of the expression that represents the restoring force is linear with shaft
deflection. However, the right side of this expression that defines centrifugal
force, varies as the speed squared. Hence, at some rotational speed, the shaft displacement must become infinite to restrain the disk and attached unbalance
mass. This point of infinite displacement is aptly termed the critical speed, or the
balance resonance speed of the rotating assembly.
Another way to demonstrate this concept of infinite amplitude is based
upon an expansion, and re-configuration of (3-75) in the following manner:
2
K r = Mr +Me
2
K rMr = Me
2
r (K M ) = M e
2
2
This expression may now be solved for the bending deflection r as follows:
2
Me
e
r = ------------------------------ = -------------------------------2
2
K M
(K M)
(3-76)
From equation (2-41), it was shown that (K/M) is equal to the undamped
natural frequency c (critical speed) squared. Substituting this expression into
equation (3-76), the deflection r may now be restated as:
2
e
r = ------------------2
2
c
(3-77)
The eccentricity e between the shaft center S, and the mass centerline G
remains constant. In addition, the natural frequency c remains fixed due to an
125
invariable mass M and spring constant K. Hence, the shaft deflection r is a function of the operating speed . Furthermore, if the numerator and denominator of
(3-77) are divided by the critical speed squared c2, the following is obtained:
2
e ( c )
r = ------------------------------2
1 ( c )
(3-78)
In an effort to further simplify this expression, recall that the undamped

critical speed frequency ratio was defined in equation (2-86) as:
Critical Speed Frequency Ratio = = c
Substituting this speed ratio into (3-78) yields the next relationship:
2
e
r = ---------------2
1
(3-79)
As the shaft rotational speed increases from some slow roll condition, the
value of the critical speed ratio increases, and the mid span shaft deflection r
increases. If the eccentricity e between the mass centerline G, and the shaft centerline S is assumed to be some realistic value such as +2, Fig. 3-32 may be plotted. As expected, the midspan deflection becomes quite large as the critical speed
(=1) is approached.
Stated in another way, as the ratio of /c approaches unity, the denominator of equation (3-79) becomes smaller, and the value of r becomes increasingly
large. When the shaft speed is equal to the critical speed c, the = c/c=1,
and the denominator of equation (3-79) becomes zero. Obviously, division by zero
will result in infinity. This is consistent with the previous logic, and the definition of an undamped system.
The response plot in Fig. 3-32 was constructed between values of 0 and
0.94 to describe the amplitude characteristics below the resonance. If the plot
Shaft Center Deflection (r)
16
Fig. 332 Plot Of Jeffcott

Rotor Approaching The
Critical Balance Resonance Speed of =1
e 2
1 2
and for : e = +2
14
r=
12
10
r=
2 2
1 2
0.2
0.3
8
6
4
2
0
0.0
0.1
0.4
0.5
0.6
0.7
0.8
Frequency Ratio (=/c)
0.9
1.0
126
Chapter-3
range is extended from 0 to 3.0, the diagram shown in Fig. 3-33 evolves. Again, a
numeric value of +2 was selected for the eccentricity e, and it is noted that the
midspan shaft deflection indeed moves off towards infinity. In a real machine,
infinitely large shaft deflections are not possible. Only two possibilities for the
Jeffcott (or any other) rotor are feasible. The first option is for the midspan
amplitudes to increase to the point where the machine destroys itself (only
choice for an undamped system). The other alternative for a real machine is to
have the displacement amplitudes at the critical speed restrained by damping.
In accordance with the earlier discussion within this chapter, positive
damping is an energy dissipater. It will limit the vibration amplitude through
the resonance. For a system with low damping, such as a structural resonance,
the amplitude at the resonance will be high, and the resonance bandwidth will
be small (high Q). Conversely, for a system with large damping, such as a fluid
film bearing with viscous damping, the resonance peak will be lower, and the
bandwidth will be wider (small Q). Thus, without damping in the system, the
machine could not survive a resonance. This discussion also identifies the logic
associated with the amplitude increase at the critical speed. It is hopefully clear
from the preceding explanation, and the general equations, that the displacement at the resonance must increase. Furthermore, the amount of the vibration
increase at the critical is dependent upon the available system damping.
Unfortunately, the characteristics of the amplitude response through the
resonance are not completely defined, because one other peculiarity must be reconciled. Specifically, Fig. 3-33 reveals a midspan amplitude r that migrates off
towards positive infinity (+) as the resonance is approached. In true mathematical fashion, the amplitude above the critical returns from negative infinity (-).
That is particularly disturbing when it is also recognized that the amplitudes
above the resonance are all negative. For example, if e=2, and =2, then by equation (3-79), r is computed to be minus 2.67. Hence, the plotted curve is mathematically correct, but it does not describe a true physical situation. That is, a
negative vibration amplitude is incomprehensible.
60

Rotor Passing Through
The Critical Balance Resonance Region of =1
To Plus
e 2
1 2
and for : e = +2
r=
40
20
r=
2 2
1 2
0
-20
-40
-60
0.0
From Minus
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
127
In retrospect, equation (3-79) used for computation of this deflection r only

contains two terms. The eccentricity e, plus the frequency ratio between the running speed and the natural frequency =/c. Above the resonant frequency c
it is obvious that this speed ratio must be a positive number that is greater than
one. Hence, there is no variation of the frequency ratio that would reverse the
sign of the midspan deflection r.
The only other alternative resides with the eccentricity e. As previously
stipulated, the magnitude of e is fixed by the geometry of the disk, and the location of the mass on the disk. However, the original definition of the Jeffcott rotor
did not restrict the direction (i.e., the sign) of this eccentricity e. Hence, if the
direction of e was reversed, the plotted response curve should flip over into the
positive domain.
In fact, that is exactly what happens with a real machine, as well as the Jeffcott model. To prove this point, the response data above the resonance will be
replotted in Fig. 3-34, and e=+2 will be replaced by e=-2. In order to maintain
reasonable amplitudes, a frequency ratio range extending from 1.08 to 3.0 will be
used. It is noted that the resultant plot illustrates the proper positive deflection
r, and it also is indicative of normal behavior on the back slope of a resonance.
16

Rotor Leaving Critical Balance Resonance Speed
Region of =1 with e=-2
14
r=
e 2
1 2
and for the condition of : e = 2
12
10
r=
-2 2 2 2
2
= 2
=
1 1 1 2
1 2
6
4
2
0
1.0
At High Speeds "r" approaches "e"

1.2
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
3.0
The sign reversal of the eccentricity e is physically equivalent to a reversal

in the positions of the shaft S, and mass G centerlines. For instance, compare the
Jeffcott rotor running below the critical in Figs. 3-30 and 3-31, with the rotor
operating above the critical as shown in Figs. 3-35 and 3-36. Below the critical,
the mass M was on the outside of the rotor. Whereas, above the critical speed, the
mass M is tucked away underneath the curvature of the shaft and the light
side of the rotor is now on the outside. This reversal of heavy side out to light side
out is the mechanism behind the nominal 180 phase shift across a resonance.
Another interesting point from the response plot in Fig. 3-34 is that the
value of r steadily diminishes with an increasing frequency ratio . At high shaft
speeds, the magnitude of r approaches the magnitude of e. Physically, this characteristic is described in Fig. 3-36 of the Jeffcott rotor at a high speed condition.
128
Chapter-3
In this state (i.e., well above the critical speed), the values of e and r are essentially equal. Since e carries a negative sign (above the critical), the centerline for
the mass G is now coincident with the bearing centerline B. This direction reversal eliminates the centrifugal force, since the radius to the mass is now equal to,
or very close to zero, and the rotor has self balanced itself.
e
G
S
e
r
B
Centrifugal
Force
Bearing
Center
Mass
Center
Bearing
Center
Mass
Center
B
Midspan Orbit
Restoring
Force
Midspan Orbit
Shaft & Disk

Center
Shaft & Disk

Center
Fig. 335 Jeffcott Rotor Operating Slightly

Above The Critical Balance Resonance
Speed
Fig. 336 Jeffcott Rotor Operating In A

Self Balanced Condition Well Above The
Critical Balance Resonance Speed
This self-balancing characteristic is the mechanism that allows so many

machines to operate successfully at speeds in excess of the rotor balance resonance, or critical speed. In this high speed condition, the system is in equilibrium, and the shaft actually rotates around the mass center M, which is
equivalent to the centerline of gravity G for the rotor. For further discussions
within this chapter, and especially the balancing chapter 11, reference will be
made to rotation about the geometric axis below the critical speed, and rotation
about the mass or inertial centerline above the critical. These continuing references are based upon the fundamental behavior and understanding of the traditional Jeffcott rotor.
At the beginning of this discussion, it was stated that the damping was
intentionally set at zero. That was a convenient and necessary presumption to
maintain simplicity of the analytical model. The inclusion of damping into the
system does substantially complicate the response of the Jeffcott rotor. For
instance, consider Fig. 3-37 of this rotor with damping. Note that the centerlines
B, S, and G are no longer collinear. An angular phase change has been invoked
upon the mass centerline G. This implies that the centrifugal force, and the shaft
129
Mass
Center
S
Bearing
Center
Fig. 337 Jeffcott Rotor With Damping
Shaft & Disk

Center
spring restoring force are no longer equal and opposite. Actually, a new force
must now be included to achieve a force balance. This new force will be a damping term that may be tangential to the disk, and may vary in magnitude with the
shaft surface velocity.
The addition of this damping does not invalidate the previous discussion,
but it does complicate the scenario. For simplicity, the damping force might be
assumed to be viscous, and the resultant force equal to the tangential velocity
times the damping coefficient. This type of force could be included, and the previous analysis repeated. However, this inclusion would not necessarily improve the
overall understanding of the critical speed phenomena. Furthermore, damping
characteristics of real machines are not a simple linear function of rotor velocity,
and a proper analysis would be substantially more intricate.
In essence, it must be recognized that all rotors have some amount of damping. The oil film in the bearings, oil in the shaft seals, or the process fluid itself
may provide the damping. In all cases, the presence of damping will influence
the shaft behavior. There will be positive contributions to the addition of damping, such as lower vibration amplitudes through the critical speed region. However, damping will also open the door to a variety of mechanisms such as nonsynchronous whirl. In all cases, the real world is always more complicated than
the models that we can build to explain physical events. Fortunately, the
undamped Jeffcott model may be used to explain the fundamental characteristics associated with the balance resonance or critical speed phenomena.
130
Chapter-3
MODE SHAPE MEASUREMENT

Previous portions of this chapter have briefly covered the deflection of static
beams with variations in weight distribution, and support location. In addition,
the influence of damping, and the effect of rotor support stiffness have been
addressed. Although these are important concepts that directly influence the
behavior of rotating machinery, there is a strong argument to be made for direct
observation of the deflected rotor mode shapes.
A visual observation of the rotor mode shapes would provide substantiation
of the previously discussed theory, and allow further study into the complexities
and subtleties of rotor dynamics. Fortunately, a method exists to provide this
visualization by the application of proximity probe transducers, and electronic
display instrumentation. For example, consider Fig. 3-38 of a simple rotor that
consists of a shaft plus a center mounted disk. This is similar to the Jeffcott
rotor. For this model, assume that the rotor and disk are initially well balanced,
and that system damping is minimal. Further presume that a single concentrated mass unbalance is placed on the disk.
Proximity Probe
315
0
Fig. 338 Simple Rotor With

Concentrated Mass Unbalance
And Midspan Radial Proximity
Probe
270
R
C W ot
270
225
90
180
Mass Unbalance
The outer diameter of the disk is observed by a vertical proximity displacement probe that measures the distance between the probe tip and the disk outer
surface. In many ways, this type of vibration transducer may be considered as
the electronic equivalent of a mechanical dial indicator. Detailed discussions of
the characteristics of this type of probe are presented in chapters 6 and 7. However, for the current discussion, the reader should accept that this probe will
accurately measure the distance between the probe tip and the observed surface.
The conditioned output from this proximity probe transducer system is a
voltage sensitive signal that may be observed on an oscilloscope. The typical
scale factor for this type of transducer is 200 millivolts per Mil. Thus, a 1 Mil
(0.001 Inch) change in distance between the probe and the target surface will
produce an electronic signal equal to 200 millivolts. If the distance between the
rotor surface and the probe remains constant at all angular positions, the oscilloscope will display a straight line as the shaft is rotated. However, if the observed
surface is eccentric (for whatever reason), then the probe will display a sine wave
on the oscilloscope. The frequency of this sine wave will be equal to the speed of
Mode Shape Measurement
131
rotation, and the amplitude of the resultant sine wave will be dependent upon
the magnitude of the eccentricity.
An angular coordinate system must be established and maintained. This
angular coordinate system must allow a definitive and repeatable relationship
between the rotating system, and a stationary reference point. Normally, this is
achieved by another proximity probe that observes a notch (keyway) or projection (key) on the rotating shaft. During shaft rotation, this timing or Keyphasor transducer produces a synchronous pulse as described in later chapters of
this text. The angular location of this pulse is determined by stopping the
machine, and physically lining up the trigger point on the notch or projection
with the timing probe. When this physical alignment occurs at zero speed, it is
equivalent to the trigger point of a pulse signal during rotation.
Again, specific details of this trigger arrangement are discussed in chapters
6, 7, 8, and 11. For the purposes of this current discussion, accept the fact that a
trigger/reference arrangement does exist. Furthermore, the vibration sensing
probe is always located at zero degrees for all phase measurements. In the example rotor shown in Fig. 3-38, the vibration sensing proximity probe is shown at
the top vertical position above the disk. The rotation is specified as counterclockwise, the angular coordinate system on the disk begins at 0 at the probe, and
the angles increase in a clockwise direction. This same logic will be used
throughout this text.
Also, it must be understood that the proximity probe cannot directly identify the angular location of an effective mass unbalance. The proximity probe can
only measure distances. That is, the probe can measure the change in distance
between transducer and rotor around the entire circumference, but it cannot
directly identify the location of a mass unbalance. This effective, or equivalent, or
lumped mass unbalance location is normally identified as the Heavy Spot for the
disk or rotor system. The circumferential point that the proximity probe does
identify is the high point of the observed surface. This is the point of peak vibration that is identified by the synchronous vector phase angle. This physical location is generally referred to as the High Spot.
Consider a closer examination of the wheel on the simple rotor described in
Fig. 3-38. The diagram presented in Fig. 3-39 represents the motion of this wheel
at low rotative speeds. For purposes of discussion, assume that the rotor is operating at a speed that is well below the first critical speed (i.e., translational balance resonance) of the system. In this drawing, the center of rotation is shown to
be coincident with the geometric center. A residual or effective mass unbalance is
shown as the Heavy Spot.
If this rotor was perfectly balanced, then the Mass Center would be identical to the Geometric Center. However, if a lumped unbalance is placed on the
disk, it is clear that a shift in the Mass Center must occur. It is also logical to recognize that the new Mass Center must reside on a line between the Geometric
Center and the Heavy Spot. Furthermore, as the rotor turns, centrifugal force
will cause the disk to deflect radially in the direction of the Heavy Spot. Under a
low speed condition, the High Spot occurs at the same angular location as the
Heavy Spot. Stating it another way, the minimum distance between the disk and
132
Chapter-3
the probe occurs at the High Spot. Clearly, this is coincident with the Heavy Spot
at speeds well below the critical speed (this is exactly true only in the complete
absence of damping).
Probe
@ 0
Probe
@ 0
Phase
= 110
High
Spot
Geometric Center of
Rotation
Center
270
90
Mass
Center
Phase
Geometric
= 290
Center
Center of
Rotation
90
270
Mass
Center
Heavy
Spot
High
Spot
180
CCW Rotation
Fig. 339 Simple Rotor Operating Well

Below The Shaft Critical Speed
Heavy
Spot
180
CCW
Rotation
Fig. 340 Simple Rotor Operating Well

Above The Shaft Critical Speed
If rotation of this example system is counterclockwise, the angles would be

measured against rotation, or clockwise from the probe. Thus, the Heavy-High
Spot angular position is shown at a physical location of 110 in Fig. 3-39. If the
shaft is turning, the phase angle obtained from a Vector Filter would be 110. On
the vibration waveform, this would be equivalent to the time lag between the
peak of the vibration signal, and the trigger point of the Keyphasor pulse. If the
shaft is not rotating, and the timing notch is positioned under the Keyphasor
probe, then a counter rotation angle of 110 from the vertical vibration probe
would locate the High Spot, and the associated Heavy Spot. Balancing of this
rotor would require removing weight at 110 or adding weight at 290. This is
the logic behind the traditional proximity probe balancing rule of:
At speeds well below the critical,
remove weight at the phase angle,
or add weight at the phase angle plus 180
Next, consider what happens at rotational speeds well above the first critical. This condition is described in Fig. 3-40. In this drawing, the center of rotation is now coincident with the actual mass center of the rotor. That is, rotation
occurs around the mass center, or principal inertia axis instead of the geometric
axis. This is identical to the behavior described by the Jeffcott rotor in the self
balanced condition in Fig. 3-36.
The residual mass unbalance or Heavy Spot remains in the same angular
133
location as the slow speed case. Intuitively, this must be true. That is, the relationship between the probe and the unbalance must remain constant, or the
rotor could never be balanced. For instance, an unbalanced and loose impeller on
a shaft would have a different phase angle during every runup. This wheel could
not be balanced because the angle between the probe and the unbalance changes.
A rotor can only be balanced when the angle between the stationary probe and
the Heavy Spot on the rotor remains constant from run to run.
At high speed conditions, the new center of rotation manifests as a new
High Spot. The rotor has self-balanced itself through the critical speed region,
and the new center of rotation about the mass center produces an eccentric rotation axis. This eccentricity of rotation about the mass center results in a new
High Spot that is 180 away from the Heavy Spot. Stating it another way, the
minimum distance between the disk and the proximity probe occurs at the identified High Spot. This location is exactly opposite the Heavy Spot at rotational
speeds well above the critical (completely true in the case of no damping).
As previously mentioned, the Heavy Spot remains at 110, and the new
location of the High Spot is shown at a physical location of 290. If the shaft is
turning, the phase angle obtained from a DVF would be 290. On the vibration
waveform, this would be equivalent to the lag between the peak of the vibration
signal, and the Keyphasor pulse. If the shaft is not turning, and the timing
notch is positioned under the Keyphasor probe, then a counter rotation angle of
290 would locate the High Spot. The Heavy Spot would remain at 290 minus
180 or at 110. Balancing of this rotor would require adding weight at 290 or
removing weight at 110. This is the logic behind the proximity probe balancing
rule of:
At speeds well above the critical,
add weight at the phase angle,
or remove weight at the phase angle minus 180
Obviously the weight corrections required below and above the critical
speed would be performed in exactly the same angular locations. This is the
property that allows slow speed balancing machines to correct the mass unbalance characteristics of high speed flexible rotors operating over one or more critical speeds. However, more important to the current topic of rotor mode shapes; it
is clear that techniques exist to measure dynamic shaft motion, and to identify
shaft critical speeds. It should also be mentioned that this example was presented as a rotor with minimal damping. As previously mentioned, damping will
influence the response, and the identification of the true location of the residual
rotor unbalance, i.e., the Heavy Spot. This determination is usually made by the
computation of balance sensitivity vectors as presented in chapter 10. In a highly
damped mechanical system, the true vertical and horizontal vibration response
characteristics must be carefully examined, and the previously stated balancing
rules of thumb judicially applied.
Up to this point, the discussion has considered vibratory motion only in the
vertical plane. This is acceptable if the horizontal support characteristics are
identical to the vertical. However, in many cases there is a definable difference
134
Chapter-3
between vertical and horizontal rotor support. This difference may be attributable to variations in oil film characteristics, variations in bearing housing stiffness from vertical to horizontal, or a combination of both. In order to
accommodate this asymmetry, it is customary to install mutually perpendicular
probes to observe both the vertical and horizontal vibration response characteristics. The installation of orthogonal proximity probe transducers provides the
capability to measure total motion of the shaft within the bearing.
In addition, the measurement of shaft mode shapes should be considered as
a three-dimensional proposition. This is due to the differences between horizontal and vertical restraints, damping, and excitations along the length of the rotor.
An informative approach towards quantification of three-dimensional mode
shapes consists of measuring shaft orbits at various points along the rotor. These
orbits may be combined into an isometric view of the shaft with the respective
orbits constructed at each measurement location.
The diagram in Fig. 3-41 illustrates such a presentation. In this example, a
3/8 inch diameter shaft Rotor Kit is configured with two sleeve bearings plus a
midspan mass. Relative shaft vibration is observed with X-Y proximity probes
mounted at six different lateral positions. The shaft orbits are plotted at each
probe location, and the Keyphasor dots are connected to describe the deformed
rotor mode shape at a specific speed.
Mass Between Bearings
Bearing Journal
Shaft
Mode
Shape
Fig. 341 Mode Shape Of

Rotor Kit With Soft Sleeve
Bearings And Center Mass
Soft Sleeve Bearings

1st Critical Speed = 3,500 RPM
This drawing depicts the rotor kit operating at the vertical translational
critical speed of 3,500 RPM. These small rotor kits are quite sensitive to support
characteristics. Hence, it is normal to observe a horizontal critical speed, followed by a vertical critical with a similar mode shape. At the measured critical
speed of 3,500 RPM, the center of the shaft is considerably more deformed than
at the bearings, and a typical first translational resonance for this type of
machine is noted.
Using the same rotor kit, and changing the sleeve bearings to ball bearings
(higher stiffness), the resultant orbital patterns and associated mode shapes are
presented in Fig. 3-42. The system resonant frequency has increased to 6,300
RPM due to the higher support stiffness, and the midpoint shaft deflection has
greatly increased. It is also noted that a zero axis crossing nodal point appears at
135
both bearings. This is analogous to the hard support case presented in Fig. 3-27
where a definite phase change occurs, and the orbits are completely reversed
passing through a nodal point.
Mass Between Bearings
Bearing Journal
Shaft
Mode
Shape

Rotor Kit With Hard Ball
Bearings And Center Mass
Hard Ball Bearings

Since the shaft motion at the bearings is quite small, it must be recognized
that bearing damping is minimal with this configuration. The absence of damping increases the amplification factor through the critical, and increases the susceptibility of the machine to a variety of instability mechanisms. For stiff
bearings, it should also be noted that the unbalance energy of the rotor is closely
coupled to the supporting structure. In this configuration, casing or structural
vibration is likely to be higher than the shaft relative motion due to the direct
transmissibility of energy from the rotating shaft to the stationary casing.
On large turbines or compressors that contains very stiff bearings, the
designers often include a circumferential squeeze film damper to provide additional damping. These squeeze film dampers consist of a non-rotating, loose fitting annulus, around the bearing outer diameter. The cavity between the bearing
outer diameter and the inner diameter of the damper is filled with lube oil. The
minor motion (velocity) of the bearing housing with respect to the stationary
damper provides a means of energy dissipation through the viscous oil film.
Hence, this type of device can be used to provide more damping for a poorly
damped system.
Another popular machine configuration consists of an overhung wheel with
a short inboard bearing span. This arrangement is used by machines such as
overhung blowers, or power turbines on dual shaft gas turbines. The behavior of
this type of mechanical system is dominated by the mass and the gyroscopics of
the overhung wheel. To demonstrate the modal differences of a mass between
bearings versus an overhung mass, the rotor kit was re-configured to the
arrangement shown in Fig. 3-43. Hard ball bearings were again used, and orbital
measurements were obtained at 5 locations. Note that stiff ball bearings produce
nodal points with a 180 phase change across each node (bearing). Also observe
136
Chapter-3
the large excursions at the outboard end of the rotor, and the new critical speed
of 7,000 RPM. Clearly, the dynamic motion outboard of the bearings describes
the anticipated conical mode shape. This result is to be expected, and it is fully
consistent with the static beam diagrams presented in Figs. 3-2, 3-4, and 3-6.
Overhung Mass
Shaft
Mode
Shape
Bearing Journals

Rotor Kit With Hard Ball
Bearings And Overhung
Mass
Hard Ball Bearings

Although the orbital technique is extremely informative, it is sometimes

graphically difficult to draw a meaningful isometric diagram of the rotor system.
This is often combined with the difficulty of installing mutually perpendicular
probes at each desired location. An alternate approach consists of installing a
single proximity probe at each location. This information provides a single vector
quantity at each measurement point, which could be plotted in an isometric
manner as previously achieved with the orbits. However, this again can be difficult from a graphics standpoint.
Another technique for handling the vector data is to resolve it into Cartesian coordinates and plot the sine or cosine component at each measurement
position. This approach will provide a single quantity, either positive or negative,
as a function of position along the rotor. This technique provides acceptable mode
shapes, and the mechanics of plotting the two-dimensional data is usually quite
straightforward. It should also be mentioned that this technique, as well as the
other proximity probe measurements, must be corrected for slow roll runout at
each probe location. Failure to perform this basic runout compensation may easily destroy the validity of the dynamic shaft mode shape measurements.
137
Case History 4: Vertical Generator Mode Shape

Application of any measurement technique is always predicated upon the
availability of accessible measurement planes. Within the refining and petrochemical industries, rotating shafts are almost totally enclosed, and the installation of additional vibration transducers is difficult if not impossible. However,
within the power generation industry, the mechanical equipment is significantly
larger, and exposed shaft surfaces are often available.
This accessibility to the rotating elements allows the direct application of
the previously discussed measurement technique. For example, consider the vertical generator rotor displayed in Fig. 3-44. This generator is driven by a hydro
turbine at a normal operating speed of 277 RPM. Physically, the turbine extends
downward from the generator, and this diagram primarily depicts the three
guide bearings plus the generator shaft. The weight of the entire rotating system
approaches 100 tons, and the vertical span between upper and lower bearings is
approximately 27 feet. Although the majority of the hydro turbine rotor is under
water large portions of the generator shaft are externally visible, and all three
radial guide bearings are accessible.
1Y
90
Elevation 1
1X &
K
Upper
Guide
Bearing
CW
Rotation
Thrust
Bearing
139"
2Y
90
Generator
2X
Lower
Guide
Bearing
Elevation 2
3Y
90
30"
Elevation 3
3X
67"
4Y
90
Elevation 4
4X
92"
5Y
90
Fig. 344 Machinery And

Vibration Transducer
Arrangement For A Vertical Generator Mounted On
Top Of A Hydro Turbine
Turbine
Guide
Bearing
Elevation 5
5X
Transducers and Rotation

Are Viewed From the Top
of the Generator Looking Down
Down To Kaplan Turbine
138
Chapter-3
The machine is normally monitored with a pair of proximity probes

mounted below the lower generator guide bearing. For the purposes of this test,
relative X-Y proximity probes were temporarily installed at five different elevations as shown in the machinery arrangement diagram, Fig. 3-44. The displacement probes mounted at elevations 1, 2, 3, and 5 were attached to each
respective bearing housing with rigid brackets. The X-Y probes installed above
the coupling assembly (Elevation 4) were quite difficult to mount since the
machine had no stationary elements in the vicinity of the coupling. To address
this situation, a uni-strut frame was constructed within the confines of the turbine pit. This aluminum structure was attached to the walls, and it was reinforced with many cross braces to provide a rigid support for the vibration
transducers observing the generator shaft above the coupling assembly.
At all five elevations, the structural motion of the bearing housings and the
uni-strut frame provided a potential error source for the shaft displacement measurements. In order to quantify the magnitude, frequency content, and timing
relationships of the structural vibration, a series of ten low frequency accelerometers were installed on the machine. One accelerometer was mounted adjacent to
each radial proximity probe, and the relative shaft motion from the displacement
probe was acquired simultaneously with the casing absolute vibration sensed by
the accelerometer. In most cases, the casing acceleration signals were insignificant, and the majority of the vibration was relative motion between the shaft
and the casing. However, under some test conditions, the casing motion had to be
included, and the techniques discussed in chapter 6 were applied to determine
the overall, or absolute shaft vibration.
Under most test conditions, the largest error source was attributed to surface imperfections in the observed shaft surfaces. All of the shaft surfaces were
pitted and/or rusted, and the radial vibration signals had to be runout compensated to minimize the shaft surface imperfections. The normal operating speed of
277 RPM required that slow roll data be obtained at a much lower speed, and a
slow speed of 40 RPM was selected. Although this slow roll speed was a substantial percentage of the normal running speed, it was observed that minimal
changes occurred between 60 and 20 RPM during coastdown. Hence, 40 RPM
was selected as the nominal slow roll speed since the Digital Vector Filter (DVF)
used for this data could correctly process this low speed value.
This particular machinery train always experienced a speed increase during load rejection. Depending on the actual generator load at the time of the trip,
the rotor speed of the unit might accelerate from the normal value of 277 to as
high as 450 RPM, and then coastdown to a stop. Dynamic vibration data could be
acquired throughout the speed range, and slow roll runout vectors may be subtracted from the database to obtain runout compensated synchronous 1X shaft
displacement. For example, consider the array at runout compensated 1X vectors
at 410 RPM presented in Table 3-5.
These compensated vectors carry the units of Mils,p-p, and the phase angles
are referenced to each individual probe location. By inspection of this tabular
summary two immediate conclusions may be drawn. First, the motion at each
139
Table 35 Vertical Generator - Summary Of Initial Runout Compensated X-Y Shaft Vibration
Vectors At Maximum Load Rejection Speed Of 410 RPM
Location
Y-Probe
X-Probe
Elevation 1
10.25 Mils,p-p @ 138
7.54 Mils,p-p @ 215
Elevation 2
4.48 Mils,p-p @ 292
5.55 Mils,p-p @ 337
Elevation 3
10.81 Mils,p-p @ 275
8.37 Mils,p-p @ 12
Elevation 4
21.73 Mils,p-p @ 271
27.41 Mils,p-p @ 357
Elevation 5
10.10 Mils,p-p @ 255
11.09 Mils,p-p @ 350
elevation is forward, and essentially circular. Second, it is apparent that the bottom four elevations are generally in phase, and the top elevation is out of phase
with respect to the other four measurement planes. Based on these observations,
it is reasonable to compute an average radial motion at each elevation (basically
circular response). Also recognize that a sign change (indicative of a phase
change) must occur between the top two planes (Elevations 1 and 2).
The vector array from Table 3-5 may be manipulated in various ways to
determine an average value for the generally circular orbits. One way would be
to rotate the initial Y-Axis probe angles by 90 to be in the same angular reference position as the X-Axis probes. Performing this simple addition of 90 to each
of the Y-Axis probe angles, Table 3-6 is easily generated. The In-Phase magniTable 36 Vertical Generator - Summary Of Initial Runout Compensated X-Y Shaft Vibration
Vectors With Common Angular Reference At Load Rejection Speed Of 410 RPM
Location
Y-Probe
X-Probe
Elevation 1
10.25 Mils,p-p @ 228
7.54 Mils,p-p @ 215
Elevation 2
4.48 Mils,p-p @ 22
5.55 Mils,p-p @ 337
Elevation 3
10.81 Mils,p-p @ 5
8.37 Mils,p-p @ 12
Elevation 4
21.73 Mils,p-p @ 1
27.41 Mils,p-p @ 357
Elevation 5
10.10 Mils,p-p @ 345
11.09 Mils,p-p @ 350
tudes of each runout compensated shaft displacement vector may now be determined by multiplying each amplitude by the cosine of the associated angle (i.e.,
In-Phase=A cos ). Performing this manipulation on the vectors in Table 3-6, and
calculating a simple arithmetic average, Table 3-7 may be produced.
The average In-Phase amplitudes from Table 3-7 may now be combined into
a rotor mode shape. If these respective plus or minus magnitudes are plotted at
each elevation, and if this is performed on an overlay of the rotor drawing, the
mode shape shown in Fig. 3-45 evolves. Further embellishment to the graphics
was performed by sweeping this shape over a complete.
140
Chapter-3
Table 37 Vertical Generator - Summary Of In-Phase Components At 410 RPM

Location
Y-Probe In-Phase
X-Probe In-Phase
Average In-Phase
Elevation 1
-6.86 Mils,p-p
-6.18 Mils,p-p
-6.52 Mils,p-p
Elevation 2
+4.15 Mils,p-p
+5.11 Mils,p-p
+4.63 Mils,p-p
Elevation 3
+10.77 Mils,p-p
+8.19 Mils,p-p
+9.48 Mils,p-p
Elevation 4
+21.73 Mils,p-p
+27.37 Mils,p-p
+24.55 Mils,p-p
Elevation 5
+9.76 Mils,p-p
+10.92 Mils,p-p
+10.34 Mils,p-p
From Table 3-7 and Fig. 3-45, it is noted that the bottom four In-Phase components are positive, and the top elevation 1 carries a negative sign. This change
in sign signifies a zero axis crossing between the upper guide bearing, and the
lower guide bearing. The resultant nodal point is visible directly above the middle of the generator in Fig. 3-45. This same general mode shape was originally
evident during load rejection, and at the normal operating speed of 277 RPM.
Upper
Guide
Bearing
(-6.52 Mils,p-p)
Upper
Guide
Bearing
(-2.25 Mils,p-p)
139"
Lower
Guide
Bearing
+4.63 Mils,p-p
30"
139"
Lower
Guide
Bearing
+0.67 Mils,p-p
30"
(-0.76 Mils,p-p)
+9.48 Mils,p-p
67"
67"
+24.55 Mils,p-p
(-2.76 Mils,p-p)
92"
Turbine
Guide
Bearing
+10.34 Mils,p-p
Average In-Phase (A*Cos )
Motion At Each Elevation
Fig. 345 Measured Mode Shape For Vertical Generator At The Maximum Load
Rejection Speed Of 410 RPM
92"
Turbine
Guide
Bearing
(-1.48 Mils,p-p)
Fig. 346 Measured Mode Shape For Vertical Generator At The Normal Machine
Operating Speed Of 277 RPM
141
After some deliberation, it was finally concluded that a constant mode shape
under all operating conditions was unusual. This was combined with abnormal
temperature characteristics of the bearing pads at each guide bearing. It was
well documented that the upper generator guide bearing and the turbine guide
bearing were functioning at temperatures that were nominally 20F higher than
the lower generator guide bearing. This inconsistency appeared for many years,
and no explanation was available for this behavior.
Eventually, the machinery train was rebuilt by mechanics that were
employees of the utility company instead of contractors hired by the OEM. During reassembly of this unit, the utility company mechanics discovered that the
OEM procedure for setting the lower generator guide bearing clearances produced excessive radial clearances. This assembly procedure was modified by the
utility company mechanics, and the final results were verified with swing checks
of the vertical rotor. The ensuing test run was very successful, and bearing temperatures were now consistent at all three guide bearings. It was concluded that
the lower generator guide bearing had been essentially ineffective in restraining
the middle of the generator rotor for many years. This change in restraint at the
generator lower guide bearing was clearly reflected in the shaft mode shapes.
Specifically, the runout compensated In-Phase components at the normal operating speed of 277 RPM are summarized in Table 3-8.
Table 38 Vertical Generator - Summary Of In-Phase Components At 277 RPM
Location
Y-Probe In-Phase
X-Probe In-Phase
Average In-Phase
Elevation 1
-2.14 Mils,p-p
-2.35 Mils,p-p
-2.25 Mils,p-p
Elevation 2
+0.62 Mils,p-p
+0.72 Mils,p-p
+0.67 Mils,p-p
Elevation 3
-0.78 Mils,p-p
-0.73 Mils,p-p
-0.76 Mils,p-p
Elevation 4
-2.68 Mils,p-p
-2.85 Mils,p-p
-2.76 Mils,p-p
Elevation 5
-1.39 Mils,p-p
-1.56 Mils,p-p
-1.48 Mils,p-p
The average X and Y-Axis In-Phase components are used to plot the shaft
mode shape shown in Fig. 3-46. The influence of an active generator lower guide
bearing is quite apparent, and a new shaft nodal point was clearly introduced
below this midspan bearing. It should also be mentioned that the synchronous
1X vectors were reduced, and the vibration amplitude scaling on Fig. 3-46 at 277
RPM is five times larger than the adjacent Fig. 3-45 at 410 RPM. This change in
scaling was considered to be appropriate to allow full visibility of the modified
mode shape at 277 RPM.
It was concluded that the excessive bearing clearance associated with the
incorrect OEM assembly procedure contributed to many of the mechanical problems on this machinery train. The validity of the end user modified bearing clearance adjustment was demonstrated by a new consistency in guide bearing
temperatures, a reduction in synchronous vibration amplitudes, and logical
142
Chapter-3
changes in shaft mode shapes. In addition, the overall vibration characteristics

were less susceptible to changes in electrical load.
Although the presented hydro turbine generator mode shape data was discussed at only two speeds, the entire 1X response during startup or coastdown
should be checked. This data should be examined for the presence of rotor and
structural resonances. The number of critical speeds, combined with the general
rotor configuration, and the generic bearing type (i.e., soft versus hard) will go a
long way towards definition of the expected mode shapes. In addition, the examination of runout compensated polar plots across a machine should confirm the
inphase behavior of a translational resonance, and the out-of-phase motion of a
pivotal mode. Analysis of this type of data will be discussed in greater detail in
subsequent chapters of this text.
Overall, it is reasonable to conclude that there are techniques available for
measurement and presentation of the dynamic shaft mode shapes. Some companies apply the techniques described in this section to produce on-line threedimensional mode shapes that are continuously displayed and monitored by both
operations and maintenance personnel. This is particularly true in the hydro
power generation industry where multiple access points to the rotating shaft are
readily available. On other machines, it is impossible to mount probes all along
the length of the shaft. However, in some cases it is possible to attach proximity
probes inboard and outboard of the bearings, and outboard of the shaft seals.
Further information may be obtained from probes in the coupling area. This is
particularly true for machines with solid (inflexible) couplings. Based on this
type of additional shaft motion data, it is often possible to obtain a realistic estimate of the deformed shaft mode shape. Again variable speed data must be considered, and the diagnostician must have an awareness of the expected type of
mode shape for the machine under examination.
ANALYTICAL RESULTS
There are situations when direct mode shape measurements are not practical due to a lack of probe locations, or during the initial design stages of a
machine. In these cases, mathematical modeling of the rotor system is not only
an extremely informative tool, it may be the only available option. These calculations provide the ability to preview machine response before construction of the
machine, and it allows examination of many parameters associated with the system behavior. In addition, if problems do develop during operation, the computer
models can provide considerable insight into the anticipated operating mode
shape of the system. In addition, mode shape changes due to various excitations,
or mechanical abnormalities can be modeled and examined in the computer.
There are three fundamental types of shaft mode shape calculations. The
simplest form consists of undamped critical speed calculations, and the associated mode shapes. This type of analysis is based upon mass and stiffness properties of the system. The undamped computations yield calculated natural
frequencies, and dimensionless mode shapes. The second type of mode shape cal-
Analytical Results
143
culation evolves from a damped stability analysis of the mechanical system. The
inclusion of damping allows the computation of stability, as well as the entire
array of forward and reverse modes. This damped stability analysis also produces dimensionless shaft mode shapes.
The third major type of computed mode shape evolves from a forced
response analysis of the system. Whereas the previous two types of calculations
do not include synchronous forcing functions (e.g., unbalance), the forced
response does incorporate input forces, and it produces dimensional mode
shapes. Hence, rotor displacement amplitudes at any location and any speed
may be calculated. Chapter 5 on analytical modeling will address these three
types of mode shape calculations in greater detail. However, for the purposes of
this current chapter on rotor mode shapes, the discussion will remain at the fundamental level of undamped modes.
Case history 3 at the beginning of this chapter on the two impeller rotor
forms the basis of performing undamped critical speed calculations. In all cases,
the rotor model consists of a series of circular cylinders with particular inner and
outer diameters, length, density, and modulus of elasticity. Lumped weights and
inertias are included for elements such as impellers or couplings. The support
stiffness characteristics are also included as part of the input. This type of analytical program for undamped critical speeds performs an iterative solution for
each of the natural frequencies. It also provides a simple mode shape for each
critical speed. The specific results that can be obtained with this type of analytical computation are best illustrated with the following example of a flexible compressor rotor.
Case History 5: Eight Stage Compressor Mode Shape Change
For this case study, consider the compressor rotor described in Fig. 3-47.
This is an eight stage centrifugal compressor operating at 10,500 RPM, and consuming 7,000 horsepower. The outboard is at the thrust end of the compressor,
and a large balance piston is mounted at the outboard discharge end of the
machine. This particular machine was operated in surge for an extended period
of time. When the unit was returned to normal operation, it was discovered that
probe gap voltages had increased, and radial vibration had increased to levels
between 3.5 and 4.0 Mils,p-p. Attempts to slow the machine down were met with
even higher vibration amplitudes, and operations personnel were understandably perplexed. It is common practice within many plants to slow down a
Fig. 347 Eight Stage

Rotor Configuration
144
Chapter-3
machine when it starts to vibrate excessively. In some companies, this is a normal instruction given to all compressor and turbine operators. Hence, when this
time honored approach of slowing down the machine to get away from a point of
high vibration results in even greater vibration amplitudes, the operations personnel find that their options are severely limited.
Examination of the full speed vibration data revealed that the machine was
apparently operating in a pivotal mode, that became more severe as the speed
was decreased. This was considered to be unusual since the compressor normally
operated above the first, but below the second critical speed. In addition, the
probe gap voltages indicated a measurable babbitt loss at both journal bearings
(i.e., increased clearances). Fortunately, an undamped critical speed analysis has
been performed, and this information was available for examination in conjunction with the vibration data. Fig. 3-48 displays the calculated mode shapes at the
translational first critical with three different bearing stiffness values. Note that
at a stiffness of 200,000 Pounds/Inch (top plot), a typical cylindrical mode is predicted with a critical speed of 3,350 RPM. On the middle plot, the stiffness was
increased to 500,000 Pounds/Inch, and this change has little modal effect, but it
Fig. 348 Mode Shapes For Compressor

Rotor Translational First Critical Speed
Fig. 349 Mode Shapes For Compressor

Rotor Pivotal Second Critical Speed
Analytical Results
145
does increase the first critical speed to 4,270 RPM. A further jump in bearing
stiffness to 1,000,000 Pounds/Inch reduces motion (vibration) at the bearings,
and raises the critical speed to 4,770 RPM. Also note the similarity between
these plots and the previously discussed Fig. 3-27.
The next set of mode shape diagrams in Fig. 3-49 describes the pivotal balance resonance, or second critical speed. Again, the top plot shows the 200,000
Pounds/Inch stiffness case, and an expected conical mode at a calculated critical
speed of 7,160 RPM. The middle plot reveals the change in mode shape as the
stiffness is increased to 500,000 Pounds/Inch. At this stiffness the critical has
been increased to 10,730 RPM, and the compressor could be operating on the
front slope of this second mode. A final step of increasing stiffness to 1,000,000
Pounds/Inch has an additional clamping influence upon the mode shape, and the
critical has been raised to 13,810 RPM.
Due to the abnormal operation of this compressor in surge, it is probable
that some mechanical damage was inflicted on the bearings. It was postulated
that the potential bearing damage resulted in a reduction in bearing stiffness,
and this allowed the second critical speed to drop below operating speed. In
essence, the unit was presumed to be running above the second critical speed.
For the sake of completeness, the third critical was also examined, and the
resultant mode shapes are presented in Fig. 3-50 as bearing stiffness is
increased through the three previous steps. Note that changes in bearing stiffness has minimal effect on the deflected shaft mode shapes. As expected, the fre-
Fig. 350 Mode Shapes

For Compressor Rotor
Bending Third Critical
Speed
146
Chapter-3
quency continues to increase with a third critical speed at 14,970 RPM at

200,000 Pounds/Inch, and 22,960 RPM for the stiffest case of 1,000,000 Pounds/
Inch. It is also worthwhile to note that the third critical is a sensitive area from a
rotor fatigue standpoint. This is due to the reverse flexure acting within the
rotor. This mode is sensitive to outboard masses. For instance, the installed coupling must be extremely well balanced to allow the machine any success of survival in this mode. In actuality, most compressors of this general configuration
would be designed to avoid this mode. With respect to the current mechanical
problem on this rotor, there is no evidence to suggest any influence from this
third critical speed upon the measured high shaft vibration.
The information from a series of undamped critical speed calculations is
often summarized into a family of curves that is typically called a critical speed
map. More specifically, this type of data array is defined as an undamped critical
speed map. In this type of presentation the combined effects of bearing stiffness,
critical speed, and shaft mode are displayed on a log-log scale. The format of this
diagram facilitates the examination of a large amount of information in a reasonably concise manner.
For instance, if the calculated critical speeds from the eight stage compressor rotor are plotted against bearing stiffness, the undamped critical speed map
displayed in Fig. 3-51 may be constructed. In order to provide improved definition of the curve shapes, two additional data points at 100,000 and 2,000,000
Pounds/Inch have been added. Hence, each curve consists of five calculated critical speeds corresponding to five different stiffness values. It is significant to note
that changes in stiffness alone can determine the mode shape of the system. The
lower order modes are generally controlled by bearing stiffness, and the higher
order modes are typically governed by shaft stiffness. The controlling parameter
is normally determined by the distribution of strain energy within the
undamped critical speed calculations.
30,000
Rotor Bending - 3rd Mode
10,000
Damaged
Bearings
Good
Bearings
Pivotal
2nd Mode
Translational - 1st Mode
1,000
100,000
1,000,000
2,000,000
Stiffness (Pounds/Inch)
Fig. 351 Undamped Critical Speed Map For Eight Stage Compressor Rotor
Analytical Results
147
However, the slope of the various critical speed curves may also provide an
indication of the controlling physical parameters. For instance, if the critical
speed changes significantly with stiffness variations, then the rotor response is
probably dependent upon the bearing stiffness. Under this condition, changes in
the journal bearings would change the resultant shaft critical. However, if the
critical speed curve is fairly flat, then shaft stiffness is probably the controlling
parameter. In that case, the rotor would have to be physically modified to alter
the shaft critical speed and/or associated mode shape.
Returning to the case history under discussion, it was concluded from the
critical speed map (Fig. 3-51), that operation at the normal speed of 10,500 RPM
could result in the compressor experiencing a change in operating modes due to a
reduction in bearing or support stiffness. Hence, the machine that ran above the
first critical speed may be altered into operating above the second mode. Any
speed reduction would then raise shaft vibration amplitudes due to the rotor
entering the pivotal balance resonance. This predicted behavior was certainly
consistent with the previous experiences of the operating personnel. This conclusion is also consistent with the strong potential of bearing damage manifesting
as a reduction in support stiffness and effective bearing damping.
In the final scenario, the unit was tripped, and it did pass through both a
pivotal and a translational resonance during coastdown. Both journal bearings
were found to be damaged with expanded clearances, and the shaft seals
reflected the bearing damage. Fortunately, the rotor journals were not scored,
and all bearings and seals were replaced. The post overhaul compressor startup
was normal, and the full speed vibratory characteristics returned to previously
documented values. Hence, the damaged bearings (due to extended operation of
compressor in surge) proved to be responsible for the increased vibration amplitudes, and the change in operating shaft mode shapes.
In summary, this mode shape analysis, and the associated critical speed
map is informative during the design stages of a machine. This information is
also quite useful during troubleshooting of a field problem. In most situations,
the field problem solving exercise does not allow sufficient time to perform this
type of analysis. Hence, it is always desirable to perform this type of detailed
critical speed and mode shape analysis before the machine starts to shake. It is
also desirable to apply some of the more advanced techniques discussed in the
following chapters to examine the influence of bearing characteristics, torsional
resonances, damped analysis, plus dimensional simulation of synchronous
response characteristics.
148
Chapter-3
BIBLIOGRAPHY
2. Belding, William G. and others, ASM Handbook of Engineering Mathematics, 4th
printing, p. 319, Metals Park, Ohio: American Society for Metals, 1989.
3. Calistrat, Michael M., Flexible Couplings, their design selection and use, p. 464,
Houston: Caroline Publishing, 1994.
4. Childs, Dara, Turbomachinery Rotordynamics Phenomena, Modeling, and Analysis, New York: Wiley-Interscience Publication, 1993.
5. Den Hartog, J.P., Mechanical Vibrations, New York: McGraw-Hill Book Company,
1934.
6. Gieck, Kurt and Reiner Gieck, Engineering Formulas, 6th edition, p. M3, New York:
McGraw-Hill Inc., 1990.
7. Harris, Cyril M., Shock and Vibration Handbook, fourth edition, pp. 1-11, 1-12, 38-5,
and chap. 32, 33, 36, and 37, New York: McGraw-Hill, 1996.
8. Jeffcott, H.H., Lateral Vibration of Loaded Shafts in the Neighborhood of a Whirling Speed - The Effect of Want of Balance, Philosophical Magazine, Vol. 37 (1919),
pp. 304-314.
9. Rankin, W.A., On the Centrifugal Force of Rotating Shafts, Engineer (London): 27,
1869.
10. Shigley, Joseph E. and Charles R. Mischke, Standard Handbook of Machine Design,
pp. 9.13, and 10.6, New York: McGraw-Hill Book Company, 1986.
11. Spotts, M.F., Design of Machine Elements, 6th Edition, pp. 18, 27, and 150, Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1985.
12. Thomson, William Tyrell, Mechanical Vibrations, Englewood Cliffs, New Jersey:
Prentice Hall, Inc., 1948.
13. Vance, John M., Rotordynamics of Turbomachinery, p 61, New York: Wiley-Interscience Publication, 1988.
14. Weaver, F.L., Rotor Design and Vibration Response, Proceedings of the First Turbomachinery Symposium, Gas Turbine Laboratories, Texas A&M University, College
Station, Texas (1972), pp. 142-147.
15. Young, Warren C., Roarks Formulas for Stress & Strain, Sixth Edition, pp. 59, 101,
and 346, New York: McGraw-Hill Book Co., 1989.
Bearings and Supports
earings provide the primary interface

between the moving and the stationary parts of a machine. Although the seals
and process fluids (or magnetic fields) coexist, the bearings provide the majority
of the stiffness and damping for the moving assembly. It is understandable that
dynamic forces developed on the moving parts are transmitted to the stationary
parts across these main support bearings. The forces may be static radial loads
due to rotor weight, or they may be dynamic forces due to mechanisms such as
mass unbalance. In either case, the radial bearings must carry the applied loads,
or the machine will fail.
Many machines are also equipped with thrust bearings to restrain the axial
loads imposed by differential fluid pressures across stages of compression or
expansion. Motors, generators, some gear boxes, and some double flow compressors do not contain thrust bearings. On other trains, a single thrust bearing may
accommodate several machines connected via a series of hard couplings. In all
cases, if the thrust bearing(s) fail, the machinery train will cease to function.
It is reasonable to conclude that bearings are one of the most vulnerable
machinery elements. In many cases, bearings are the scapegoat for other malfunctions. It has been repeatedly demonstrated that bearings are often redesigned to handle higher and higher loads, when the problem actually originates
within some other part of the machine. In these situations it is desirable to
return to basic engineering principles, and examine the behavior of the entire
mechanical system, including the interaction with the main bearings. The influence of loads, physical dimensions, bearing clearances, lubricant properties, and
various geometric configurations will be reviewed in this chapter. The primary
emphasis of this discussion will be on fluid film radial bearings. However, general comments for thrust bearings and roller element bearings will be included.
As noted in the preface to this text, excessive bearing clearances represent
a major category of common machinery malfunctions. Hence, various methods
for accurate measurement of bearing clearances will be presented. In addition,
the measurement of bearing housing support characteristics will also be
addressed within this chapter.
149
150
Chapter-4
FLUID FILM RADIAL JOURNAL BEARINGS

In the development of analytical machinery models, there is always a temptation to begin the project with a detailed rotor analysis. However, the machinery diagnostician soon discovers that bearing characteristics must be defined
and included within the rotor model. In some cases, such as the optimization of a
mechanical system, many different bearing types may be considered with different rotor models. Hence, it is often an iterative procedure to determine the
proper combination of bearings and rotor configuration.
One of the easiest, and informative starting points is the determination of
the static shaft loading upon the planar area of the bearing. This calculation consists of dividing the static journal load by the plane area of the bearing as shown
in the following equation (4-1):
Journal Weight
Bearing Unit Load = BUL = ------------------------------------------------------Length Diameter
where:
Bearing Unit Load

Journal Weight
Length
Diameter
=
=
=
=
(4-1)
Static Bearing Loading (Pounds/Inch2)

Static Load on Bearing (Pounds)
Bearing Length (Inches)
Bearing Diameter (Inches)
If the weight of a rotor is evenly distributed between two bearings, then the
Journal Weight is equal to 50% of the rotor weight. The product of bearing length
times diameter yields the planar area of the bearing. That is, a top view of the
bottom half of a bearing will have a projected area that is determined by the
bearing dimensions. For example, consider a 14,400 pound rotor supported by
two bearings that are each 6.0 inches long and 8.0 inches in diameter. The static
bearing unit load may be computed from (4-1) as follows:
( 14, 400 2 ) Pounds
Journal Weight
BUL = ------------------------------------------------------- = ----------------------------------------------------
Length Diameter 6.0 Inches 8.0 Inches
7, 200 Pounds
2
BUL = ------------------------------- = 150 Pounds/Inches
2
48.0 Inches
A loading of 150 Psi is acceptable for most industrial machines. Generally,
the range of allowable bearing loads for fluid film bearings varies between 100
and 300 Psi. For lightly loaded bearings (i.e., <100 Psi), the rotor bearing system
may be susceptible to instability. In addition, the bearing size with respect to the
shaft would violate common sense. If the above example had a loading of 50 Psi,
and the rotor weight plus bearing diameter remained constant then the bearing length would have to be 18.0 inches, which is ludicrous.
Conversely, for heavily loaded bearings (i.e., >300 Psi), the bearing may fail
prematurely due to the excessive radial loads. Also, the bearing size with respect
to the shaft would not make sense. If the original example had a loading of 600
Psi, and the rotor weight plus bearing diameter remained constant then the
bearing length would have to be 1.5 inches. This is likewise unreasonable for a
Fluid Film Radial Journal Bearings
151
seven ton rotor with an 8.0 inch diameter shaft.

In addition to these static loads, forces due to unbalance, misalignment, fluids, gears, etc., must be considered in the bearing design. These cyclic forces do
not lend themselves to the simple analysis that was just performed. In order to
evaluate these parameters, plus the oil film characteristics, the bearing analysis
must be significantly expanded. Many excellent papers on dynamic bearing characteristics have been published by investigators such as Edgar Gunter1, Jim
McHugh2, Paul Allaire and Ron Flack3, and Dana Salamone4.
In most cases, determination of machine support coefficients is generally a
two step effort. The first part consists of the computation of oil film characteristics, and the second task requires the measurement of the bearing structural
support. The overall or effective rotor support is based upon a combination of
these individual, yet interrelated parameters.
Bearing oil film coefficients can be determined with two different types of
computer programs. The first type is a bearing look up program that outputs
principal stiffness Kxx, Kyy and damping Cxx, Cxx coefficients, plus the cross-coupling coefficients of Kxy, Kyx, Cxy, and Cyx. These programs also display Sommerfeld numbers with associated speeds, loads, and journal eccentricities. The
bearing parameters can be calculated for specific conditions, or over a defined
speed domain. Programs of this type run quite rapidly, and are useful for examining multiple cases prior to the final definition of parameters, and the detailed
bearing calculations.
This second type of bearing coefficient program computes the equilibrium
position, plus the stiffness and damping coefficients for a defined bearing geometry. Programs exist for cylindrical bearings, multi-lobe bearings, pressure dam
bearings, and tilting pad bearings. These programs are often sophisticated finite
element solutions that allow variable oil viscosity within the bearing, accept oil
turbulence, plus the application of vertical and horizontal external forces, and
variations in preload (where appropriate).
Typical program output data includes the principal stiffness Kxx, Kyy and
damping Cxx, Cxx coefficients, plus the cross-coupling coefficients Kxy, Kyx, Cxy,
and Cyx as required. These results are usually displayed graphically, as illustrated in Fig. 4-1. This plot describes dimensional stiffness and damping coefficients as a function of rotating speed. The presented data was calculated for a
five (5) shoe tilting pad bearing. This bearing has a length/diameter (L/D) ratio
of 0.4, a 60 arc length, load on pad (LOP), a 50% offset, and a 0.25 preload. The
1 Edgar J. Gunter, Dynamic Stability of Rotor Bearing Systems, NASA Report SP-113, 1966.
2 James D. McHugh, Principles of Turbomachinery Bearings, Proceedings of the Eighth Tur-
bomachinery Symposium, Gas Turbine Laboratories, Texas A&M University, College Station, Texas
(November 1979), pp. 135-145.
3 Paul E. Allaire, and Ronald D. Flack, Design of Journal Bearings for Rotating Machinery,
Proceedings of the Tenth Turbomachinery Symposium, Turbomachinery Laboratories, Texas A&M
University, College Station, Texas (December 1981), pp. 25-45.
4 Dana J. Salamone, Journal Bearing Design Types and Their Applications to Turbomachinery, Proceedings of the Thirteenth Turbomachinery Symposium, Turbomachinery Laboratories,
Texas A&M University, College Station, Texas (November 1984), pp 179-188.
152
Chapter-4
applied bearing static load was 1,750 pounds, and diametrical clearance was 9.0
Mils (0.009 inches). Since this is a tilting pad bearing, the cross-coupling coefficients are zero, and only the principal (xx and yy) coefficients are shown. It
should also be mentioned that due to the extended amplitude range, this type of
information is normally plotted with a log-log scale. Based on the relative magnitudes of the coefficients, the stiffness parameters appear on the top half, and the
damping curves are towards the bottom portion of the plot.
Fig. 41 Dimensional Oil

Film Bearing Stiffness And
Damping Coefficients Versus Rotor Speed
In many instances, the bearing coefficients are presented in a non-dimensional format. The customary form used for non-dimensional stiffness coefficients may be expressed by:
Cb
K NonDim = K Dim -------
W
where:
KNonDim
KDim
Cb
W
= Non-dimensional Stiffness
= Stiffness (Pounds/Inch)
= Bearing Radial Clearance (Inches)
= Static Bearing Load (Pounds)
(4-2)
153
Similarly, the non-dimensional bearing damping coefficients are calculated by:

Cb
C NonDim = C Dim -----------------
W
where:
(4-3)
CNonDim = Non-dimensional Damping

CDim = Damping (Pounds-Seconds / Inch)
= Shaft Rotative Speed (Radians / Second)
Non-dimensional coefficients are often used to define a particular bearing

design, and they permit direct comparison between bearings. Non-dimensional
coefficients also allow reasonably easy conversions between test cases for variations in bearing clearances, loads, or speed. Hence, the designer may evaluate
changes in parameters without repeating the full array of bearing calculations.
Another non-dimensional parameter that is computed for fluid film bearings is the Eccentricity ratio. In bearing terminology, eccentricity is defined as
the distance between the bearing center, and the shaft centerline position. Dividing this distance by the bearing clearance yields the dimensionless quantity of
eccentricity ratio. This physical location represents the calculated equilibrium
position of the journal. It may be presented as a vector, or as a horizontal and a
vertical eccentricity (or offset), from the bearing center.
These terms are easier to understand if an example of a fluid film bearing is
examined. For instance, consider a shaft journal rotating within a cylindrical
bearing as illustrated in Fig. 4-2. In this example, the bearing clearance circle is
Bearing Radial
Clearance
Bearing
Center
Shaft
Orbit
Fig. 42 Journal Eccentricity

Position Within A Plain Circular Fluid Film Bearing
CCW Rotation
Eccentricity
Vector
Shaft
Centerline
Position
Shaft
Centerline
Shift Vector
shown, and the bearing geometric center is identified. The shaft is defined as
rotating in a counterclockwise direction, and the shaft orbit is indicated in the
lower right quadrant of the bearing. The center of the shaft orbital motion is
commonly referred to as the shaft centerline position. The physical distance
between this shaft centerline position and the geometric center of the bearing is
defined as the shaft eccentricity. As previously noted, eccentricity may be stated
as a vector quantity, or as X-Y Cartesian coordinates.
The eccentricity ratio consists of the eccentricity magnitude divided by the
bearing clearance. Most analysts use radial bearing clearance to compute the
eccentricity ratio. As a precautionary point, eccentricity should not be confused
154
Chapter-4
with the shift in shaft centerline position from some initial rest position (e.g., at
the bottom of the bearing). Although these two vectors are directly related via
the bearing clearance, the eccentricity vector is a calculated parameter based
upon the bearing center. Whereas, the centerline position vector is measured
with shaft sensing orthogonal proximity probes, and it is generally referenced to
the bottom of the bearing (for a horizontal machine). Ideally, these two vectors
should terminate at the same point within the clearance circle.
Since the eccentricity ratio is associated with the minimum oil film thickness, it is important information for the bearing designer, as well as the machinery diagnostician. It should be recognized that each particular bearing type or
configuration displays a unique shaft centerline position of the journal within
the bearing. This running position is a function of physical parameters such as
bearing geometry, operating speed, shaft weight, and lubricant characteristics.
The actual running position may be influenced by the application of shaft preloads originating from normal sources such as gear contact forces, or abnormal
forces such as coupling misalignment.
In many machinery analysis problems, it is difficult to separate normal versus abnormal forces acting on the shaft. The dynamic motion of the shaft (vibration) is altered, and the running position of the journal within the bearing is
influenced. Hence, one must evaluate the dynamic as well as the static information. This type of evaluation is often predicated upon a comparison between normal behavior and the current motion and/or position characteristics of the shaft
position within the journal bearing. More specifically, the diagnostician must be
aware of normal shaft position characteristics in order to identify an abnormal
position. For instance, Fig. 4-3 describes the normal shaft centerline running
position for three different types of common industrial journal bearings.
The plain journal bearing shown on the left side of Fig. 4-3 is a typical bearing installed in many types of horizontal machines. On smaller machines, this
type of bearing may consist of upper and lower thin bearing liners restrained by
a heavy bearing housing. On larger machines, the babbitt bearing surface may
be integral with the bearing housing. In either case, this type of bearing generates an oil wedge in the lower right bearing quadrant (with CCW rotation). The
shaft is supported at the minimum oil film, and journal weight is supported by
the hydrodynamic forces within the bearing. In most cases, the shaft centerline
vector pivots up from 20 to 40 above the bottom horizontal plane of the bearing.
Plain Journal Bearing
Tilt Pad Bearing - Load on Pad
Shaft
Orbit
Shaft
Orbit
CCW Rotation
Shaft
Orbit
Shaft
Centerline
Position
Shaft
Centerline
Position
Shaft
Centerline
Shift Vector
Elliptical Bearing
CCW Rotation
Shaft
Centerline
Shift Vector
Shaft
Centerline
Position
CCW Rotation
Shaft
Centerline
Shift Vector
Fig. 43 Shaft Centerline Position With Three Different Types Of Fluid Film Bearings
155
The tilt pad bearing displayed in the middle of Fig. 4-3 consists of a series of
floating pads that surround the journal. A common configuration for a horizontal
machine consists of three pads in the bottom half of the bearing, and two pads
located in the top bearing half. The three bottom pads are usually configured
with one pad directly below the shaft (6 oclock position). This physical pad location is commonly referred to as a Load on Pad (LOP) arrangement. If the bearing
pads were repositioned or rotated on the shaft circumference to allow the two
bottom pads to straddle the true vertical centerline, this would be considered as
a Load Between Pad (LBP) configuration.
For a LOP arrangement, the shaft supporting oil wedge is established and
maintained on the bottom pad. Due to the location of this oil wedge, the shaft
rises essentially straight up into the normal running position. In most cases, the
shaft centerline vector pivots up from 80 to 100 above the bottom horizontal
plane of the bearing. The normal shaft running position is slightly offset from the
true vertical centerline. Usually this offset is in the direction of rotation as noted
in the center diagram of Fig. 4-3.
Shaft centerline position for a lemon bore or elliptical bearing are shown in
the diagram located on the right side of Fig. 4-3. Within this type of fixed lobe
bearing the horizontal clearances are much larger than the vertical clearances.
Typically, a ratio of 1.5:1 or 2:1 is maintained between horizontal and vertical
clearances. This physical configuration allows the rotating shaft to slide over into
lower right bearing quadrant for a counterclockwise rotating shaft (as shown), or
the lower left bearing quadrant for a clockwise shaft rotation.
Fig. 44 Non-Dimensional
Stiffness And Damping
Coefficients Versus Journal
Eccentricity Ratio
156
Chapter-4
Although there are many types and configurations of journal bearings, the
calculation of expected journal location within a bearing may be compared with
the running position as determined by DC measurements with proximity probes.
Again, it should be mentioned that the probes measure shaft position from an
initial point such as the bottom of the bearing. Thus, the vector algebra for the
probe calculation is based upon the rest point of the shaft in the bearing,
whereas the analytical calculations are referenced to the bearing center. Significant deviations between computed and measured shaft centerline positions may
be useful in the identification of a machinery problem. Conversely, if the radial
position calculations and measurements agree, the validity of the computations
are reinforced, and the diagnostician should consider looking into other aspects
of abnormal behavior of the machinery.
Refer back to the five shoe tilt pad bearing data from Fig. 4-1, and the nondimensional stiffness and damping parameters defined by equations (4-2) and (43). Clearly, the dimensional coefficients may be converted into non-dimensional
values. It is common to plot these dimensionless parameters against the nondimensional eccentricity ratio as shown Fig. 4-4. Since this is a tilt pad bearing,
the shaft will rise vertically from the bottom pad, and the angle associated with
the eccentricity vector will be in the vicinity of 90. For this common bearing, it is
noted that vertical stiffness increases as the minimum oil film decreases (i.e.,
larger eccentricity ratio). Hence, the computed results are consistent with intuitive logic and the expected behavior for this type of mechanical system.
Bearing analytical programs also compute the dimensionless Sommerfeld
number based upon the inlet viscosity, speed, length, diameter, load, and clearance. This parameter is widely used as a characteristic number for journal bearing performance. Typical values for the Sommerfeld number vary from 0.01 to
10.0. The common format for the Sommerfeld number calculation is presented in
the following expression:
2
N So
where
NSo
L
D
R
W
Cb
=
=
=
=
=
=
=
=
L D R
= ----------------------------------- -------
W
Cb
(4-4)
Sommerfeld Number (Dimensionless)

Absolute or Dynamic Oil Viscosity (Pounds-Seconds / Inch2)
Shaft Rotational Speed (Radians / Second)
Bearing Length (Inches)
Shaft Diameter (Inches)
Load on Bearing (Pounds)
Bearing Radial Clearance (Inches)
It should be mentioned that other forms of the Sommerfeld number are

used. Although the general intent of the expression remains intact, individual
designers may use different values for this dimensionless number. For instance,
the rotational speed may be stated Revolutions per Second instead of Radians
per Second. In all cases, it is mandatory to identify the parameters and engineering units when attempting to compare the Sommerfeld numbers generated by
157
Fig. 45 Non-Dimensional
Stiffness And Damping
Coefficients Versus Sommerfeld Number
two or more bearing designers.

A typical plot of Sommerfeld number versus the non-dimensional stiffness
and damping is presented In Fig. 4-5. This plot is based upon the same five shoe
tilting pad bearing depicted in Figs. 4-1 and 4-4. Generally, the bearing designer
will consider the characteristics displayed by the Sommerfeld plots, and the previously discussed eccentricity ratio plots, plus the dimensional plots of bearing
coefficients versus speed. Bearing designers also examine plots of Sommerfeld
numbers versus other non-dimensional quantities such as whirl or speed ratios.
Clearly, individual design techniques and computer programs yield a large
assortment of plot formats.
Another common calculation is the Reynolds number at the minimum oil
film. This non-dimensional number is the ratio of inertia to viscous forces, and it
may be computed by:
H R
N Re = ------------------------------------
G
where:
NRe
R
G
=
=
=
=
=
=
=
Reynolds Number (Dimensionless)

Oil Density (Pounds / Inch3)
Minimum Oil Film Height (Inches)
Shaft Rotational Speed (Radians / Second)
Acceleration of Gravity (386.1 Inches / Second2)
Absolute or Dynamic Oil Viscosity (Pounds-Seconds / Inch2)
(4-5)
158
Chapter-4
The Reynolds number allows characterization of the oil flow at the minimum oil film. This is useful in determination of the fluid flow regime. In most
instances, laminar flow through the minimum oil film is encountered. Cavitation
often occurs above the journal, but within the load carrying bottom half of the
bearing, laminar flow is the normal and desired situation. In most cases, if the
minimum oil film Reynolds number is less than 1,000, then laminar flow should
be expected. Conversely, if this value exceeds 1,000, then turbulent flow would be
a concern. The computational software should be able to handle both types of
flow regimes, and provide meaningful bearing coefficients, plus proper equilibrium position and force balance.
As noted on the Sommerfeld number, various forms of these equations are
in use, and dimensional analysis should always be performed to verify the consistency of units. In fact, some analytical computer programs do not yield true
non-dimensional values for parameters such as the Reynolds or Sommerfeld
numbers. Often a residual unit remains that alters the magnitude of the number.
Again, to avoid confusion, the machinery diagnostician should make sure that
dimensionless numbers are truly non-dimensional, or at least consistent
between comparative cases.
A non-dimensional quantity that most bearing designers agree upon is the
Preload factor for lobed or segmented bearings (e.g., tilt pad). These types of
bearings display a pad curvature that is greater than the shaft curvature. This
physical configuration forces the oil to converge close to the middle of each pad
due to the reduced clearance. In essence, preload produces or forces an oil wedge
in the bearing pad. For these types of bearings, the preload may be determined
by the ratio bearing clearance Cb, and the pad clearance Cp as follows:
C p Cb
Cb
Preload = --------------------- = 1 -------
Cp
C p
(4-6)
Either radial or diametrical clearances may be used for (4-6), but both variables must be the same. That is, Cb and Cp must be both radial, or both diametrical clearances. Another way to express the shaft preload is to convert equation
(4-6) into equivalent diameters. By substitution in the previous expression, it can
be easily shown that the Preload may also be calculated with the following:
D p Db
Preload = ----------------------
D p Ds
where:
(4-7)
Dp = Diameter of Pad Curvature (Inches)

Db = Diameter of Bearing Clearance (Inches)
Ds = Diameter of Shaft (Inches)
For a positive preload, the shaft diameter Ds is the smallest number, and
the pad diameter Dp is the largest number. If the pad and bearing clearances are
equal, then the preload is zero. The bearing is circular, with the pad and bearing
159
sharing the same center of curvature. At the other extreme, if preload is equal to
1.0, then the bearing clearance is zero, and the shaft is in direct contact with the
bearing pad. In practice, bearing preloads are typically found to be in the range
of 0.1 to 0.5. As preload increases, the bearing stiffness increases, and the damping often decreases. This relationship between the principal coefficients and the
bearing pad preload may be used to optimize the stiffness and damping characteristics of the bearing. Conversely, if the bearing pad is damaged during installation, or the babbitt is scraped by a well intentioned millwright, the preload
may be seriously altered, and the bearing characteristics totally corrupted.
The normal preload is a positive number, indicating that the bearing pad
radius of curvature is greater than the bearing radius. If this situation was
reversed, and the bearing pad displayed a radius that was smaller than the bearing, a negative preload would result. Physically, this means that the shaft is
riding on the pad edges (i.e., bearing pad is edge loaded), and premature failure
of the bearing is a certainty. On a questionable installation, bluing on the shaft
may be used to determine the actual contact area between the journal and the
bearing pads.
Additional visibility of bearing characteristics may be obtained from the
calculated pressure and temperature profiles for each design. For example, consider Fig. 4-6 of radial pressure distribution within an elliptical bearing. This
type of radial bearing is also referred to as a two lobe, or lemon bore bearing with
25 Psi oil supply pressure.
90
120
60
180
30
CCW Rotation
150
-200
0
200 400 600 800
Pressure (Psi)
210
Fig. 46 Oil Pressure Distribution Around An Elliptical Journal Bearing
330
240
300
270
The polar coordinate plot displays the circumferential pressure distribution

within the bearing at normal load and speed. Note that the maximum developed
pressure occurs at the bottom of the bearing. This point is slightly upstream of
the minimum oil film, and is consistent with expected bearing behavior and theory. Further examination of this diagram reveals another positive pressure
buildup at an angle of 25. This pressure buildup is due to the development of
another converging oil wedge between the journal and the top half of the bear-
160
Chapter-4
ing. This secondary oil wedge is due to the fixed lobe bearing geometry. In some
cases, the influence of forces originating in the top half of the bearing may be
responsible for driving an instability, or providing a positive stabilizing force
(e.g., pressure dam).
The difference between the physical location of the maximum oil film pressure, and the minimum oil film thickness points out an interesting fact in the
world of rotors and bearings. Specifically, a non-rotating shaft with a vertical
load will deflect downward in the direction of the applied load. Now assume that
the shaft is turning at a constant speed, and that an oil wedge has developed. In
this condition, the application of a downward force on the shaft will be greeted by
a vertical shift in the direction of the applied vertical load, plus a horizontal shift
of the shaft within the bearing. The horizontal shift will occur in the direction of
rotation. That is, a shaft rotating in a counterclockwise direction will move to the
right, and a shaft turning clockwise will slide to the left. This cross-coupling
mechanism across the oil film is responsible for many bearing behavioral characteristics including the self-excited instabilities discussed in chapter 9 of this text.
It should also be mentioned that most fluid film bearings are constructed
with a steel base or backing, and a babbitt coating that provides the actual bearing surface. The babbitt may be either tin or lead based, and various compositions are in common use. Since babbitt is softer than the steel journal, it is the
first sacrificial element in a bearing assembly. Ideally, during a bearing failure,
the babbitt will sustain the majority of the distress, and the steel journal will not
be damaged. Thus, the bearings may be replaced, and the rotor may be reused
without any repairs to the journals. Of course, during a major failure, the steel
journals may contact the steel backing on the bearings, and substantial damage
may be inflicted on both the journals as well as the bearings.
The babbitt thickness on journal bearings may range from micro-babbitt
thickness of 0.005 to 0.015 Inches (5 to 15 Mils) as a minimum, to 0.050 or 0.060
Inches (50 or 60 Mils) as a maximum. The thick babbitt will be a better choice for
conditions of dirty lube oil, or anticipated wear on the bearings. Unfortunately,
thick babbitt layers are susceptible to damage from impact loads, various bearing instabilities, and shaft misalignment. In many instances, a malfunction can
break off a chunk of thick babbitt, and carry it around the entire bearing, with
disastrous consequences. Babbitt is also a poor conductor of heat, and a hot bearing will generally result in a premature fatigue failure. On the other hand, the
thin micro-babbitt bearings will transmit heat to the backing material more
readily, and this type of bearing is more resistant to impact loads, and other
dynamic forces. However, the oil system must be maintained in a very clean condition. Any dirt or foreign objects in the oil may seriously damage a micro-babbitt coating. The oil film dynamics of thin versus thick babbitt bearings are
essentially the same. The size and the geometry of the bearing is often more
important than the babbitt thickness. However, the diagnostician should not
ignore this important bearing parameter.
161
Case History 6: Shaft Position In Gas Turbine Elliptical Bearings

In most cases, it is technically difficult (if not impossible) to directly check
the validity or accuracy of the computed bearing coefficients. However, each calculation must conclude with a force balance, plus a position balance of the journal within the bearing clearance. It is reasonable to believe that if the calculated
eccentricity position is correct, then the other computed parameters are also representative of the bearing characteristics. Since journal position within an oil
film bearing can be measured directly with proximity probes, it is logical to perform a check of the analytical predictions versus actual machine data.
For this case history, consider a group of four single shaft gas turbines that
operate between 5,000 and 5,350 RPM. These units are rated at 40,000 HP, and
they are used to drive high pressure centrifugal compressors through a single
helical gear box. The shaft sensing proximity probes are mounted at 45 from
the true vertical centerline as shown in Fig. 4-7. At the turbine inlet end #1 bearing, the probes are mounted above the shaft. Conversely, at the exhaust end #2
bearing, the probes are located below the shaft.
CCW
Y-Axis
X-Axis
45
Fig. 47 Angular Arrangement Of Radial Proximity

Probes On A Single Shaft
Gas Turbine
45
45
X-Axis
45
Y-Axis
CCW
Inlet Bearing
Exhaust Bearing
The eight inch diameter journals are supported in elliptical bearings. These
bearings have an average vertical diametrical clearance of 16 Mils (0.016
inches), and a normal horizontal diametrical clearance of 32 Mils (0.032 inches).
These physical dimensions are consistent with a nominal 2:1 clearance ratio previously mentioned in this chapter.
The shaft centerline position for these machine journals was determined by
measuring the proximity probe DC gap voltages at a stop condition, and at full
speed. The difference between these DC voltages is divided by the transducer
scale factor to determine the position change in the direction of each transducer.
This X-Y change in radial position may be plotted on a graph that displays the
bearing clearance, plus the calculated journal position in the X and Y directions.
Fig. 4-8 depicts the radial journal positions for the turbine inlet bearings.
Shaft centerline locations for the A unit were obtained on different dates, and at
slightly different speeds varying between 5,100 and 5,340 RPM. Three additional
machines identified as the B, C, and D units are also included in this survey.
Speeds for these last three units varied between 5,010 and 5,350 RPM. It is
noted that excellent agreement has been achieved between the calculated position at 5,340 RPM, and the six sets of field data.
The same position information for the exhaust end #2 bearing is contained
in Fig. 4-9. Notice that the scatter of data is much greater at this bearing, and
Chapter-4
Vertical Clearance (Mils)
162
T
V
W
Inlet End #1 Bearing

4
1VWT
A-5,100 RPM
A-5,340 RPM
B-5,350 RPM
C-5,010 RPM
-4
CCW Rotation
-8
-16
A-5,330 RPM
-12
-8
-4
12
16
D-5,350 RPM
Calculated
Horizontal Clearance (Mils)

Fig. 48 Shaft Centerline Position On Four Gas Turbines At Inlet End #1 Bearing
the deviations from the calculated position are substantial. Initially, it might be
concluded that the theory does not support the actual machinery behavior. However, a partial explanation for these aberrations resides within the characteristics of the proximity probe measurements. Specifically, the early vintage of both
the proximity probes, and the companion drivers are sensitive to operating temperature. The temperature limit specification for this specific probe and cable
was 350F; and the oscillator-demodulator operating limit was specified as 150F
for a standard unit, or 212F for an extended temperature range version.
As shown in Fig. 4-7, the exhaust end probes are mounted outboard of the
#2 bearing, and below the horizontal centerline. These probes are subjected to a
high temperature environment that can easily heat the transducers to temperatures in excess of 200F. The oscillator-demodulators are mounted in an explosion proof housing. Although a heat shield is installed between the turbine
exhaust and this box, the electrical components often operate at temperatures
above 130F. Thus, the exhaust end probes, cables, and drivers are all exposed to
elevated temperatures that affect the calibration curve slope.
For many years, the instrumentation vendors have recognized that operat-
T
V
W
8
Exhaust End #2 Bearing
4
0
-4
CCW Rotation
-8
-16
-12
-8
-4
T
4
A-5,330 RPM
A-5,100 RPM
A-5,340 RPM
B-5,350 RPM
C-5,010 RPM
VW
8
1
12
D-5,350 RPM
Calculated
16

Fig. 49 Shaft Centerline Position On Four Gas Turbines At Exhaust End #2 Bearing
Based On Direct Probe Gap Measurements Without Temperature Correction
163
ing temperature will influence probe calibration. For instance, Fig. 4-10 depicts
the variation in calibration curves at temperatures of 75, 200, and 350F. This
Fig. 410 Influence Of

Operating Temperature On
Proximity Probe System
Output Voltage
data was published by the manufacturer of the proximity probes installed on

these particular gas turbines. The plotted data is for a 0.300 inch diameter
probe. Larger excursions are normally exhibited by smaller diameter probe tips.
From this family of curves, it is clear that the calibration curve will bend downward as temperature increases. At 200F, the calibration curve is nominally 0.5
volts below the normal curve for gaps in the vicinity of -9.0 to -10.0 volts DC.
Hence, for a given distance between the probe and shaft, the output DC voltage
from the Proximitor is reduced by about 0.5 volts. Since the measurement system operates with a negative bias, the gap voltages are likewise negative.
The correction for this temperature behavior requires adding the incremental voltage to the Proximitor output voltage. Thus, the measured output DC
voltage should be corrected by -0.5 volts DC to yield a temperature compensated
value. Specifically, Table 4-1 summarizes the cold (at stop) gap voltages, plus the
hot (running) gap voltages for the B machine. The differential gap voltages are
merely the cold minus the hot gap voltages at the turbine exhaust bearing.
Dividing the Y-Axis (vertical) probe differential gap voltage by the normal transducer sensitivity of 0.2 Volts/Mil (200 mv/Mil) yields a displacement change of
2.15 Mils towards the probe. Similarly, the X-Axis (horizontal) transducer exhibits a -1.24 volt change, which is equivalent to a 6.20 Mil position shift away from
the probe. This is equivalent to an overall shaft vector shift of 6.56 Mils at 26
Table 41 Direct Proximity Probe Gap Voltages At Turbine Exhaust End #2 Bearing
Probe and
Angular Location
Cold Gap
Voltage
Hot Gap
Voltage
Differential
Gap Voltage
Differential
Position
Y-Axis @ 315
-9.66 volts DC
-9.23 volts DC
+0.43 volts DC
+2.15 Mils
X-Axis @ 225
-9.23 volts DC
-10.47 volts DC
-1.24 volts DC
-6.20 Mils
164
Chapter-4
from the cold to the hot position.

However, if the measured hot probe gap voltages are corrected by -0.5 volts
DC to compensate for the transducer temperature sensitivity, the results are
shown in Table 4-2. The initial cold gap voltages (zero speed) remain the same as
Table 42 Corrected Proximity Probe Gap Voltages At Turbine Exhaust End #2 Bearing
Probe and
Angular Location
Cold Gap
Voltage
Hot Gap
Voltage
Differential
Gap Voltage
Differential
Position
Y-Axis @ 315
-9.66 volts DC
-9.73 volts DC
-0.07 volts DC
-0.35 Mils
X-Axis @ 225
-9.23 volts DC
-10.97 volts DC
-1.74 volts DC
-8.70 Mils
before. The displacement shift is again determined by dividing the differential

gap voltages by 0.2 Volts/Mil to determine the distance shift. For the Y-Axis
probe, this yields a displacement change of 0.35 Mils away from the probe. The
X-Axis transducer now displays a -1.74 volt change, which is equal to an 8.70 Mil
position shift away from the transducer. The total shift of the journal centerline
position is therefore equivalent to a vector shift of 8.71 Mils at 47 (cold to hot
position). Thus, the temperature correction reveals that the shaft is really riding
higher in the bearing than the uncorrected data revealed.
Certainly the accuracy of this correction may be improved by detailed temperature sensitivity calibration of each transducer on each machinery train.
However, that type of information is often not available, or the expense of producing and maintaining this database might be cost prohibitive. Hence, the use
of a reasonable voltage correction is considered to be adequate and acceptable for
this situation.
Correcting each of the hot gap voltages from the initial shaft centerline diagram produces the journal positions presented in Fig. 4-11. Again, the exhaust
end probes are mounted on the bottom of the shaft, and the corrected DC voltages reveal a shaft rise. It is evident that agreement between the calculated and
T
V
W
8
Exhaust End #2 Bearing
with Hot Probe Gap Correction
4
0
-4
1 VW
-12
-8
-4
A-5,340 RPM
C-5,010 RPM
1
8
A-5,100 RPM
B-5,350 RPM
CCW Rotation
-8
-16
A-5,330 RPM
12
D-5,350 RPM
Calculated
16

Fig. 411 Shaft Centerline Position On Four Gas Turbines At Exhaust End #2 Bearing
Based On Temperature Corrections To The Proximity Probe Gap Voltages
Fluid Film Radial Bearing Clearance Measurements
165
measured journal position has been significantly improved by this simple probe
gap temperature correction. The remaining deviations in measured radial position between both ends of the turbine may now be attributed to the presence of
external loads, moments, or other influences acting upon the shaft.
Since the inlet end #1 bearing is adjacent to the accessory coupling, very little torque is transmitted during normal operation. Thus, the presence of external
forces, and misalignment loads are minimal at the front end bearing. As previously observed, the measured positions agree very well with the theoretical calculations that consider only the load due to the applied journal weight.
However, at the gas turbine exhaust end bearing, the full power output
from the turbine is transmitted across the load coupling. Dependent upon coupling type, alignment position and associated external forces, the actual journal
location would probably deviate from the predicted eccentricity that was computed with only the journal weight. In fact, the reverse statement might also be
appropriate. That is, since the exhaust end shaft centerline position agrees with
the computed location, the influence of external forces may be considered to be
minimal (i.e., indicative of a well-aligned Load coupling).
Overall, the eccentricity calculations at both ends of the turbine appear to
be realistic and representative of average machine behavior. This correlation
between the measured journal positions, and the computed equilibrium location
is considered to be supportive of the accuracy of the analytical fluid film bearing
calculations. Similar measurements and comparisons with calculated results
may be performed at other speeds or different oil supply conditions. In most
cases there should be a respectable correlation between the measured and the
calculated shaft centerline position. This technique may also be used as a diagnostic tool. For example, if the measured shaft operating position is substantially
different from the calculated position, the diagnostician should give strong consideration to the presence of internal or external shaft preloads.
FLUID FILM RADIAL BEARING CLEARANCE MEASUREMENTS

Assuming a proper bearing design, constant mechanical configuration, and
the availability of a suitable lubricant at the required flow, temperature, and
pressure most of the variables shown in the bearing equations migrate
towards constant values. The one parameter that generally does not remain constant is the bearing clearance. Although fluid film bearings are often touted as
lifetime bearings, the reality is that these babbitt bearings are subject to physical damage whenever the oil film collapses. This could be caused by heavy shock
loads on the bearings, loss of lubricant, or the detrimental long-term effects from
excessive unbalance or misalignment. Many other physical mechanisms will also
produce attrition in bearing babbitt thickness. In all cases, it is necessary to
monitor journal position at each radial bearing with X-Y proximity probes, and
to compare and trend this data within accurate bearing clearance diagrams.
The total diametrical clearance between the stationary bearing and the
rotating shaft may appear to be an easily measurable value. Unfortunately, it is
166
Chapter-4
often quite difficult to accurately determine the true assembly clearance of a

journal bearing. For a simple bearing configuration such as a plain circular bore,
the clearance is the difference between the shaft diameter and the inner diameter bore of the bearing. Dependent on the length of the journal, the shaft diameter is normally measured at two or more axial locations (minimum fore and aft
positions). Each axial location is typically measured at three to five different
diameters. Besides providing the necessary average shaft diameter, this data
checks for any gross diameter variations, or taper across the length of the journal. If the bearing is assembled (without the shaft), the inner diameter of the
bearing should be measured in a manner similar to the shaft. The average diametrical clearance for a plain circular bore bearing is therefore:
Cd
where:
plain
= 1, 000 ( D b D s )
(4-8)
Cd-plain = Average Diametrical Bearing Clearance (Mils)

Db = Average Bearing Inner Diameter (Inches)
Ds = Average Shaft Outer Diameter (Inches)
If the shaft is resting solidly in the bottom half of the bearing, Plastigage
may be placed on top of the shaft, and the upper bearing half installed, bolted
down, unbolted, and then removed. Comparison of the deformed width of the
Plastigage against the Width Chart supplied on each package of Plastigage will
identify the diametrical clearance. Care should be taken to insure that the correct thickness of Plastigage be used for the bearing clearance measurement. The
common colors and measurement ranges are as follows:
Green Plastigage .................1 to 3 Mils
Red Plastigage .....................2 to 6 Mils
Blue Plastigage ................... 4 to 9 Mils
If these ranges are not appropriate, or if Plastigage is not available, then
lead wire (or soft solder) may be used. For this measurement, the lead wire may
be placed on top of the shaft, the top half of the bearing installed, bolted down,
unbolted, and then removed. The thickness of the lead wire may then be measured with a 0 to 1 inch micrometer. The resultant thickness will correspond to
the diametrical bearing clearance.
For long journal bearings, a strip of Plastigage or lead wire should be placed
at either end of the bearing (i.e., fore and aft). Ideally, the clearances should be
the same at both ends of the bearing. If variations do appear, the journal should
be checked for a possible taper, and the bearing should be checked for wear or
any evidence of a conical bore. In addressing this type of incongruity, it might be
desirable to run a strip of Plastigage or lead wire axially along the top of the
shaft. This would help to identify if the dissimilar bearing clearances vary uniformly along the length of the journal, or if some type of step change in clearance
has occurred somewhere within the bearing.
For a fixed two lobe bearing such as an elliptical or lemon bore bearing, the
167
previous techniques may be used to measure the vertical bearing clearance. Horizontal clearances are somewhat more difficult to determine. One approach is to
measure across the assembled width of the bearing to determine the horizontal
bearing diameter. Since these types of bearings are often provided with axial oil
supply grooves at the splitline, it is important to measure the distance from
above one groove to below the opposite groove as shown in Fig. 4-12. Two mea-
Oil Inlet Groove
D h1
h2
Fig. 412 Horizontal Diameter Of Elliptical Bearing

With Axial Oil Grooves
surements are obtained at each end of the bearing, and they are averaged to
determine the horizontal bearing inner diameter. Subtracting the shaft diameter, and multiplying by 1,000 will yield the average horizontal clearance in Mils.
This should be compared with the vertical clearance to verify that a proper ratio
exists. Typically, a large steam turbine will have a horizontal to vertical clearance ratio of 1.5:1, and an industrial gas turbine will generally be in the vicinity
of 2:1. Another way to measure horizontal clearance on an elliptical bearing is to
use feeler gages between the shaft and the bottom half of the bearing (top half
removed). Measurements must be made on both sides of the journal, and their
sum is a good approximation of the horizontal diametrical bearing clearance.
As bearing complexity increases, the techniques used to measure bearing
clearances become more sophisticated. For stationary multi-lobe bearings,
devices such as custom taper gauges, various multipoint measurement devices,
or profile measurement machines may be used to determine bearing dimensions.
Once again, when the minimum inner bearing dimensions are determined, subtraction of the shaft diameter yields the effective diametrical clearance. In
essence, an expression similar to equation (4-8) may be used to determine the
running clearance of fixed pad bearings. For pressure dam bearings, care should
be taken to insure that the primary clearance measurement is based upon the
dam lip, and not the pressure dam depth.
A further complication is introduced when tilt pad radial bearings are considered instead of fixed geometry bearings. In these assemblies, the clearance is
influenced by pivoting of the bearing shoes. For instance, Fig. 4-13 depicts a five
shoe bearing with an internal shaft journal. This is the same type of bearing that
was used for computation of the Fig. 4-1 data. For illustration purposes, the journal in Fig. 4-13 is drastically undersized to allow an improved graphical visualization of the pad motion. If the five bearing pads are uniformly positioned
168
Chapter-4
Shaft CL
Bearing
Centerline
Shaft CL
C + Drop
b
Fig. 413 Total Vertical Clearance In A Five Pad LBP Tilting Pad Journal Bearing
around the journal (i.e., no tilt), and the journal center is coincident with the
bearing center the distance from the journal surface to the center of each pad
is equal to the radial bearing clearance Cb. If the journal is lifted vertically
upward (left side of Fig. 4-13), the shaft will stop at the center of the upper pad.
The total vertical travel from the bearing center will be equal to the radial bearing clearance Cb. If the shaft is now allowed to sink into the bottom half of the
bearing, the condition shown on the right side of Fig. 4-13 will occur. In this diagram, the shaft will sink below the physical bearing clearance circle due to the
rotation of the two bottom pads. The amount of the vertical shift will be equal to
the radial bearing clearance Cb, plus an additional Drop due to the pad pivot.
It should be noted that this Drop only occurs in the static shaft condition
depicted in Fig. 4-13. During machine operation, the diametrical bearing clearance is twice the radial bearing clearance Cb, and the static Drop does not occur.
However, if bearing clearance is to be extracted from the static journal Lift, the
pad Drop must be subtracted. Stated in another way, if a dial indicator is used to
measure the total vertical Lift of the shaft within the bearing, the indicator reading will exceed the bearing clearance. Clearly, the Lift value must be reduced by
the pad Drop in order to determine the actual diametrical bearing clearance.
Fig. 414 Shaft Drop In A

Radial Tilting Pad Bearing
Due To Pad Pivot
Drop
169
A direct approach for determining bearing clearance from shaft Lift was
presented in the 1994 papers by Nicholas5, plus Zeidan and Paquette6. These
authors used diagrams similar to Figs. 4-13 and 4-14 to explain this characteristic of radial tilting pad bearings. Specifically, within Fig. 4-14, the dotted line
describes the radial bearing clearance Cb, and is the angle between the vertical
centerline and the pad pivot point. The trigonometric relationship within this
right triangle is as follows:
Cb
cos = ----------------------------
C b + Drop
or
Cb
C b + Drop = ------------
cos
From Fig. 4-13, the total shaft Lift for this tilt pad bearing is the summation of the movement in the top half plus the bottom half of the assembly. This
may be stated and combined with the last expression as:
Lift = ( Top Half Clearance ) + ( Bottom Half Clearance )
Lift = ( C b ) + ( C b + Drop )
Cb
1
Lift = ( C b ) + ------------ = C b 1 + ------------
cos
cos
The diametrical clearance Cd may be determined by solving the last equation for the radial bearing clearance Cb, and multiplying by 2 to yield:
Cd
odd
2 Lift
= 2 C b = --------------------------1
1 + ----------
cos
(4-9)
A sub-subscript odd has been added to the diametrical clearance solution of

equation (4-9). This identifies the fact that this solution is for a tilt pad bearing
with an odd number of pads. The assembly diagram shown in Fig. 4-13 represents a load between pad (LBP) configuration. However, if this bearing was
turned upside down, a load on pad (LOP) arrangement would be depicted. Thus,
equation (4-9) is correct for either a LOP or a LBP tilt pad bearing with an odd
number of pads. For an even number of pads with LOP configuration, = 90,
5 John C. Nicholas, Tilting Pad Bearing Design, Proceedings of the Twenty-Third Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University, College Station, Texas
(September 1994), pp. 179-194.
6 Fouad Y. Zeidan, and Donald J. Paquette, Application of High Speed and High Performance
Fluid Film Bearings In Rotating Machinery, Proceedings of the Twenty-Third Turbomachinery
Symposium, Turbomachinery Laboratory, Texas A&M University, College Station, Texas (September 1994), pp. 209-233.
170
Chapter-4
and the cosine of 90 is equal to 1. Thus, equation (4-9) may be simplified as:
Cd
LOP even
2 Lift
2 Lift
2 Lift
= --------------------------- = -------------------- = -------------------- = Lift
2
1
1
1 + ------------
1 + ---
cos
1
(4-10)
The final common configuration for a radial tilt pad bearing would be a LBP
with an even number of pads. This bearing type would display excessive clearance or Drop in both the upper, and the lower halves, and the total Lift would be:
Lift = ( Top Half Clearance ) + ( Bottom Half Clearance )
Lift = ( C b + Drop ) + ( C b + Drop )
Cb
Cb
2 Cb
Lift = ------------ + ------------ = --------------- cos cos
cos
Once more, the diametrical clearance Cd may be determined by solving for
the radial bearing clearance Cb, and multiplying by 2 to produce:
Cd
LBP even
2 Lif t cos
= 2 C b = ---------------------------------------- = Lif t cos
2
(4-11)
For symmetrical tilting pad bearings with an odd number of pads, the angle
between the vertical centerline and the pad pivot point is fixed. The following
values are typically used for this bearing angular dimension:
3 Pads ..................... = 60
4 Pads ..................... = 45
5 Pads ..................... = 36
6 Pads ..................... = 30
7 Pads ..................... = 25.7
8 Pads ..................... = 22.5
For purposes of simplification and easy reference, the developed equations
(4-9) and (4-11) are combined with the standard bearing pad pivot angles, and a
diametrical clearance to lift ratio calculated for each configuration. These results
are summarized in Table 4-3, and the trivial case represented by equation (4-10)
Table 43 Tilt Pad Bearing Diametrical Clearancea Based On Shaft Lift Measurements
Load
3 Pads
4 Pads
5 Pads
6 Pads
7 Pads
8 Pads
LBP
0.667xLift
0.707xLift
0.894xLift
0.866xLift
0.948xLift
0.924xLift
LOP
0.667xLift
Lift
0.894xLift
Lift
0.948xLift
Lift
aDiametrical
Clearance = Numerical Factor x Lift
171
has also been included. If the shaft Lift in a radial tilt pad bearing is measured
with a dial indicator or a proximity probe, the measured Lift may be converted to
a diametrical clearance based upon the factors in Table 4-3 for the specific bearing configuration. For example, if a Lift of 14 Mils was measured, and the bearing was a five pad assembly, the diametrical bearing clearance is determined
from Table 4-3 as follows:
Cd
odd
= 0.894 Lift = 0.894 16 Mils = 14.3 Mils 14 Mils
Table 4-3 may also be applied to the situation where a lift check of the journal within the bearing is not physically possible. In these cases, a separate mandrel may be used to measure the allowable motion within the bearing (i.e., the
Lift). The shaft bearing journal diameter must be accurately measured as previously discussed, and a suitable mandrel cut on a precision lathe to exactly the
same dimensions. Depending on the bearing configuration, the mandrel and
assembled bearing may be mounted either vertically or horizontally. Typically,
the mandrel is fixed, and the bearing housing is physically moved back and forth
across the pads. For a fixed mandrel, a dial indicator is used to measure the overall motion of the bearing housing about the mandrel.
Conversely, if the bearing housing is mounted in some rigid fixture, the fabricated mandrel may be moved between pads, and the overall motion of the mandrel measured with the dial indicator. In either case, care must be exercised to
insure that the stationary element remains fixed, and that the mandrel and
bearing housing are collinear (i.e., the axial centerline of the mandrel is parallel
to the bearing axial centerline). In addition, an accurate dial indicator reading to
tenths of a Mil should be used for these measurements. The resultant Shift or
Lift may then be multiplied by the appropriate geometric factor from Table 4-3 to
determine the diametrical bearing clearance of the tilt pad bearing.
When checking tilt pad bearing clearances with a mandrel and an assembled bearing, it is desirable to check clearances in more than one direction. For
instance, a four pad bearing should be checked at orthogonal diameters. That is,
the updown, and the leftright pads should be measured to verify that uniform
clearances exist in both directions. For a five pad bearing, a total of three positions should be checked to insure that clearances are measured with respect to
each pad. Obviously, the same concept may be extended to bearings with a larger
number of pads.
This discussion of bearing Lift checks was predicated upon the assumption
that a vertical Lift or Shift measurement could be made directly at the bearing.
Obviously, this is an ideal condition. In many instances it is physically impossible to both mount a dial indicator next to the bearing housing, and position the
indicator on top of the shaft. A more common condition is shown in Fig. 4-15.
This diagram depicts a three stage rotor, horizontally supported between two
radial journal bearings. In the sketch at the top of Fig. 4-15, a vertical dial indicator is located close to the coupling end of the rotor. The axial distance between
the adjacent bearing and the indicator is identified as Zb-i. The span or axial distance between bearing centerlines is specified as Zb.
172
Chapter-4
Proximity
Probe
Dial
Indicator
Coupling
Proximity
Probe
Rotor in Bottom of Bearings
b-i
Rotor with Coupling End Elevated
Bearing
Bearing
Vertical Force
Lift @ Dial Indicator
Lift @ Proximity Probe
Fig. 415 Field Lift Check

Of Coupling End Bearing
On A Horizontal Rotor
Mounted Between Two
Journal Bearings
Lift @ Bearing
Assumed
Pivot
Point
Lift
Lift
Lift
b-i
In addition, a vertical proximity probe is located on the outside of each

bearing in Fig. 4-15. If an upward vertical force is applied at the coupling end of
this rotor, the shaft will move towards the top of the coupling end bearing. It is
assumed that the left end shaft remains reasonably stationary at the bottom of
the outboard end bearing. This may be verified by another dial indicator at the
outboard bearing, or the DC gap of the proximity probe at this bearing.
If no interference occurs, and if the rotor remains rigid, a straight lift
should occur between the zero motion pivot point at the outboard end of the
rotor, and the lifting point. This elevated rotor position is shown in the middle
sketch of Fig. 4-15. Additionally, the vertical change in shaft centerline along the
length of the rotor is presented at the bottom of Fig. 4-15. In this diagram, the
vertical axis is expanded for clarity. It is noted that a series of similar right triangles are present in this ideal lift diagram. By simple proportion of these right triangles, the following expression evolves:
Lif t b
Lif t i
--------------- = -------------------------Zb
Zb + Zb i
173
This equation of proportions may be solved for the bearing Liftb, as follows:
Lif t i
Z b L if t i
Lif t b = --------------------------- = ---------------------------Zb + Zb i
Z b i
1 + -----------
Z
(4-12)
Thus, the bearing shaft Liftb may be determined based upon a dial indicator Lifti obtained at a different axial location on the shaft. Next, the bearing diametrical clearance may be determined by applying the appropriate correction
factor from Table 4-3 for a tilt pad bearing, or by equating the Lift to the vertical
clearance for a fixed pad bearing. If a vertical proximity probe is mounted adjacent to the bearing (e.g., Fig. 4-15), the change in DC gap voltages may also be
used to determine the lift as shown in equation (4-13):
Mils
Lif t p = 5.0 ----------- { DC Ga p rest DC Ga p elevated } Volts
Volt
(4-13)
In many cases, the vertical shaft lift measured by a proximity probe Liftp
may be very close to the shaft shift within the bearing Liftb. This is due to the
short axial distance between the bearing and the probe location (e.g., the configuration shown in Fig. 4-15). In fact, it is highly desirable to compare the corrected
dial indicator readings from equation (4-12) with the differential probe gap readings computed with equation (4-13). This logic also applies to the opposite end of
the machine. For instance, in Fig. 4-15, the outboard bearing is the assumed
pivot point for lifting the rotor. At this location a vertical dial indicator should
show zero motion as the shaft is lifted. In many cases, this non-motion is taken
for granted, and an indicator is seldom positioned at the bearing opposite the
unit subjected to lift check. However, proximity probes are often installed, and
these probes should be monitored to verify that the shaft is not moving at the
opposite end of the rotor. In practice, the DC gaps at this opposite bearing should
not change as the shaft is raised.
On many installations, the machinery is equipped with the preferable combination of X-Y proximity probes. Often these transducers are mounted at 45
from the vertical centerline, and a true vertical proximity probe does not exist. In
this situation, the distance changes with respect to each probe should be vectorially summed to determine the overall shaft lift at that location. The specific steps
are outlined in the following case history 7.
174
Chapter-4
Case History 7: Expander Journal Bearing Clearance

A 5,000 HP hot gas expander operates at 8,016 RPM with 4.00 inch diameter journals mounted in tilting pad bearings. The bearings are four pad with a
load between pad (LBP) configuration. This machine is equipped with X-Y proximity probes adjacent to each bearing at 45 from vertical. A dial indicator was
mounted 11 inches from the bearing, and the distance between bearing centerlines was measured to be 53 inches. The shaft was lifted with a pry bar, and the
indicator showed a vertical lift of 10.5 Mils. The probe gap voltages measured
during the lift are summarized in Table 4-4:
Table 44 Summary Of Probe Gap Voltages During Lift Check
Shaft Physical Position
Y-Axis
X-Axis
45 Left of Vertical
45 Right of Vertical
At Rest - Bottom of Bearing
-10.73 Volts DC
-9.69 Volts DC
Elevated - Top of Bearing
-9.96 Volts DC
-8.51 Volts DC
Probe Location
The lift at the bearing may be calculated based upon the external lift measurement, and the axial distances between bearings and indicator position.
Using equation (4-12), it is easily determined that:
Lif t i
10.5 Mils
10.5 Mils
Lif t b = ---------------------------- = ------------------------------------ = --------------------------- = 8.69 Mils
Z
(
1
+ 0.208 )
11
Inches
b i
1 + ---------------------
1 + -----------
53 Inches
Z
b
This mechanical result should now be compared with the lift measurements
obtained with the shaft proximity probes. Applying equation (4-13) for each
transducer, the shaft shift detected by each probe may be computed in the following manner:
Lif t p
Y Axis
Mils
= 5.0 ----------- { ( 10.73 ) ( 9.69 ) } Volts = 5.20 Mils
Volt
and
Lif t p
X Axis
Mils
= 5.0 ----------- { ( 9.96 ) ( 8.51 ) } Volts = 7.25 Mils
Volt
The negative signs indicate that the shaft movement was towards the
probes. If a standard coordinate system is used, the true horizontal axis would be
at 0, and true vertical would be at 90. Within this coordinate system the X-Axis
probe would be located at 45, and the Y-Axis transducer at 135. If the measured
shifts are considered as vectors towards each probe, the overall motion may be
175
expressed as the following two vectors:

V y = A = 5.20 Mils 135
V x = B = 7.25 Mils 45
The sum of horizontal vector components are determined with (2-31):
V add
V add
horiz
horiz
= A cos + B cos
= 5.20 cos 135 + 7.25 cos 45 = 3.68 + 5.13 = 1.45 Mils
Similarly, the sum of vertical vector components are computed with (2-32):
V add
V add
vert
vert
= A sin + B sin
= 5.20 sin 135 + 7.25 sin 45 = 3.68 + 5.13 = 8.81 Mils
From these shaft position changes it is noted that the shaft did not come
straight up in the bearing. The horizontal shift of nominally 1.5 Mils indicates
that the shaft moved sideways. This is not a surprising result since the pry bar
used for the lift was not completely level, and some horizontal force was probably
applied to the rotor. For bearing clearance purposes, the vertical lift of 8.8 Mils
should be used for further calculations. However, before addressing the bearing
clearances, it is desirable to conclude the vector addition computations of the
shifts measured by the proximity probes. If equation (2-33) is used to determine
the combined magnitude shift, the following result is obtained:
V add =
( V add
horiz
) + ( V add
vert
) =
( 1.45 ) + ( 8.81 ) =
79.72 = 8.93 Mils
Note that the vector sum of 8.9 Mils is very close to the vertical shift of 8.8
Mils determined in the previous group of calculations. Finally, the angle of the
shaft lift is determined from equation (2-34) as:
V add vert
8.81
- = atan ---------- = atan ( 6.076 ) = 80.6 81
add = atan ----------------------
1.45
V add horiz
Ideally, the lift angle should be 90. Since some horizontal shift was
imposed, a slight variation in angles does occur. If the lift angle is between 75
and 105 the total vertical lift error will be less than 4%. In many cases, it is
more convenient to add the shift vectors on a handheld calculator rather than go
through the detail required in the previously outlined steps. For this application,
the diagnostician should make sure that the calculator is capable of easily performing vector addition (e.g., HP 48SX).
The vertical lift readings based upon the dial indicator should be close to
the values measured by the proximity probes (assuming that the probes are
mounted next to the bearing). If the deviation between the two values is greater
than approximately 5 to 10% then there is something wrong, and the entire
176
Chapter-4
measurement scenario should be re-examined. In this case, the calculated lift

from the X-Y probes (8.8 Mils) should be compared with the mechanical lift as
measured with the dial indicator (8.7 Mils). Since the probes are mounted outboard of the bearing, the indicated vertical lift from the probes is slightly greater
than the mechanical lift corrected to the center of the bearing. It is reasonable to
conclude that the vertical bearing lift is equal to 8.7 Mils. Based on this information, Table 4-3 or equation (4-11) may be used to determine the vertical diametrical bearing clearance as follows:
Cd
LBP even
= Lif t cos = 8.7 cos 45 = 8.7 0.707 = 6.2 Mils
The final step is to verify the general validity of this measurement. Typically, a bearing clearance ratio (BCR) is calculated as follows:
Bearing Diametrical Clearance ( Mils )
BCR = ----------------------------------------------------------------------------------------------------------Journal Diameter ( Inches )
(4-14)
Since this expander had 4.00 inch journals, the BCR is simply:
6.2 Mils
Bearing Diametrical Clearance ( Mils )
BCR = ----------------------------------------------------------------------------------------------------------- = ------------------------ = 1.6 Mils/Inch
4.00 Inches
Journal Diameter ( Inches )
A clearance to diameter ratio of 1.6 makes good sense for this bearing configuration in a horizontal machine. Table 4-5 describes the general behavior of
key parameters as the BCR is varied. Most bearing designers agree that a BCR
of 1.0 is generally on the tight side. Small bearing clearances result in high oil
film stiffness, and this is accompanied by low shaft vibration, and potentially
high bearing temperature. If the BCR is increased to 2.0, the stiffness and damping will decrease, vibration will increase, and the bearing would probably run
cooler. In addition, the machine with larger bearing clearances would be more
susceptible to a variety of instability mechanisms. In most horizontal industrial
machines, the BCR is seldom less than 1.0, and it generally does not exceed 2.0.
In specialized applications, with exotic mechanical designs and metallurgy, these
traditional limits may be extended. However, in most cases, the BCR runs
between 1.0 and 2.0.
On large vertical machines, the radial bearing loads are low, and the weight
of the rotating element is supported by a massive thrust bearing that is usually
located at the top of the machine. On these units, the radial bearing clearances
Table 45 General Trends Of Key Bearing Parameters With Variations In Bearing Clearance
Ratio (BCR) For Horizontal Machines Mounted In Fluid Film Bearing
Bearing Clearance Ratio
(BCR)
Oil Film
Stiffness
Oil Film
Damping
Shaft
Vibration
Bearing
Temperature
1.0 Mil/Inch
Increases
Increases
Decreases
Increases
1.5 Mils/Inch
Nominal
Nominal
Nominal
Nominal
2.0 Mils/Inch
Decreases
Decreases
Increases
Decreases
177
are much tighter, and Table 4-5 is not applicable. For these vertical machines,
the bearings are basically a flooded oil bath, with diametrical clearances that
generally vary between 10 and 20 Mils (0.010 and 0.020 inches). These are typically referred to as guide bearings, and their fundamental function is to keep the
shaft running in a vertical position. The clearance of these bearings are normally
obtained by physically swinging the rotor back and fourth in orthogonal directions (e.g., North-South and East-West). In this case, the upper thrust bearing
becomes the pivot point, and bearing clearance is measured with dial indicators
at each bearing. For vertical machines equipped with tilt pad bearings, the individual pads are often radially adjustable in position to provide the capability to
change the overall bearing clearance. On fixed geometry bearings, the proper
clearance has to be built into the bearing based upon actual diameter.
In any lift measurement on assembled machines, consideration must be
given to physical configurations or conditions that could cause measurement
errors. For instance, close clearance seals, or a long balance piston might restrict
the rotor lift, and appear as reduced bearing clearances. On gear boxes, if an element is partially supported by a mating gear, the lift check will be erroneous
since the starting point will not be at the bottom of the bearing. Similarly,
installed couplings, governor drive gears, and engaged turning gears will all
inhibit the shaft lift, and may be incorrectly interpreted as reduced bearing
clearances.
Conversely, excessive clearances in other machinery parts associated with
the bearings may look like large clearances. Loose hold down bolts, or housing
attachment bolts can produce inordinate shaft lift readings. On electric
machines such as motors or generators, the bearings are normally insulated with
some type of non-conducting material. This electrical insulation isolates the
rotor voltage from passing to ground through the machine bearings. These insulating blocks are usually installed with zero clearance. However, clearances can
expand with time and excessive vibration, with an overall reduction in support
stiffness. The same argument applies to the fit between the bearing assembly
and the housing. Although wide variations may be encountered for this dimension, most machines operate somewhere between an interference fit, or crush, of
1 or 2 Mils; and a clearance of 1 or 2 Mils. Clearly, excessive crush can distort the
bearing assembly resulting in premature failure, whereas excessive clearance
will reduce the support stiffness. This stiffness reduction may allow a rotor resonance that normally resides above operating speed to creep back into the operating speed domain. When this occurs, shaft vibration increases, and the
propensity towards early failure of the bearing increases.
It is generally advisable to refer to the OEM specifications for guidance in
establishing the proper clearances between the bearing assembly and the bearing cap. If this information is not available, then a zero to 1 Mil clearance should
be used as a reasonable starting point. Determination of this clearance may be
difficult due to the possibility of a zero clearance. If Plastigage or lead wire in
installed between machine parts that have essentially no clearance, the measurement media becomes smeared, and essentially useless. The solution to this
situation resides in providing an initial, or reference, clearance at the split line.
178
Chapter-4
Blue Plastigage
or Lead Wire
5 Mil
Shim
Bearing
Shaft
Fig. 416 Typical Clearance

Measurement Between
Bearing Liner And Bearing
Housing
Bearing Pedestal
For example, a 5 Mil shim has been installed at the housing split line shown in
Fig. 4-16. This shim elevates the entire upper half of the bearing cap by 5 Mils,
and allows the use of Blue Plastigage (4 to 9 Mil range) to measure the remaining clearance. If the Plastigage shows a 4 Mil clearance, then subtraction of the 5
Mil shim reveals an interference fit of 1 Mil. Conversely, if the Plastigage indicates a 6 Mil clearance, then subtraction of the 5 Mil split line shim results in a
bearing to cap clearance of 1 Mil. For clearances that exceed the measurement
range available from Plastigage, lead wire may be used. In either case, when the
measurement checks are completed, the Plastigage (or lead wire) remnants, plus
the split line shims must be removed before final assembly of the housing.
If excessive cap to bearing clearances are encountered, the best permanent
solution is to re-machine the offending stationary element(s) to restore proper
clearances. In some cases, this is not a viable option due to production or maintenance demands. In this situation, a temporary stainless steel shim may be
installed between the cap and the bearing to tighten up the assembly. If this correction technique is used, then the machine history records should clearly indicate the installation of this shim.
In all cases, the success of the lift check is highly dependent upon the
method used to mechanically lift the shaft. For light rotors, a simple pry bar is
quite adequate for this task. For heavier rotors, a screw jack, or an overhead
chain hoist might be used. On very heavy rotors, a hydraulic jack may be necessary to lift the rotor. It must be recognized that this is a potentially dangerous
practice. Rotors have been permanently bent, and bearing housings have been
cracked or broken due to the aggressive use of a hydraulic jack. This type of lift
should be performed carefully, and with full knowledge of the expected clearances. Multiple dial indicators might be installed axially on the shaft to verify
that a linear (straight line) lift is occurring. This information would help to minimize potentially bending the shaft. It might also be desirable to mount a sepa-
Bearing Supports Measurements and Calculations
179
rate dial indicator on the outboard end of the bearing housing. This vertical
indicator would be used to reveal any tendency towards a vertical lift of the housing. This information would help to minimize any damage to the bearing housing
from the vertical hydraulic jack under the shaft.
BEARING SUPPORTS MEASUREMENTS AND CALCULATIONS

The previously discussed bearing characteristics are associated with the
properties of the oil film between the rotating shaft, and stationary bearings of
different configurations. This is an acceptable description of the rotor support
system if the bearings are rigidly supported. Industrial machines with heavy
cases, and light rotating elements fall within this category. Barrel compressors
with internal bearings, rigid gear boxes, high pressure pumps, and many older
pieces of equipment operate with structural stiffness that are substantially
greater than the oil film stiffness.
However, this is not the case for many other machines that have flexible
supports and/or foundations. Units such as induced draft or forced draft fans,
steam or gas turbines, horizontally split centrifugal compressors, and pumps
with external bearings are just a few examples of machines that operate with
flexible supports. For these types of machines the remainder of the mechanical
system must be included. In a general case, the effective support stiffness for a
typical rotor on a flexible support may be defined by equation (4-15) that
describes the relationship as a group of springs in series:
1
1
1
1
1
1
------------ = ----------- + ------------- + ---------------- + ------------- + -------K eff
K oil K hsg K base K fnd etc
where:
Keff
Koil
Khsg
Kbase
Kfnd
=
=
=
=
=
(4-15)
Effective Rotor Support Stiffness (Pounds/Inch)

Oil Film (Bearing) Support Stiffness (Pounds/Inch)
Bearing Housing Support Stiffness (Pounds/Inch)
Baseplate Support Stiffness (Pounds/Inch)
Foundation Support Stiffness (Pounds/Inch)
This expression will be subjected to substantial modification if the support

structure is in a resonant condition, or if the support is highly flexible. However,
these are rare occurrences, and the above equation (4-15) is considered to be generally representative of the normal rotor support parameters.
Quantification of the structural support terms in equation (4-15) is a formidable technical feat. The calculation of these individual stiffness terms is difficult at best, and in some cases it is virtually impossible. The most reasonable
approach for determination of the support coefficients is a direct measurement of
the dynamic stiffness of the support structure. This measurement requires the
application of a defined force to the structure, and the determination of the
resultant movement. In the simplest case, a dial indicator is used to measure the
displacement in thousands of an inch, and a calibrated hydraulic jack provides
the force. Division of the applied force by the total movement yields a static stiff-
180
Chapter-4
ness in Pounds/Inch. This is a zero frequency technique that often fails to provide
the correct structural stiffness since the characteristics vary with frequency.
It is possible to measure variable frequency structural stiffness by exciting
the system with an appropriate device, and measuring the response with a vibration transducer. A frequency response function (FRF) measurement (a.k.a.,
transfer function) may be performed between the signal emitted by a force transducer, and the resultant displacement motion signal. This FRF measurement
should include the amplitude relationship between force and motion at each frequency bin, plus phase and coherence information. The force applied to the structure would typically be measured in Pounds, and the structural response would
be measured in Inches. The vibration or motion measurement could be made
with a proximity probe mounted on an isolated stand, or with a seismic transducer that is integrated to displacement. In most cases, the field motion measurements are obtained with an accelerometer, and this signal is double
integrated to obtain casing displacement. The engineering units for the frequency response function are Pounds/Inch, and this measurement is commonly
referred to as Dynamic Stiffness.
The device used to excite the structure may vary from an electromechanical
shaker to an impact hammer. The use of an electromechanical shaker provides a
highly controllable excitation source, whereas an impact hammer is easily
applied in a variety of situations. The physical installation of any shaker is often
hampered by limited access to the assembled machinery bearing housings. In
some cases, an electromechanical shaker with a stinger attached may be used to
reach specific mechanical elements. In other situations, the selection of an
impact hammer provides the necessary size and flexibility to excite a machine
bearing housing with an acceptable and definable impact force. A typical
arrangement for measuring the horizontal stiffness of a bearing housing with an
Bearing
High Sensitivity
Accelerometer
ring
Bea
Shaft
Be
ar
ing
Pe
de
sta
Fig. 417 Typical Test

Arrangement For Bearing
Housing Horizontal Impact
Test
Impact Hammer Output to DSA

Accelerometer Output to DSA
Cap
181
impact hammer is shown in Fig. 4-17.

Manual impact hammers come in various sizes for different testing applications. For instance, small hammers weigh between 1 and 2 Grams, and exhibit a
frequency limit of nominally 900,000 CPM (15 KHz). These miniature impact
hammers are used for static testing of items such as turbine blades. At the other
end of the scale a 12 pound sledge hammer, or an instrumented battering ram
may be used for low frequency tests on large structures such as foundations or
buildings. For bearing housing measurements, a typical impact hammer weighs
between 0.3 and 3 Pounds, and it is capable of producing a concentrated 5,000
pound force upon the test element. The dynamic force produced by the hammer
is generally measured with an integral piezoelectric force transducer. Frequency
response characteristics typically vary from 300 CPM (5 Hz), to a usable maximum frequency of about 60,000 CPM (1,000 Hz). Thus, the dynamic characteristics of a small to medium sized force hammer adequately cover the operating
speed range of most machines.
During structural impact tests, the casing response is usually measured
with an accelerometer attached to the bearing cap. The accelerometer should
have a frequency range that is compatible with the force transducer. In addition,
the accelerometer signal must be double integrated to convert acceleration to
casing displacement. This double integration may be performed in an external
analog device, or by application of wave form math in the DSA. In either case,
the final FRF measurement between the applied force (Pounds) and the resultant displacement response (Inches) yields an equivalent support stiffness for the
bearing housing (Pounds/Inch). In all cases, the validity of the frequency
response data is checked with the coherence function, and the relative phase
between signals should be examined.
For measurements of structural natural frequencies, the test setup is identical to the dynamic stiffness measurements, but the transducers are reversed.
In this type of test, a FRF is performed between the measured acceleration
divided by the input force. Double integration of the acceleration signal is neither required, nor desirable and the FRF output units are typically Gs/Pound.
This type of data is commonly referred to as Inertance, and it should be performed whenever a structural resonance is suspected.
Case History 8: Measured Steam Turbine Bearing Housing Stiffness
For demonstration purposes, a typical data set is presented in Fig. 4-18.
This information was obtained with a field instrumentation setup identical to
Fig. 4-17. This test was performed on the exhaust bearing housing of an 8,000
HP steam turbine that normally operates at 8,520 RPM. The data was acquired
with a 3 pound impact hammer, and a high sensitivity accelerometer mounted in
a horizontal plane. Both the impact hammer and the accelerometer were directly
connected to an HP-35665A Dynamic Signal Analyzer (DSA). The power source
within the DSA was used to drive both piezoelectric transducers, and the resultant data was stored on a floppy disk. The data was later examined on an HP35670A, and the data displayed in Fig. 4-18 committed to hard copy format with
182
Chapter-4
an HP-5L LaserJet printer. The FRF yields the dynamic stiffness plot at the bottom of Fig. 4-18. Since this data covers a wide amplitude range, a log scale was
used for the stiffness. It is noted that a reasonably flat region exists between
4,800 and 13,000 CPM. At the normal operating frequency of 8,520 RPM the
FRF reveals a dynamic stiffness value of 1,210,000 Pounds/Inch. This is judged
to be a realistic value for the heavy cast steel bearing housing. The FRF also
shows a substantial drop in stiffness at frequencies of 660 and 16,560 CPM.
If the center phase plot in Fig. 4-18 is examined, the large phase shift at
16,560 CPM might be interpreted as a structural resonance. However, when the
coherence plot at the top of Fig. 4-18 is considered, it is evident that coherence
between force and motion signals has dropped to below 0.2 at 16,560 CPM. This
indicates that the FRF data is not valid, and the significance of the change at
16,560 CPM should be removed from further consideration.
At the turbine speed of 8,520 RPM the computed coherence was 0.97. Generally, coherence values greater than 0.9 are indicative of acceptable FRF data.
Hence, the information in the vicinity of the turbine running speed is considered
Fig. 418 Frequency

Response Function (FRF)
Of Steam Turbine Bearing
Housing Horizontal
Dynamic Stiffness
183
to be excellent data. When coherence drops to levels below 0.9, the FRF data
should be cautiously applied. If coherence drops below 0.7, the FRF data should
generally be ignored.
The data array shown in Fig. 4-18 is easily acquired, and rapidly processed.
From the previous discussion it is summarized that the dynamic stiffness at turbine speed of 8,520 RPM was obtained directly from the FRF plot, and verified by
the coherence. The validity of the amplitude and phase change at 16,560 CPM
was found to be highly questionable due to the low coherence. However, the drop
in FRF amplitudes at low frequencies was not fully explained. For an improved
understanding of this behavior, and examination of the component force and displacement signals is required. This supplemental data is presented in Fig. 4-19
over the same frequency range used for the FRF data in Fig. 4-18.
Fig. 419 Force And Displacement Data Used to

Develop Steam Turbine
Frequency Response
Function (FRF) Dynamic
Stiffness Plot
The upper diagram in Fig. 4-19 is the force (in Pounds) applied across the
frequency domain of 0 to 24,000 CPM. If this same data was viewed in the time
domain, a sharp initial pulse would be observed. Within the frequency domain,
this pulse provides a reasonably uniform excitation across the selected analysis
bandwidth. Hence, it may be properly concluded that the low frequency drop off
of the FRF data is not due to any significant variations in the applied force.
However, the measured displacement presented in the bottom diagram of
Fig. 4-19 reveals a large increase in the response at 660 CPM. There might be a
tendency to consider the 660 peak as a resonance, but this conclusion is not supported by the differential phase data of Fig. 4-18. Furthermore, structural resonances have a narrow bandwidth, and the 660 CPM peak show in Fig. 4-18 does
not display this characteristic. In all probability, the 660 CPM peak is due to a
measurement anomaly. More specifically, an accelerometer was used to make the
bearing housing response measurements. The acceleration signal was double
integrated to obtain displacement. This conversion is accomplished within the
DSA by dividing the acceleration signal by frequency squared.
At the boundary condition of zero frequency, the integrated displacement
would have a value of infinity. This does not appear in the data because there is
184
Chapter-4
virtually no measurable acceleration output until the vibration transducer

becomes active around 180 CPM (3 Hz). However, it is a fundamental fact that
displacement amplitudes at low frequencies may be abnormally amplified due to
the double integration process. This is further complicated by any noise in the
acceleration signal that might also be erroneously amplified during double integration. Hence, the mechanical significance of the 660 CPM peak is eliminated.
In addition, the displacement drop at 16,560 CPM is not meaningful information
due to the previously mentioned low coherence at this frequency.
Since the dynamic stiffness FRF consists of force divided by displacement,
the increased displacement at low frequencies produces a reduction in the
dynamic stiffness. This is common behavior in all of these measurements, and
low frequency data is generally ignored. This is acceptable since the low speed
stiffness is considerably less important than the housing stiffness within the
operating speed domain.
From a measurement standpoint, it should be mentioned that the casing
displacement resulting from an impact hammer test is very small. For example,
the peak displacement at 8,520 CPM on Fig. 4-19 is only 3.36 x 10-7 Inches,o-p.
This value is equivalent to 0.000672 Mils,p-p. Using equation (2-21), this displacement converts to an accelerometer output of 0.000693 Gs,o-p at the machine
frequency. Fortunately, a high sensitivity accelerometer was used for this data,
and the scale factor of 10,000 mv/G resulted in a signal strength of only 6.93 millivolts,o-p. If the DSA is set for a full scale range of 1.0 volt,o-p, the acceleration
signal would appear at -43 dB. Although this is a small voltage, it is still within
the range of most analyzers.
As an alternate scenario, if the response accelerometer had a scale factor of
only 100 mv/G, the electrical signal would be proportionally reduced. In this situation, the analyzer would have to accommodate a low level signal of -83 dB.
Unfortunately, many instruments do not have an adequate dynamic range to
handle this variation in amplitudes. In all cases, it is recommended that a high
sensitivity accelerometer (1,000 or 10,000 mv/G) be employed for this type of
measurement. In addition, the data should be processed with a DSA that has a
sufficiently large dynamic range (e.g., HP-35670A).
If an electromechanical shaker was substituted for the impact hammer, the
applied force would be greater, and the resultant casing motion measured by the
accelerometer would also increase. Thus, the measurement problems would
diminish. However, the larger signal amplitudes must be compared with the
potential difficulty and time required to properly mount an electromechanical
shaker in the field. Regardless of the excitation source, the data processing and
examination techniques are essentially the same. Overall, the user must be fully
aware that this type of measurement is subject to a variety of errors, and all
aspects of the FRF must be validated.
In retrospect, the primary objective of this exercise is directed at a measurement of bearing housing dynamic stiffness. As discussed, the bottom plot of
Fig. 4-18 depicts the variation of this parameter with frequency. In many cases it
is desirable to develop an equation that describes this behavior. By using appro-
185
priate curve fitting software within the DSA, a suitable polynomial equation may
be defined that relates frequency to stiffness. This is an important consideration
during the accurate modeling of rotating equipment as discussed in chapter 5.
However, there are situations when this type software is not available, and
another approach must be used to develop the characteristic equation.
Case History 9: Measured Gas Turbine Bearing Housing Stiffness
The data presented in Fig. 4-20 was acquired on the inlet bearing housing
of a natural gas fired 40,000 HP gas turbine. This machine operates at 5,300
RPM, and casing stiffness information was required to enhance the accuracy of
the analytical rotor model. Data was obtained in the vertical direction (Y-Y) with
a vertical accelerometer, and a vertical impact hammer excitation. Information
was also acquired in the horizontal plane (X-X) with a horizontal accelerometer,
and a horizontal impact. For consistency with previous examples, the horizontal
FRF data is shown in Fig. 4-20. This information was derived with a three pound
Fig. 420 Horizontal Stiffness Measurement Of

Gas Turbine Inlet End #1
Bearing Housing
force hammer directly connected to an HP-3560A portable analyzer. Casing

motion was measured with a high sensitivity accelerometer that was subjected to
analog double integration prior to DSA processing. A comparison of this FRF
with the previous example reveals a somewhat jagged curve in Fig. 4-20. After
examining the various supplemental plots, it was concluded that the deviations
are attributed to the analog double integration. Fortunately, coherence was
above 0.9 at all frequencies above 1,000 CPM. Hence, the FRF data was considered to be acceptable, but a polynomial equation of this FRF was still required.
Since the HP-3560A does not have curve fitting capabilities, the FRF data was
exported to a Microsoft Excel spreadsheet, and dynamic stiffness values were
listed at 30 CPM intervals. This data was then subjected to a sixth degree polynomial curve between frequencies of 1,000 and 6,000 CPM, and the following
characteristic equation was generated:
186
Chapter-4
KS
xx
Cpm 6
Cpm 5
Cpm 4
= 4, 317 --------------- 86, 566 --------------- + 678, 621 ---------------
1, 000
1, 000
1, 000
Cpm 3
2
6
6
2.6269 10 --------------- + 5.2612 ( Cpm ) 5, 165 ( Cpm ) + 2.448 10
1, 000
Within the specified frequency range, this polynomial expression may be
used to calculate the horizontal structural stiffness of this inlet end #1 gas turbine bearing housing as a function of speed. As an example, the horizontal stiffness at the machine running speed of 5,300 RPM may be computed in the
following manner.
KS
xx
5, 300 6
5, 300 5
5, 300 4
= 4, 317 --------------- 86, 566 --------------- + 678, 621 ---------------
1, 000
1, 000
1, 000
5, 300 3
2
6
6
2.6269 10 --------------- + 5.2612 ( 5, 300 ) 5, 165 ( 5, 300 ) + 2.448 10
1, 000
KS
xx
= { 95.684 362.015 + 535.464 391.085 + 147.787 27.347 + 2.448 } 10

KS
xx
= 0.909 10 = 909, 000 Pounds/Inch
This calculated value of 909,000 Pounds/Inch is consistent with the FRF

measurement in Fig. 4-20. This type of curve fitting may also be used to determine the oil film stiffness as a function of rotating speed. The calculated stiffness
versus speed curves (e.g., Fig. 4-1) may be converted to a polynomial equation,
and the resultant expression used within the rotor response programs. In most
cases, a third or fourth degree polynomial is sufficient to describe the stiffness
curves, but the use of fifth or sixth degree equations are common.
The oil film stiffness, and various structural stiffness are combined in a
reciprocal manner as described in equation (4-15). However, the field FRF tests
essentially combine the stationary structural elements into a single housing
stiffness. Thus, it is reasonable to simplify the overall or effective rotor support
equation into the following common format:
1
1
1
------------ = ----------- + ------------K eff
K oil K hsg
(4-16)
In this expression, the oil film stiffness is identical to the previous discussion, and the housing stiffness reflects the results of the field impact or shaker
test. That is, the measured housing stiffness incorporates the flexibility of all of
the bearing housing support elements. From a practical side, it is reasonable to
assume that baseplate and foundation stiffness values are much greater than
the bearing housing stiffness. Hence, the lower bearing housing stiffness is the
dominant or controlling structural stiffness. This simplification provides the
familiar format for the effective stiffness for two springs in series. It is understandable that the effective support stiffness upon the rotor is always lower than
Bearing Housing Damping
187
either the oil film stiffness, or the housing support stiffness.

The diagnostician should always be cautious in the application and interpretation of structural dynamic stiffness measurements. The excitation source,
the elements influenced by the excitation source, the specific measurement
points, method of accelerometer double integration, plus the final signal windowing and processing within the DSA can all contribute to substantial errors. These
informative measurements should be carefully performed, and thoroughly
understood. As previously mentioned, these FRF housing measurements are
combined with the speed dependent oil film coefficients. Both of these coefficients
are used within the stability and damped critical speed analysis, plus the forced
response analysis. This combination of parameters provides a substantially
improved model of the rotor support system across the operating range.
BEARING HOUSING DAMPING

The foregoing discussion has centered on the determination of bearing
housing support stiffness. Although stiffness is a major consideration, it is not
fully definitive of the complete mechanical system. In chapter 2 of this text, a
minimum mechanical system description consisted of mass and damping in addition to the stiffness. In most cases, the mass may be considered as the weight of
the bearing housing assembly. However, the determination of structural damping is considerably less defined.
Technical papers by Gunter and Kirk7, Barrett and Nicholas8, and Nicholas, Whalen and Franklin9 generally include some type of support damping. The
most common form recommended for structural damping consists of 10% of the
critical damping for the housing or pedestal. By expanding equation (2-62) and
dividing by 10, the structural damping for the bearing housing may be estimated
with the following expression:
K hsg W hsg
C hsg 0.2 ---------------------------------G
where:
Chsg
Khsg
Whsg
G
=
=
=
=
(4-17)
Housing Damping (Pounds-Seconds / Inch)

Housing Stiffness (Pounds / Inch)
Housing Weight (Pounds)
Acceleration of Gravity (386.1 Inches / Second2)
7 Edgar J. Gunter, and R.G. Kirk, The Effect of Support Flexibility and Damping on the
Dynamic Response of a Single Mass Flexible Rotor in Elastic Bearings. Report No. ME-4040-10672U, University of Virginia, 1972.
8 L.E Barrett, and John C. Nicholas, The Effect of Bearing Support Flexibility on Critical
Speed Prediction, ASLE Transactions, 1984 Joint Lubrication Conference, San Diego, California,
February 1984 (revised June, 1984).
9 John C. Nicholas, John K. Whalen, and Sean D. Franklin, Improving Critical Speed Calculations Using Flexible Bearing Support FRF Compliance Data, Proceedings of the Fifteenth Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University, College Station, Texas
(November 1986), pp. 69-78.
188
Chapter-4
The housing stiffness in (4-17) is the measured value from the FRF measurements at a particular speed. In practice, the support stiffness polynomial
equations are modified in accordance with equation (4-17), and plots of housing
damping versus speed are generated. Alternatively, structural damping plots
may be subjected to a polynomial curve fit to develop the equations for support
damping as a function of speed. It should be noted that the housing damping values are quite small when compared to the oil film characteristics. In one case, the
structural damping was in the range of 200 to 300 Pounds-Seconds/Inch,
whereas the oil film damping coefficients were calculated to vary between 6,000
to 20,000 Pounds-Seconds/Inch. Although the support damping is a minor correction, inclusion of this parameter should improve the analytical model accuracy.
FLUID FILM THRUST BEARINGS

In addition to the radial journal bearings, most industrial machines contain
some type of thrust bearing to restrain axial motion of the shaft. As the journal
bearings support the static rotor weight plus dynamic loads the thrust bearings must contain constant thrust loads plus various axial dynamic forces. The
constant thrust loads are primarily due to differential pressure across wheels in
fluid handling machines (turbines and compressors), or axial components of gear
contact forces. Additional dynamic axial forces are caused by items such as
cocked wheels, bent rotors, or misaligned shafts. It should be noted that electric
machines such as motors or generators do not contain thrust bearings. The magnetic forces across the air gap center the rotor within the stator.
In addition to restraining axial forces, the thrust bearing must also hold the
rotor in a fixed axial position with respect to the stationary casing. This is necessary from a machine efficiency standpoint. For instance, in a steam turbine the
critical clearance between the first stage wheel and the nozzle block is maintained by the thrust bearing. Similarly, the axial position between wheels and
diaphragms in a centrifugal compressor is maintained by the thrust bearing
assembly. It should be recognized that axial clearances in most machines are relatively small, and the thrust bearing is required to maintain the axial position,
and prevent catastrophic contact between the rotating and the stationary parts.
In all cases, the thrust bearing assembly must fulfill this requirement for
maintaining axial position, plus axial float within the thrust bearing. These two
basic requirements demand two adjustments to the thrust bearing assembly. In
many machines, solid shims are installed behind the thrust shoes to establish
these critical axial dimensions. These shims are precision ground to tight tolerances, and they are normally segmented into two 180 sections. Typically, the
axial position shim located behind the active thrust shoes is ground to establish
the axial location of the rotor. Then the shims behind the inactive shoes are
ground to establish the total thrust float within the bearing. Typical installations
of thrust shims are shown in Figs. 4-21 and 4-22. It should also be mentioned
that other mechanical schemes are used on process machines. However, in virtually all applications, one adjustment must be available for the rotor position, and
Fluid Film Thrust Bearings
189
a second adjustment must be available to set the thrust float.

On journal bearings, maximum shaft surface velocities are often limited to
200 or 250 feet per second. By comparison, the larger diameter thrust collars
may exhibit peripheral speeds in the vicinity of 450 to 500 feet per second. These
higher speeds demand proper lubrication to cool the higher temperatures associated with the increased surface velocities. In most machines, thrust bearings
consist of segmented floating pads that are often used in conjunction with
directed lubrication between the pads (e.g. Kingsbury type bearing). The stationary thrust bearing pads are assembled in a circular pattern, and they operate
with an oil film between the stationary pads and a rotating thrust collar. As
shown in Fig. 4-21, a machine normally contains a set of active thrust shoes, plus
a set of inactive shoes. Under normal operating conditions, the machine runs on
the active shoes. During process or mechanical upsets, the rotor may slam up
against the inactive thrust shoes. To protect the machine during this type of
transition, the inactive thrust bearing is often identical to the active thrust bearing in terms of mechanical construction, and load carrying capability.
Journal
Bearing
Thrust Float
Shaft
Thrust Probe
Fig. 421 Typical Single

Thrust Collar Assembly
On Horizontal Machinery
Shims for Axial Position

Active Thrust Shoes
Thrust Collar
Stationary
Casing
Normal
Thrust
Shims for Thrust Float

Inactive Thrust Shoes
The diagram in Fig. 4-21 is representative of the thrust bearing arrangement in most centrifugal compressors, many large pumps, and small turbines.
This configuration consists of a removable thrust collar that is mounted on the
shaft with an interference (shrink) plus a keyed fit. If the thrust collar becomes
damaged, it may be replaced with a spare collar. The active and inactive thrust
shoe assemblies are rolled into the casing on either side of the single thrust collar. Precision ground shims are installed behind the shoes to establish the proper
axial position of the rotor with respect to the stator, and to establish the thrust
float indicated on Fig. 4-21. For fluid film thrust bearings, the thrust or axial
float is normally in the vicinity of 15 Mils (0.015 Inches). Thrust floats of less
than 10 Mils (0.010 Inches) or greater than 20 Mils (0.020 Inches) are seldom
encountered on fluid film thrust bearings.
Due to the critical nature of the rotor thrust position, a proximity probe system is employed to monitor the relative rotor axial position via changes in the
probe DC gap voltage. As shown in Fig. 4-21, an axial probe is used to observe
190
Chapter-4
the end of the shaft. During bearing assembly and setup, a dial indicator is also
positioned axially, and changes in probe gap readings are compared with axial
movements measured with the dial indicator. For system redundancy during
operation, two thrust probes are normally installed, and the readings from both
probes are compared with the dial indicator. In virtually all cases, all three readings (2 probes plus 1 indicator) must compare within 1.0 Mil. Although this
might seem like a tight tolerance, it is almost 7% of a normal float zone. Hence,
this is a precision measurement that must be accurately calibrated and setup in
a static condition. If this is not done properly, then the information obtained from
the electronic probes when the machine is running will always be questionable.
Normal
Thrust
Inactive Thrust Collar
Shaft
Active Thrust Collar
Thrust Float
Thrust Probe
Fig. 422 Typical Dual

Thrust Collar Assembly
On Horizontal Machinery
Journal
Bearing
Active Thrust Shoes
Shims for Axial Position

Shims for Thrust Float
A thrust bearing configuration commonly used on large steam and gas turbines is depicted in Fig. 4-22. In this arrangement, dual thrust collars are
employed. These collars are generally integral with the shaft forging. Hence, if
these thrust collars are damaged, the rotor or stub shaft might have to be
scrapped and replaced. The journal bearing in Fig. 4-22 is mounted between the
thrust shoes, and the stationary casing supports the bearings, as well as the
shims mounted behind each set of thrust shoes. Once again, one or more axial
proximity probes are used to measure rotor thrust position. Due to the differences in thrust bearing configurations, the axial probe shown in Fig. 4-22 will
display an increase in gap voltage as the thickness of the active thrust pads
diminish under normal thrust loads. However, the thrust bearing shown in Fig.
4-21 will display a decrease in axial probe gap voltage as the thickness of the
active pads diminish under load. Depending on the thrust bearing configuration,
the direction of the normal thrust loads, and the location of the axial proximity
probes the probe gap voltage may either increase or decrease as thrust bearing attrition occurs. Both situations can occur, and the machinery diagnostician
must have full documentation of the mechanical setup in order to be confident in
the on-line DC probe gap information.
Thrust bearings in large horizontal machines nominally contain 40 to 80
Mils (0.040 to 0.080 inches) of babbitt. In most cases, the babbitt is considered as
the sacrificial element. During a thrust failure, the machine should coastdown
Fluid Film Thrust Bearings
191
on the available babbitt, and not allow contact between other parts of the rotor
and the casing. The specific machine and probe setup for proper thrust monitoring is presented in chapter 6.
Another common application for a fluid film thrust bearing occurs on large
vertical machines such as motors or generators. These vertical units are usually
directly coupled to pumps or turbines, and the entire weight of the machinery
train is supported by a single thrust assembly at the top of the upper machine. A
typical installation is shown in Fig. 4-23. The radial guide bearings in this type
of machinery often consist of a series of tilt pad bearings, and the thrust bearing
usually consists of segmented pads. The thrust bearing may be a fixed geometry
or a tilting pad configuration. In either case, an oil bath is provided for the thrust
bearing, and high pressure oil lifting jets are often used to supplement the oil
bath during startup and shutdown.
CW Rotation
Axial Probes
1Y
Thrust Runner
A-1
A-3
1X
K
Guide Bearing
Thrust Bearing
120
45
A-2
75
Casing
Fig. 423 Typical Fluid

Film Thrust Bearing On
Large Vertical Machinery
Shaft

Are Viewed From the Top
of the Machine Looking Down
Vertical load support for this machine occurs at the thrust runner shown in
Fig. 4-23. Shaft weight is essentially hung from the center of the thrust runner,
and this mechanical element is positioned over the fluid film thrust bearing.
Since smooth operation of this machine often depends upon proper behavior of
the thrust runner, it is an ideal location to measure vibration and position. As
noted in Fig. 4-23, X and Y probes are mounted in a true horizontal direction to
observe radial motion of the thrust runner. A Keyphasor probe is installed at
the same angular location as the X axis probe for train timing measurements.
A group of three axial probes are mounted in the true vertical direction
observing the top of the thrust runner. On the horizontal machines previously
discussed, only one axial thrust position probe was required for thrust position
measurement. Two thrust probes were normally installed on horizontal
machines to provide full redundancy for the axial measurement. This is particularly necessary to support the voting logic used on monitoring systems setup for
automatic trip. However, on a large vertical machine, there are very few installations that incorporate automatic trip based upon changes in axial position. For
192
Chapter-4
Fig. 424 Variation In

Vertical Thrust Position Of
Thrust Runner As A Function Of Generator Load
Vertical Thrust Runner Position (Mils)
-6
-7
J
B
P
-8
-9
-10
-12
-13
-14
Probe A-2
Probe A-3
BJ
P
-11
BPJ
Probe A-1
BJ
P
BJ
P
J
BP
J
BP
BJ
P
BPJ
P
BJ
0
10
12
14
16
18
20
Generator Load (Megawatts)
these vertical machines, three probes are installed above the thrust runner to
determine the relative level of the thrust runner. Since three points determine a
plane, any significant deviation in differential gap voltages from the three axial
probes is easily interpreted as a tilting of the thrust runner.
For instance, Fig. 4-24 describes the thrust position changes measured by
three axial probes mounted on a unit configured like Fig. 4-23. In this case, the
three proximity probes provide consistent information. Clearly, the level of the
thrust runner remains constant as load is varied on this turbine generator set.
The same transducers may also be used to examine the axial vibration of the
thrust runner. For example, running speed 1X vectors were obtained at slow roll
speeds of 40 to 50 RPM, and at synchronous speed of 277 RPM at 14 megawatts.
The full speed vectors were corrected for slow roll, and the runout compensated
1X vectors are presented in Table 4-6.
The probe angle corrections represent the angular spacing between the
three probes. These static angles may be used to correct to dynamic vectors to a
common location. In this case, the probe A-1 was selected as the common reference point, and the vectors from A-2 and A-3 were adjusted to the equivalent
angular position of A-1. The resulting three corrected vectors are totally consistent in terms of both amplitude and angle. The final 1.5 Mil synchronous vector
is considered to be quite acceptable for this class of machine.
Table 46 Runout Compensated Axial 1X Vectors With Probe Angle Correction At 14 MW
Axial Transducer
Identification
Runout Compensated
1X Vector
Probe Angle
Correction
1X Vector with Static

Angle Correction
Probe A-1
1.49 Mils,p-p @ 307
1.49 Mils,p-p @ 307
Probe A-2
1.43 Mils,p-p @ 65
240
1.43 Mils,p-p @ 305
Probe A-3
1.49 Mils,p-p @ 188
120
1.49 Mils,p-p @ 308
193
ROLLING ELEMENT BEARINGS

Most large process machines rely on fluid film bearings to support the rotating or reciprocating elements. When fluid film bearings are properly designed,
installed, and operated, their life expectancy is quite long. However, another
class of bearings are normally installed on smaller machines, or machines that
require close tolerances, or minimal relative motion between parts (e.g., precision machine tools such as lathes or milling machines). Categorically, these bearings are identified as rolling element bearings, but they are commonly referred
to as ball bearings, roller bearings, tapered roller bearings, etc. These bearings
all contain a series of internal elements that roll between an inner and an outer
race. They may depend on external lubrication, or they may be internally packed
with grease. Bearings of this type run with very close clearances, and they provide high stiffness combined with low damping to the rotor system. The application of high loads, shock loads, or bearing attrition due to extended run times
will expand the internal clearances, and result in bearing failure. Due to these
fundamental characteristics, this class of bearings is considered to have a finite
lifespan.
Since rolling element bearings have a limited life, it is desirable to monitor
these bearings for early indication of impending failures. It has been historically
demonstrated that vibration analysis is an effective tool for this type of monitoring. In addition, vibration analysis is useful for analysis of bearing faults. In
many instances, the machinery diagnostician may be able to successfully predict
the remaining life of a bearing. This allows maintenance to be performed in a
scheduled and cost-effective manner rather than in reaction to a failure.
Contact Angle
Outer
Race
Cage
Fig. 425 Typical Rolling

Element Bearing Configuration
Ball Diameter - Dball
Bearing
Centerline
Pitch Diameter - Dpitch
Inner
Race
Shaft
Centerline
194
Chapter-4
Within any bearing, the primary dynamic excitation occurs at shaft rotational frequency. Analysis of the vibratory behavior of a shaft in a fluid film bearing often centers around the running speed vibration, and harmonics of this
primary excitation. However, in a rolling element bearing, the fundamental shaft
vibration is supplemented by the mechanics inherent with the additional moving
bearing parts. For instance, consider Fig. 4-25 that depicts a typical ball bearing.
In this example, assume that the outer race remains stationary, and that the
inner race rotates at shaft frequency. It is clear that the supplemental frequencies emitted by this type of bearing must consider the geometry of the bearing in
addition to the fundamental shaft rotational speed RPM. This would include the
number of rotating balls N, the ball diameter Dball, and the ball pitch diameter
Dpitch. The contact angle between the balls and the races is also significant in the
determination of the fundamental defect frequencies. This contact angle is identified as angle in Fig. 4-25. Intuitively, the running position of a rolling element
bearing is dependent upon the radial and axial forces applied to the bearing. If
this ratio is changed due to variations in either the radial or the axial forces, the
load contact angle across the bearing will be influenced.
In addition, the basic model for this bearing requires a tight clearance fit,
with the internal balls rolling (and not sliding) in the raceways as the shaft
turns. If the balls are sliding, this is indicative of excessive bearing clearance,
and the general equations for specific defects are no longer applicable.
The traditional equations for the repetition rate of various defects were formulated in the 1960s. Based upon the bearing geometry, and the rotational
speed, it can be shown that a defect in the bearing outer race will generate a frequency Ford that may be computed with the following expression:
D ball
N RPM
F ord = --------------------------- 1 ------------------ cos
D pitch
2
(4-18)
For bearing defects on the inner race, the emitted frequency Fird may be
determined with the next equation:
D ball
N RPM
F ird = --------------------------- 1 + ------------------ cos
2
D pitch
(4-19)
From a vibration measurement standpoint, the outer race defects normally

appear at higher amplitudes than the inner race defects. These types of flaws are
often combined with vibration at the ball spin frequency that is coincident with
the frequency for a ball defect. This ball defect frequency Fbd may be calculated
with equation (4-20):
D ball
RPM D pitch
2
F bd = --------------- ------------------ ------------------ ( cos )
2
D
D
ball
pitch
(4-20)
195
The fundamental train frequency, or more commonly the cage defect frequency Fcd may be determined by:
D ball
RPM
F cd = --------------- 1 ------------------ cos

D pitch
2
where:
Ford
Fird
Fbd
Fcd
N
RPM
Dball
Dpitch
(4-21)
= Frequency of Outer Race Defect (Cycles / Minute)

= Frequency of Inner Race Defect (Cycles / Minute)
= Frequency of Ball Defect (Cycles / Minute)
= Frequency of Cage Defect (Cycles / Minute)
= Number of Contained Balls or Rollers (dimensionless)
= Rotational Speed of Bearing Inner Race (Revolutions / Minute)
= Diameter of Ball or Roller (Inches)
= Pitch Diameter of Balls or Rollers (Inches)
= Bearing Load Contact Angle (Degrees)
The defect frequencies computed with equations (4-18) through (4-21) carry
the engineering units of Cycles per Minute (CPM). If frequencies in Cycles per
Second (Hz) are desired, the results from the last four equations may be divided
by 60. Once again, these equations are for a stationary outer race, and a rotating
inner race, with rolling and not sliding balls or rollers. It must always be recognized that ball bearings in real machines with reasonable loads do actually slip,
and the measured and calculated frequencies will probably not be identical.
Clearly, the frequencies computed with these equations are non synchronous. For a typical 100 HP motor with less than 20 balls, the outer race defect frequency computed by (4-18) will be a high frequency component in the vicinity of
40 to 45% of rotating speed times the number of balls ( 0.45xRPMxN). The inner
race defect frequency from equation (4-19) will also be a higher frequency component at about 55 to 60% of speed times the number of balls ( 0.6xRPMxN). The
ball spin or ball defect frequency will be in the order of 3 to 4 times running
speed ( 3.5xRPM), and the cage defect frequency will be approximately 45% of
running speed ( 0.45xRPM).
For a stationary inner race and a rotating outer race, equations (4-18)
through (4-20) remain the same. The expression for a cage defect with a fixed
inner race is given by equation (4-22). In this situation, the frequency of the cage
defect will be greater that one half of running speed ( 0.55xRPMout).
D ball
RPM out
F cd_ fir = ----------------------- 1 + ------------------ cos
2
D pitch
(4-22)
where: Fcd_fir = Frequency of Cage Defect with Fixed Inner Race (Cycles / Minute)
RPMout = Rotational Speed of Bearing Outer Race (Revolutions / Minute)
Bearing dimensional data is often available directly from the bearing manufacturer. Some suppliers publish lists of these fault identification frequencies
for their bearings. Whether tabular lists or discrete calculations are used, the
machinery diagnostician should recognize that a new bearing will probably
196
Chapter-4
exhibit some, or all of these frequencies at very low amplitudes. As defects occur,
the amplitudes at the associated defect frequencies will increase. As the defects
continue to grow with time, the observed frequencies will often shift as the load
distribution changes, and the balls (or rollers) begin sliding instead of rolling. In
this condition, overall vibration levels are probably unacceptable, and the unit
should be shutdown for bearing replacement.
The commonly observed types of damage on rolling element bearings result
in craters or spalls on the raceways. As the rolling elements pass over these
indentations, impact or shock pulses are generated. In some cases, the defect frequencies identified by the previous equations may not be visible in a frequency
spectrum. However, as noted by John Mitchell10, if this housing vibration data is
observed in the time domain, the repetitive pulses are easily distinguished. In
time domain analysis, the period of the four defect frequencies should be used to
identify the origin of the pulse patterns.
For the sake of completeness, it should be mentioned that other transducers, measurements, and data processing techniques are used to examine the
vibratory behavior of rolling element bearings. Methods such as spike energy,
envelope detection, shock pulse, and various demodulation techniques are commonly employed. From a practicality standpoint, the diagnostician should evaluate any bearing analysis tool on the basis of performance. Specifically, the
following question should always be asked: can the instrumentation successfully
and consistently identify mechanical failures on the rolling element bearings?
BEFORE CONSIDERING BEARING REDESIGN

In the opinion of the senior author, the Original Equipment Manufacturers
(OEMs) do a credible job of designing, building, and installing bearings. Within
the industrial end user community there seems to be an overt tendency to redesign, and continually attempt to improve on the OEM bearings. In multiple
cases, the bearing redesign has rendered the machinery inoperative. In other situations, the operating speed range of the machinery has been severely limited
following the installation of presumably improved bearings. In all cases, the
machinery diagnostician must approach bearing problems carefully, and conduct
a methodical engineering analysis of the problem. Specifically, the following
items should always be thoroughly examined and in some cases re-examined:
Check that the oil console or reservoir contains the correct lubricant.
Check the oil quality for proper density, viscosity, water content, etc.
Check the oil for the presence of any foreign materials.
Check for proper oil supply pressure, temperature, and system control.
Check the oil flow rate to each bearing, and verify that orifices are properly
installed, and that orifice diameters are both reasonable and correct.
10 John S. Mitchell, Introduction to Machinery Analysis and Monitoring, second edition (Tulsa,
OK: Pennwell Publishing Company, 1993), pp. 241-249.
Before Considering Bearing Redesign
197
Check the oil drain temperatures, and relative flow rates.

Check that the bearing is properly installed with respect to shaft rotation.
Check that anti-rotation pins are properly installed with respect to rotation.
Check that the shaft to bearing clearance is correct.
Check that the bearing to housing clearance is correct.
Check that the bearing liner is not distorted or warped.
Check that the bearing splitline is not sealed with RTV, silicone, or other
incompressible sealants. Use a thin grade of Permatex sealant for this job.
Check for other mechanical changes in the train that would influence bearing load (e.g., changing a gear coupling to a large diaphragm coupling).
Check rotor balance records, and the last set of transient startup data.
Check coupling alignment for proper cold offset and hot running position.
Check for proper temperatures from imbedded thermocouples or RTDs.
Check bearing temperature trends (day to night, week to week, etc.).
Check to be sure that shaft is level when hot and running.
Check bearings, seals, and couplings for evidence of electrical discharge.
Check pads and backing for evidence of wear, cracking, or fretting.
Check bearings for evidence of edge wear across bearings and machines.
Check for proper position of the journal within the bearing with prox probes.
Check shaft vibration for normal 1X running speed vibration vectors.
Check shaft vibration for any abnormal frequency components.
Check the attachment of the bearing housing to the casing and/or baseplate.
Check grout condition, and the attachment of baseplate to foundation.
If these checks are followed, and all identified problems corrected, the
necessity to redesign or continually replace bearings will be greatly reduced. Bad
habits seem to develop over time, and both operations and maintenance personnel have a tendency to get complacent. In many instances this will allow small
oversights to turn into major problems. Hence, before jumping into a major redesign effort, the use and abuse of the current bearings should be examined.
There are situations when the bearings really do require an upgrade. If
rotor or coupling changes are to be implemented, if the process loads or the lube
and seal oil system are to be modified, or if greater reliability is required then
the existing bearings should be audited for potential areas of improvement. The
addition of ball and socket bases for tilt pad bearings, the use of micro-babbitt, a
change in bearing metallurgy for improved heat transfer, or providing directed
lubrication are all common modifications that may benefit a particular bearing
installation. In some applications, the installation of new bearing designs such
as the Flexure Pivot Bearings described by Zeidan and Paquette11 may be highly
beneficial. In other cases, an additional five gallons per minute of oil flow may be
all that is required. Once again, the machinery diagnostician is advised to proceed with logic, and proper engineering discipline.
11 Fouad Y. Zeidan, and Donald J. Paquette, Application of High Speed and High Performance
Fluid Film Bearings In Rotating Machinery, Proceedings of the Twenty-Third Turbomachinery
Symposium, Turbomachinery Laboratory, Texas A&M University, College Station, Texas (September 1994), pp. 209-233.
198
Chapter-4
BIBLIOGRAPHY
1. Allaire, Paul E. and Ronald D. Flack, Design of Journal Bearings for Rotating
Machinery, Proceedings of the Tenth Turbomachinery Symposium, Turbomachinery
Laboratories, Texas A&M University, College Station, Texas (December 1981), pp.
25-45.
2. Barrett, L.E and John C. Nicholas, The Effect of Bearing Support Flexibility on
Critical Speed Prediction, ASLE Transactions, 1984 Joint Lubrication Conference,
San Diego, California, February 1984 (revised June 1984).
3. Gunter, Edgar J., Dynamic Stability of Rotor Bearing Systems, NASA Report SP113, 1966.
4. Gunter, Edgar J. and R.G. Kirk, The Effect of Support Flexibility and Damping on
the Dynamic Response of a Single Mass Flexible Rotor in Elastic Bearings. Report
No. ME-4040-106-72U, University of Virginia, 1972.
5. McHugh, James D., Principles of Turbomachinery Bearings, Proceedings of the
Eighth Turbomachinery Symposium, Gas Turbine Laboratories, Texas A&M University, College Station, Texas (November 1979), pp. 135-145.
6. Mitchell, John S., Introduction to Machinery Analysis and Monitoring, second edition, pp. 241-249, Tulsa, OK: Pennwell Publishing Company, 1993.
7. Nicholas, John C., Tilting Pad Bearing Design, Proceedings of the Twenty-Third
Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University,
College Station, Texas (September 1994), pp. 179-194.
8. Nicholas, John C., John K. Whalen, and Sean D. Franklin, Improving Critical
Speed Calculations Using Flexible Bearing Support FRF Compliance Data, Proceedings of the Fifteenth Turbomachinery Symposium, Turbomachinery Laboratory,
Texas A&M University, College Station, Texas (November 1986), pp. 69-78.
9. Salamone, Dana J., Journal Bearing Design Types and Their Applications to Turbomachinery, Proceedings of the Thirteenth Turbomachinery Symposium, Turbomachinery Laboratories, Texas A&M University, College Station, Texas (November
1984), pp. 179-188.
10. Zeidan, Fouad Y. and Donald J. Paquette, Application of High Speed and High Performance Fluid Film Bearings In Rotating Machinery, Proceedings of the TwentyThird Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University, College Station, Texas (September 1994), pp. 209-233.
Analytical Rotor Modeling
achinery shaft vibration characteristics

reflect the combined interaction of rotating assemblies with various fluids, and
stationary machine elements. Often these characteristics can be segregated and
quantified with appropriate measurement techniques. However, there are situations where the required data cannot be obtained due to other restraints. For
instance, the critical speed of a rotor cannot be determined due to the economic
impact of shutting down the unit. In another case, the machine running speed
cannot be increased to investigate the effects of higher order modes. Similarly,
changing mechanical parts, such as bearings or couplings, cannot be realistically
evaluated based upon hardware trial and error substitutions. From a problem
solving standpoint, the machinery diagnostician may begin an investigation by
comparing the predicted machinery behavior with the measured vibratory characteristics. In these types of situations, the development and utilization of mathematical models to simulate the mechanical systems may be mandatory.
During the latter half of the twentieth century, various computational techniques have been developed and refined into working tools. Techniques such as
Transfer Matrix, Computational Fluid Dynamics (CFD), and Finite Element
Analysis (FEA) provide significant capability for modeling physical systems. In
order to provide any overview of some analytical techniques and the generic
types of available software for the evaluation of rotating machinery behavior, the
following descriptions and examples of machinery calculations are presented.
MODELING OVERVIEW
The mathematics associated with simple mechanical systems were presented in chapter 2. For a single degree of freedom system consisting of a
undamped mass hanging from a spring, it was concluded that the system natural
frequency was a function of stiffness and effective mass. Paraphrasing equation
(2-44), it can be stated these three variables are related in the following manner:
Stiffness
Natural Frequency -------------------------Mass
(5-1)
199
200
Chapter-5
From this expression it is clear that stiff elements have high natural frequencies, and flexible parts have lower resonances. Similarly, heavy elements
will display low natural frequencies, and lighter components will exhibit higher
values. For example, the natural frequency for a large steel girder may be 5 Hz,
and a small tuning fork may emit a tone equal to a frequency of 500 Hz. In either
case, the geometry and mechanical configuration define a combination of stiffness and mass that yield a discrete natural frequency.
As system complexity increases, the intricacies of the descriptive equations
also expand. The simple expression of equation (5-1) is replaced by a matrix solution, and items such as inertia and rotational forces are included. It is apparent
that mass, and the distribution of that mass is critical to the solution. Furthermore, shaft stiffness must be determined, and combined with the mass properties. When these elements are defined, the calculation of undamped critical
speeds may be performed. These calculations do not include damping, they do
not allow asymmetric stiffness, and they do not consider specific forcing functions. However, an undamped analysis provides an overview of the natural frequencies associated with a mass distribution at a selected support stiffness, plus
the shaft mode shapes for each resonance and stiffness combination.
The next level of analytical modeling programs incorporates asymmetrical
stiffness coefficients, plus damping from the bearings, foundation, or process
fluid. The oil film coefficients are calculated for the bearing configuration, and
support coefficients are normally measured. It is important to include damping
into the calculations. This energy dissipater allows the examination of damped
critical speeds, the computation of rotor stability, plus damped mode shapes.
Although this calculation refinement does a credible job of finding the Eigenvalues (natural frequencies and damping), it does not accept actual forcing functions such as unbalance, skewed wheels, or bowed rotors.
Adding forced vibration mechanisms requires another evolution of the program structure. Within forced synchronous response programs, dimensional
forces are used to compute rotor response in displacement units. Hence, the
anticipated motion (vibration) at any speed, and at any position along the rotor
may be computed. The accuracy of these calculations is often determined by a
comparison with measured shaft vibration data at specific locations. This verification of calculations by measured vibration response characteristics is an often
ignored step. In actuality, the verification of results is vital to the development of
confidence in the calculations. It also helps to define areas where the custom
analytical programs require improvement or modification.
Clearly, the construction of a successful analytical model requires the integration of numerous calculations into a cohesive set of results. A single computer
program does not contain the entire model. In fact, many calculations are performed in separate environments from the rotor dynamics calculations. For
instance, cross-sectional inertias may be computed in a mathematical program,
and rotor dimensional configuration may be initially established in a spreadsheet program. In most cases, several different programs are required to perform
the full array of calculations. In the remainder of this chapter, the primary rotor
dynamics programs will be discussed, and illustrated with field examples.
Undamped Critical Speed
201
UNDAMPED CRITICAL SPEED

In order to understand the fundamental behavior of a rotor system, it is
mandatory to determine the frequency of the system critical speeds, and the
associated mode shapes. One of the easiest tools to begin such an investigation is
an analysis of the undamped critical speeds. To appreciate the current array of
computations it is meaningful to briefly review the origin of these calculations.
Historically, one of the earliest procedures for critical speed calculations
was developed by A. Stodola circa 1925. This graphical construction technique
required the estimation of a rotor deflection curve, followed by the computation
of relative kinetic and potential energy. By equating the kinetic to the potential
energy, a first critical speed was approximated. This technique assumed rigid
bearings, gyroscopic effects were ignored, affects of coupling weights could not be
determined, and higher order modes could not be successfully addressed.
In 1944 Myklestad1 published a paper on calculating natural modes of airplane wings and other beams. That was followed in 1945 by Prohl2 and his paper
on critical speeds of flexible rotors. By 1954, Prohl and Myklestad initiated
development of calculation techniques for lateral critical speeds. This was followed by the combined work of Holzer, Prohl and Myklestad for torsional critical
speeds. In both cases, the transfer matrix method was applied, and this provided
a significant improvement in the ability to predict rotor resonant behavior.
Although the required matrix calculations are quite complex, the final results
reflect the sophistication of the model. Evolution of these techniques have progressed through various stages. During the 1980s, progress has been closely
associated with the development of smaller and faster computers that can adequately process the matrix calculations.
The undamped critical speed program (CRITSPD) used in this text was
developed by Edgar J. Gunter3. In the formulation of this program, the transfer
matrix is divided into a point matrix containing the mass, inertia, and bearing
properties, plus a massless field matrix containing the shaft properties. The use
of a massless field transfer matrix for shaft properties has been used by Lund4 in
rotor stability and unbalance computer codes, and it is also described by Thomson5. Although the massless field matrix is a considerable improvement over the
hyperbolic continuum formulation, it suffers from numerical difficulties when
large numbers of stations are employed. Gunter incorporated a unique automatic scaling procedure to minimize the errors, and to allow the successful mod1 N.O. Myklestad, A New Method of Calculating Natural Modes of Uncoupled Bending Vibration of Airplane Wings and Other Types of Beams, Journal of the Aeronautical Sciences, Vol. 11, No.
2 (April 1944), pp. 153-162.
2 M.A. Prohl, A General Method for Calculating Critical Speeds of Flexible Rotors, Journal of
Applied Mechanics, Vol. 12, Transactions of the ASME, Vol. 67 (September 1945), pp. A142-148.
3 E. J. Gunter and C. Gareth Gaston, CRITSPD-PC, Version 1.02, Computer program in MSDOS by Rodyn Vibration, Inc., Charlottesville, Virginia, August, 1987.
4 J.W. Lund, Modal Response of a Flexible Rotor in Fluid Film Bearings, Transactions American Society of Mechanical Engineers, Paper No. 73-DET-98 (1973).
5 William T. Thomson, Theory of Vibration with Applications, 4th Edition, Prentice-Hall, Englewood Cliffs, New Jersey, 1993.
202
Chapter-5
eling of 100 station rotors on desktop computers.

The CRITSPD program includes the primary effects of rotor mass, shaft
and bearing flexibility, plus transverse and polar inertia. In addition, the
undamped analysis can incorporate synchronous or non-synchronous gyroscopic
effects, shear deformation, plus variable bearing and seal stiffness. Couplings,
impellers, thrust disks, and shaft spacers can be included by several different
modeling schemes. Hollow shafts plus variable material densities, and various
boundary conditions for each end of a rotor are accommodated. It is possible to
compute synchronous critical speeds, planar modes, plus order tracking. The program calculates, and compares total kinetic and potential energy for each mode,
plus the undamped critical speeds, and associated rotor mode shapes. In addition, the strain energy distribution, deflection, slope, moment, and shear at each
station is computed and presented. Graphical outputs include the rotor cross section, a summary mode shape diagram for all criticals, and a mode shape for each
individual critical speed, as illustrated in Fig. 5-1.
Fig. 51 Undamped
Steam Turbine Mode
Shape Output From
CRITSPD Program
Although undamped critical speed analysis can provide significant insight

into the behavior of the machinery, it does have inherent limitations. For example, it does not include the effects of forces such as mass unbalance or internal
synchronous mechanisms such as shaft bows. This analysis does not consider
damping, or cross-coupled influences from bearings, seals, or aerodynamics.
Undamped critical speed calculations are a simplification of the system mathematical model, which in turn is a further simplification of the real mechanical
system. The parameters neglected in an undamped analysis can alter system
resonances. These parameters can control the amplitudes at the resonant frequencies, and they may be responsible for instabilities. Hence, the results from
an undamped analysis must be considered in the proper context.
One of the main utilizations for undamped critical speed calculations
resides in the ability to quickly compare the change in natural frequencies as a
function of support stiffness. As mentioned in chapter 3, the undamped critical
speeds may be computed for a large range of stiffness values, and the results
plotted as a family of curves. This type of summary plot is normally referred to
as an undamped critical speed map. For instance, Fig. 5-2 depicts this type of
plot for a 22,500 pound gas turbine rotor. Calculations were performed at 1, 2,
203
20,000
10,000
Rotor Bending
Pivotal
5th
1,000
4th
3rd
Translational
2nd
1st
200
100,000
1,000,000
10,000,000
50,000,000
Fig. 52 Undamped Critical Speed Map For Single Shaft Gas Turbine
and 5 intervals to allow for even spacing of points on the logarithmic stiffness
scale. The stiffness calculation range began at 100,000 Pounds/Inch that is certainly less than any machine stiffness, and it extends to 50,000,000 Pounds/Inch
that is greater than any potential machine element stiffness.
The first five critical speeds for each stiffness were plotted, and the points
connected for each natural frequency. In this manner, a log-log plot of stiffness
versus each critical speed is produced. The first two modes reveal a variation of
critical speeds with stiffness. It is reasonable to conclude that these are bearing
dependent modes. This observation is confirmed by the detailed calculations that
show the majority of the strain energy contained within the bearings. For the
higher order criticals, and the stiffer portion of the 1st and 2nd modes, the natural frequencies display minimal variation with support stiffness. These conditions are indicative of resonant modes that are primarily controlled by shaft
stiffness. This is important information, since in the first scenario, bearing
changes could alter the rotor critical speed(s). This type of mechanical change is
reasonably inexpensive to perform. However, in the second situation, shaft modifications would be required to change the critical speeds, and this type modification can be expensive as well as technically complicated.
The undamped critical speed map is also used to examine the relationship
between the calculated resonant frequencies, and the operating speed range as
shown in Fig. 5-3. In this diagram, the rotor support stiffness is shown for three
different conditions. First, the minimum or soft condition of 500,000 Pounds per
Inch is shown. Second, the horizontal bearing stiffness Kxx is plotted, followed by
the third line of vertical bearing stiffness Kyy. The support stiffness values help
to define the potential operating range of the machine. This information allows
204
Chapter-5
20,000
10,000
5th
4th
3rd
2nd
5th
4th
3rd
1st
Operating
Speed
Range
1,000
K
xx
2nd
1st
K
min
K
yy
Bearing Stiffness Curves
200
100,000
1,000,000
10,000,000
50,000,000
Fig. 53 Undamped Critical Speed Map For A Single Shaft Gas Turbine With Principal
Bearing Stiffness Curves And Normal Operating Speed Range Identified
further quantification of the anticipated turbine characteristics.

The computation of natural resonant frequencies must also be compared
with the potential excitation frequencies. It is perfectly understandable that natural frequencies will remain essentially dormant until some type of excitation
coincides with these critical speed(s). The applied excitation may be a broad band
forcing function, such as a steam turbine subjected to a slug of water. The excitation may also be a discrete frequency such as rotational speed mass unbalance
(1X), or an excitation due to specific geometry within the machine (e.g., pump
vane passing activity).
This type of evaluation is performed on an interference plot that is commonly referred to as a Campbell diagram. Fig. 5-4 depicts such a diagram based
upon the previous critical speed map. In this diagram, the natural frequencies
are plotted on the vertical axis, and the excitation frequencies are shown along
the horizontal axis. For instance, the five critical speeds at normal operating conditions were extracted from the critical speed map, and they are shown as horizontal lines in the Campbell Diagram.
Excitations at running speed 1X unbalance, 2X misalignment, plus 5X, and
10X excitations are presented as the slanted lines. The intersection between the
horizontal natural frequency lines, and the slanted excitation lines represents a
potential case for the appearance of the specific resonance. Naturally, this type of
diagram is often expanded to include higher frequency excitations, and other resonant frequencies such as structural or torsional critical speeds.
The number of resonant frequencies that exist on most large machinery
trains can be staggering. This is particularly true for bladed machines such as
205
12,000
1X
2X
2X
5X
5X
4th
8,000
Excitations
10,000
1X
10X
6,000
Rotor Bending - 3rd

10X
5th
4th
4,000
3rd
Pivotal - 2nd
2,000
2nd
Translational - 1st
1st
Natural Frequencies
Natural Frequency (Cycles/Minute)
5th
0
10,000
20,000
30,000
40,000
50,000
Excitation Frequency (Cycles/Minute)
Fig. 54 Campbell Diagram For Single Shaft Gas Turbine At Design Stiffness Values
steam or gas turbines where the individual blades and the segmented groups of
blades exhibit a variety of tangential and axial modes of vibration. Other complications, such as separate and distinct horizontal and vertical rotor balance resonances, disk or impeller resonances, plus external resonances can substantially
increase the number of potential resonant frequencies.
Furthermore, when the sources of excitation are examined, the problem
becomes even more complicated. For instance, harmonics of the fundamental
excitations may be generated within the machine. In addition, specific frequencies may interact to form distinct beat or modulation frequencies. When all of the
possible excitation frequencies are considered, the potential for exciting the
expanded group of resonances increases dramatically.
In essence, an interference diagram such as the Campbell plot of Fig. 5-4
may become congested with all of the inherent excitations and resonances. It
might even be concluded that the machinery cannot be operated at any reasonable speed due to the coincidence of excitations and natural frequencies. Obviously, this is not an acceptable conclusion, and it is not representative of the
varied array of operating machinery trains.
In order to address this dilemma, it is suggested that machinery behavior
be examined in two different categories. The first category would consist of the
major rotor balance resonances (lateral critical speeds), and the potential low
frequency excitations. This part of the analysis follows the scheme presented in
the Campbell plot of Fig. 5-4, but additional detail is necessary to determine the
severity of the interference points.
For example, the rotor model should be expanded to include a forced
response analysis, as discussed later in this chapter. By varying the definable
206
Chapter-5
excitations (e.g., rotor bow, unbalance at various rotor locations, disk skew, etc.),
the machinery diagnostician should be able to evaluate the vibration severity for
anticipated forcing functions. This approach will identify the major or significant
resonances, and allow other interference points to be discounted.
The second category of machinery behavior considers the higher frequency
characteristics associated with turbine blades or compressor wheels. In this complex mechanical domain, the traditional two-dimensional Campbell diagram
should be expanded into a three-dimensional SAFE6 diagram (acronym for
Singhs Advanced Frequency Evaluation). This analytical tool combines the twodimensional Campbell plot with a third dimension of nodal diameters or mode
shapes. The three-dimensional intersection of natural resonant frequencies, excitation frequencies, and nodal diameters are then used to identify potential resonant conditions. The inclusion of the blade mode shape allows the diagnostician
to ignore the majority of the interfere points, and identify the frequencies and
modes of greatest potential vibration.
Case History 10: Mode Shapes For Turbine Generator Set
Undamped critical speed calculations are relatively easy to setup and run.
As noted, they do not include synchronous forcing functions, and the support
stiffness characteristics represent a simple condition. However, these calculations can provide significant visibility into the behavior of rotating systems.
For rotors supported between bearings, and for overhung assemblies, the
mode shapes discussed in chapter 3 make intuitive sense. Armed with the knowledge of the general rotor configuration, and the relative bearing stiffness, the
anticipated mode shapes may be estimated. Even though the frequencies may
not be calculated, the mode shapes for simple systems can be deduced. However,
for more complicated systems, the shaft mode shapes may not be obvious.
For example, consider the turbine generator set depicted in Fig. 5-5. This is
a three bearing machine that runs at 3,600 RPM. The combined weight for both
rotors is 21,400 pounds, and a solid coupling is used between the two shafts. The
turbine is an extraction unit, with a surface condenser at the exhaust. The synchronous generator has collector rings mounted at the outboard end, and a sepa-
Fig. 55 Rotor Arrangement For Three Bearing Turbine Generator Set
6 Murari P. Singh and others, SAFE Diagram - A Design and Reliability Tool for Turbine Blading, Proceedings of the Seventeenth Turbomachinery Symposium, Turbomachinery Laboratory,
Texas A&M University, College Station, Texas (November 1988), pp. 93-101.
207
rate exciter. The entire machinery train is mounted on a mezzanine deck, and
the structure plus the bearing supports are compliant with stiffness values
approaching a minimum value of 600,000 Pound/Inch.
Over the operating history of this machinery train various problems have
occurred. The majority of the difficulties have been traced to generator unbalance problems, or high eccentricity at the solid coupling. Although successful
field balance corrections have been performed on both the turbine and the generator, the logic behind some of the weight corrections was not fully understood.
In an effort to resolve some of the issues, the train was retrofitted with X-Y
proximity probes as shown in Fig. 5-5. During startup, these transducers
revealed critical speeds that were in direct contradiction with historical conclusions. For example, the local personnel believed that the T/G Set had a critical
speed that began at 1,000 RPM, and lasted until well above 2,000 RPM. This
behavior was considered to be inconsistent with any expected response through a
single critical speed. In addition, the three planes of installed X-Y proximity
probes provided additional contrasting information.
In order to address these anomalies, the system was eventually subjected to
an undamped critical speed analysis. The computed mode shapes for the first
three critical speeds are presented in Fig. 5-6. The first mode was calculated to
Fig. 56 Undamped Shaft

Mode Shapes For The First
Three Critical Speeds On A
Three Bearing Turbine Generator Set
208
Chapter-5
be 1,210 RPM, which agreed with the measured value of 1,250 RPM. This critical
was always visible by the proximity probes installed at the #2 and #3 Bearings.
The probes at the #1 bearing are close to a nodal point, and this translational
resonance is not particularly visible at the front turbine bearing.
The measured second critical speed occurred at 1,650 RPM. This value is
virtually identical with the calculated second critical of 1,660 RPM. In addition,
the proximity probes mounted at bearings #2 and #3 always displayed an out of
phase behavior. This was not fully understood until the analysis was performed,
and the calculated pivotal mode shapes produced. From this second critical mode
shape, it is clear that the probes at #2 and #3 bearings are on opposite sides of a
shaft node, and a phase reversal must exist. Again, the analytical calculations
are consistent with the field shaft vibration response measurements.
Finally, eccentricity problems at the coupling always caused high vibration
amplitudes at slow speeds, and at speeds just below 3,600 RPM. The reason for
this behavior is evident from the calculated mode shape plots where a large
deflection in the coupling area is visible for the first critical at 1,210 RPM, plus
the third critical speed at a computed value of 3,400 RPM. This resonance was
fully corroborated by the proximity probe transient data that exhibited a resonance at 3,440 RPM.
Additional supporting evidence concerning the behavior of this turbine generator set was documented when balance weights were placed on each end of the
generator. The anticipated modal response with the balance weights was consistent with the shaft mode shapes described in Fig. 5-6. There are other correlations that may be extracted from this data set. However, the main point is that
undamped critical speed calculations provide an analytical tool that is directly
applicable to existing machinery. In many situations, it can provide the diagnostician with significantly more insight into the dynamic behavior of the rotating
machinery, and it also provides valuable modal information for field balancing.
Case History 11: Torsional Analysis of Power Turbine and Pump
The same general techniques used for undamped lateral calculations may
also be applied towards the computation of undamped torsional frequencies and
mode shapes. As stated in chapter 2, the basic equations for lateral and torsional
characteristics are similar. However, the calculation scheme, and interpretations
of results are somewhat different. In a lateral system, stiffness is expressed as
Pounds per Inch, and mass carries the units of Pound-Seconds2 per Inch. Within
a torsional analysis, the torsional stiffness is a torque per unit angle, with common units of Inch-Pounds per Radian. Polar inertia carries the units of PoundInch-Seconds2 per Radian, and this is analogous to mass in a lateral analysis.
Within a lateral analysis, the rotor stations with maximum motion are significant (as in the previous case history). During a torsional analysis, the nodal
points are meaningful since stress reversals occur across each torsional node.
Furthermore, in a lateral analysis, the mass and stiffness properties are typically confined to a single rotor, or rotor system with hard couplings. The influ-
209
ence of the flexible couplings between rotors are seldom included in a lateral
analysis. During a torsional analysis the inertia and torsional stiffness properties of the entire train are considered. Since inertia elements are essentially
fixed, the mechanical element used for alteration of torsional natural frequencies
often reverts to the coupling spool piece between machines.
Fig. 57 Mass Elastic Data For Power Turbine, Drive Through Gear Box, And Pump
For example, consider the mass elastic data shown in Fig. 5-7. This diagram
describes the lumped inertia at eight major machinery sections, plus the seven
interconnecting torsional springs. This data was based upon OEM specified values, plus independent calculations of cylindrical sections with the equations presented in chapter 3 of this text. The actual machinery in Fig. 5-7 represents a
processed water injection pump that is gas turbine driven via a straight through
gear box. The gas turbine is a two shaft unit, with no mechanical connection
between the gas generator and the power turbine. Hence, the drive end of this
train begins with the power turbine wheel.
Considering the relative size and mass of the machine elements, it is understandable that a majority of the system inertia is contained within this turbine
drive wheel. As noted, a gear box was attached to the turbine output shaft. Since
turbine and pump speeds were compatible, the gear box consisted of a single
drive through shaft element, with no speed change. The horizontally split pump
was originally a six stage unit that was de-staged to five stages to meet process
demands. The expected operating speed range for this pump varied from 5,180 to
6,800 RPM.
Various standards (e.g., API 617) recommend a 10% separation between
any torsional resonance and the operating speed range. Application of this criteria expands the above speed range to include a minimum torsional frequency of
4,660 RPM, combined with a maximum of 7,480 RPM. Before performing any
extensive calculations, it would be reasonable to estimate the first torsional fre-
210
Chapter-5
quency based upon the available mass elastic data. For instance, the major inertia occurs at the gas turbine wheel (40.13 Pound-Inch-Sec.2/Radian), and the
main coupling stiffness (6.67x106 Inch-Pound/Radian) would normally be varied
to control the torsional resonance frequency. If these values are placed in equation (2-103), a first torsional critical speed may be estimated as follows:
6
K tor
1
1
6.67 10 Inch-Pound/Radian
------ ---------------- ------ ----------------------------------------------------------------------------2
tor
2
J mass 2
40.13 Pound-Inch-Second /Radian
1
Cycles
Seconds
F c ------ 166, 210 = 64.89 ---------------- 60 ------------------ = 3, 890 CPM
Minute
Second
tor
2
Fc
This estimated speed of 3,890 CPM does fall below the 4,660 to 7,480 RPM
exclusion range, but the differential is uncomfortably small. Obviously, a full set
of undamped torsional resonance calculations are required to obtain sufficient
precision in the torsional natural frequency calculations. The most significant
results of these computations are in Fig. 5-8. At the top of this diagram, the torsional mode shape at the calculated first critical speed of 5,030 RPM is shown.
This frequency is much higher than the simple model estimate of 3,890 RPM. In
addition, the computed first critical speed falls well within the exclusion speed
range of 4,660 to 7,480 RPM. Clearly, this deviation demonstrates that a very
simple model may not properly represent the actual mechanical system.
The second undamped torsional critical speed appears at 16,990 RPM, as
indicated at the bottom of Fig. 5-8. This frequency is considerably higher than
Fig. 58 Undamped First

and Second Torsional Mode
Shapes For Power Turbine,
Drive Through Gear Element, and Coupled Five
Stage Centrifugal Pump
211
the operating speed range, and it is beyond excitation by twice rotational speed
oscillation (2 x 7,480=14,960 CPM). Hence, this torsional resonance should not
cause any distress. Although higher order torsional frequencies are not shown,
they should also be computed, and compared against potential excitations with a
Campbell or a SAFE diagram.
From this data, it is evident that the major problem resides with the coincidence of the first torsional critical and the previously described exclusion range
(operating speed range 10%). This predicted relationship is unacceptable, and
physical changes must be implemented to correct the deficiency. Unfortunately,
the machinery under discussion included existing equipment that was in the process of re-configuration to accommodate new operating conditions. Although
changes in these existing rotors were feasible, the economic considerations voted
heavily against any significant changes to the turbine or pump rotors. The drive
through gear box shaft was identical to other units at the same facility. Hence,
there was reluctance to change this drive shaft to a one-of-a-kind assembly.
The last candidate for modification was the load coupling to the pump. The
available couplings all contained hollow spool pieces with outer diameters that
varied between 4.16, and 5.20 Inches. The torsional stiffness values for these
couplings ranged from 6.67x106 to 7.00x106 Inch-Pound/Radian. This coupling
stiffness range was previously judged unacceptable due to the frequency interference. Hence, a new coupling must be provided that is both torsionally softer,
and still able to transmit the torque with acceptable stress values.
After the examination of potential stiffness changes, a nominal value of
2.00x106 Inch-Pound/Radian was selected as acceptable. The complete array of
undamped torsional calculations were repeated, and the results summarized in
Table 5-1. With this reduction in torsional stiffness, the calculated undamped
first critical was reduced from 5,030 to 4,160 RPM. This is approximately 20%
below the minimum operating speed, and outside of the exclusion speed range.
The higher order torsional resonances were influenced by this reduction in load
coupling stiffness, but the variations were insignificant. More importantly, there
was minimal interference between higher order excitations (e.g., pump vane
passing) and the undamped torsional criticals.
The final mechanical design by the coupling manufacturer included a solid
coupling spool piece with an overall length of 16.80 Inches, and an outer diameter of 2.27 Inches. The flanges at each end of this spacer were 0.53 Inches wide,
and 8.25 Inches in diameter. It is always a good idea to check the torsional stiffness provided by the vendor. In the vast majority of the cases, the stiffness proTable 51 Comparison Of Undamped Torsional Critical Speeds For Two Different Couplings
Coupling Type
Original Hollow Spool
New Solid Spool
Stiffness
1st Mode
2nd Mode
3rd Mode
4th Mode
(Inch-Lb/Rad)
(RPM)
(RPM)
(RPM)
(RPM)
6.67 x
106
5,030
16,990
27,220
30,680
2.00 x
106
4,160
16,450
25,530
29,630
212
Chapter-5
vided by the coupling vendor is accurate. However, a quick check of this critical
parameter is always warranted. The torsional stiffness of this solid spool piece
can be closely approximated by computing the stiffness of the main torque tube,
and the mounting flanges. These values are summed in a reciprocal manner to
determine torsional stiffness of the spool piece. Since this is a solid assembly,
equation (3-66) may be used to compute the spool stiffness as follows:
4
K tor
spool
4
6
2
G shear D
= ----------------------------------------- = ------------------------------------------------------------------------------------------------------------32 L
32 ( 16.80 2 0.53 Inches )
6
K tor
spool
6
992.66 10
= ----------------------------- = 1.97 10 Inch-Pound/Radian
503.68
Similarly, the torsional stiffness of each end flange is calculated as:

4
K tor
flange
4
6
2
G shear D
= ----------------------------------------- = ------------------------------------------------------------------------------------------------------------32 L
32 ( 0.53 Inches )
6
K tor
flange
9
173.19 10
= ----------------------------- = 10.21 10 Inch-Pound/Radian
16.92
These individual torsional stiffness values may now be combined in a reciprocal manner to determine the overall or effective torsional stiffness of the entire
solid spool piece in the following manner:
1
1
1
1
----------------- = ---------------------- + ------------------------ + -----------------------K tor
K tor
K tor
K tor
eff
spool
flange
flange
1
1
1
1
----------------- = ----------------------- + -------------------------- + -------------------------6
9
9
K tor
1.97 10
10.21 10
10.21 10
eff
or
K tor
eff
= 1.97 10 Inch-Pound/Radian
From these calculations, it is clear that the spool piece torsional stiffness is
governed by the center torque tube. The end flanges are very stiff, and the most
flexible member (center tube) controls the effective stiffness. This coupling torsional stiffness value is consistent with the required 2.00x106 Inch-Pound/
Radian determined from the undamped analysis. Overall, this proved to be a
mechanically acceptable field retrofit that performed with good reliability.
Stability and Damped Critical Speed Calculations
213
STABILITY AND DAMPED CRITICAL SPEED CALCULATIONS

The inclusion of damping into the analysis significantly expands the usefulness of the analytical calculations. This manifests as improved correlation
between the rotor dynamics computations, and the real machinery behavior. In
many ways, a damped analysis is similar to undamped calculations, with the
important inclusion of damping and cross-coupled stiffness from bearings, seals,
and aerodynamic and/or fluid interactions. Most programs allow the consideration of non symmetric bearing and support coefficients, plus an output scheme
that displays representative vertical and horizontal motion. This type of analysis
also determines stability, and it provides a comparison of stability between
modes. Some programs, such as ROTSTB by Gunter7 use a complex matrix
transfer method, and other programs such as DYROBES by Chen, Gunter, and
Gunter8 are based upon finite element analysis (FEA) numerical methods. It
should be noted that the matrix transfer method may skip roots that are closely
spaced, but this problem does not occur with FEA.
Damped calculations may be considered as somewhat of a pure analysis due
to the fact that these programs calculate all lateral critical speeds of the mechanical system, including all potential forward, reverse, and mixed modes. In addition, the damped analysis determines the stability characteristics of each mode.
This is quite significant since reverse modes and stability are not computed by
other methods. For a rotor mounted between bearings with small wheel diameters, this feature may not be particularly important. However, for a rotor with
large overhung wheels, the potential of various reverse modes is significant, and
this behavior must be visible to the diagnostician. On the limitation side, this
type of program does not include external forces from mass unbalance, shaft
bows, or other external excitations.
Damped critical speed programs include multiple gyroscopic effects, shear
deformation, rotary inertia, plus a full set of eight bearing and support coefficients at defined speeds. Couplings, impellers, thrust disks, shaft spacers, and
hollow shafts are accommodated. Hysteretic shaft damping may be included as
well as aerodynamic cross-coupling effects. These programs allow variable density of rotor materials, and various boundary conditions for the rotor. Forward,
backward, or mixed criticals are computed, plus the complex Eigenvalue for each
resonance. The Eigenvalues consist of real and imaginary portions, and the normal output units are Radians per Second. The real portion of the Eigenvalue is
the modal damping or growth factor. The imaginary portion of the Eigenvalue
represents the damped natural frequency in Radians/Second. Multiplication by
30/ converts this value from Radians/Second to RPM as shown in equation (5-2):
7 Edgar J. Gunter, ROTSTB, Stability Program by Complex Matrix Transfer Method - HP Version 3.3, Computer Program in Hewlett Packard Basic by Rodyn Vibration, Inc., Charlottesville,
Virginia, March, 1989, modified by Robert C. Eisenmann, Machinery Diagnostics, Inc., Minden,
Nevada, 1992.
8 W.J. Chen, E. J. Gunter, and W. E. Gunter, DYROBES, Dynamics of Rotor Bearing Systems,
Version 4.21, Computer Program in MS-DOS by Rodyn Vibration, Inc., Charlottesville, Virginia,
1995.
214
Chapter-5
30 Imag
Damped Critical Speed = ---------------------------
(5-2)
This class of analytical program also computes the Log Decrement for each
resonance to allow an evaluation of rotor stability. The log decrement is determined by multiplying -2 times the ratio of real to imaginary portions of the
Eigenvalue as follows:
2 Real
Log Decrement = = -----------------------------Imag
(5-3)
A positive log decrement identifies a stable system, whereas a negative

value signifies an unstable mode. Since the log decrement is a direct indication of
the damping and the decay rate through a resonance, it may also be used to
determine the amplification factor of the resonance. Dividing by the log decrement will result in the amplification factor Q as shown in the next expression:
Amplification = Q = ------------------------ = --Log Dec
(5-4)
On machines with split criticals, each individual resonance will be calculated, and the dominant direction will be identifiable from the mode shape plots.
For instance, consider Fig. 5-9 that describes the non-dimensional damped vertical and horizontal mode shapes (Eigenvectors) of a gas turbine rotor. Based on
the relative amplitudes of the vertical versus the horizontal mode shapes, it is
self-evident that the described resonance is predominantly a vertical mode. On
some programs, two levels of normalization are provided for each non-dimensional mode shape. Within these programs, the peak displacement for both
orthogonal directions is always 1.0, and there is no visibility of any dominant
motion in either the vertical or the horizontal directions. The data presented in
Fig. 5-9 contains only one level of normalization, and the dominant direction of
the computed motion is maintained. From this diagram, it is noted that the
Fig. 59 Damped Gas

Turbine Mode Shape
215
Eigenvalue for mode was calculated to be:

Eigenvalue = -10.34 + 235.5 Radians/Second
The real portion of this Eigenvalue is the first term of -10.34, and the imaginary part is the second term, or +235.5. Based on these values, the last three
equations may be applied to determine the natural frequency in RPM, the log
decrement for this mode, and the associated amplification factor. The damped
critical speed is determined by converting units with (5-2) as follows:
30 Imag
30 235.5
Damped Critical Speed = ---------------------------- = ------------------------- = 2, 249 RPM
This value is identical to the damped frequency on the mode shape plot Fig.
5-9. Next, consider the calculation of the log decrement with equation (5-3):
2 Real
2 ( 10.34 )
Log Decrement = ------------------------------ = ------------------------------------- = +0.276
Imag
235.5
This positive log decrement indicates a stable mode, and the value of the log
decrement may now be used to determine the amplification factor of the resonance with equation (5-4) in the following manner:
Amplification = ------------------------ = ------------- = 11.4

Log Dec
0.276
The influence of bearing clearance upon the damped critical speeds is
always a question to be addressed. By calculating bearing oil film coefficients
under various clearance conditions, and combining this information with the
support coefficients, the anticipated machinery response may be computed. For
instance, consider the data presented in Table 5-2.
Table 52 Gas Turbine Damped Critical Speeds Versus Bearing Clearance
Journal Bearing
Clearance
1st Mode
Translational
2nd Mode
Pivotal
3rd Mode
Bending
Minimum
1,146 RPM
1,877 RPM
5,732 RPM
Average
1,104 RPM
1,826 RPM
5,727 RPM
Maximum
1,073 RPM
1,760 RPM
5,719 RPM
These damped critical speed calculations were performed at average, minimum, and maximum allowable bearing clearances. These computations were
performed for a 40,000 HP gas turbine at 5,100 RPM. They reveal that the bearing dependent 1st and 2nd modes are moderately influenced, but the frequency
of the 3rd critical is insensitive to bearing clearance variations. Generally,
changes in journal bearing clearances are not a major factor in the resonant
behavior of these machines. However, for long-term operation, it is always desir-
216
Chapter-5
able to begin with the minimum bearing clearances to allow room for babbitt
attrition during the run. Another perspective of the machine characteristics may
be obtained by examining the variations in log decrement of each mode at each
bearing clearance condition. For this gas turbine, Table 5-3 summarizes these
stability parameters for the first three critical speeds.
Table 53 Gas Turbine Log Decrement Versus Bearing Clearance
Journal Bearing
Clearance
1st Mode
Translational
2nd Mode
Pivotal
3rd Mode
Bending
Minimum
0.418
0.497
0.041
Average
0.415
0.519
0.044
Maximum
0.338
0.496
0.046
If the log decrement is positive, vibration amplitudes will decay with time.
Conversely, if the log decrement carries a negative sign, then the mode is unstable, and amplitudes will increase with time. Within Table 5-3, all values are positive. That indicates stable modes within the operating speed domain of the gas
turbine. The magnitude of the log decrement describes the rate of oscillation
decay. Specifically, a large positive log decrement delineates a well damped system that will rapidly attenuate vibratory motion. A well damped resonance will
display a low amplification factor, and will persist over a broader frequency
range. On the other hand, a small log decrement identifies a poorly damped resonance, with a higher amplification factor, and a smaller bandwidth.
With respect to the log decrements presented in Table 5-3, it is clear that
the second pivotal critical speed exhibits the largest values. As such, this second
mode would be the most difficult to excite, and the resultant motion would be
quickly suppressed (i.e., damped out). The first translational mode has somewhat lower log decrement values. This critical would be slightly easier to excite,
and the resultant motion would continue for a longer time. Finally, the rotor first
bending mode (3rd critical) has the lowest log decrement. This resonance is the
easiest to excite, and the motion would decay at a slower rate. A low log decrement is indicative of a high amplification factor at the resonance. This manifests
as rapidly increasing vibration amplitudes with minimal phase change as the
skirt of the resonance is approached.
217
Case History 12: Complex Rotor Damped Analysis
Journal 4.000"
5 Pads - LOP
6 Mil Diam. Clearance
445# Static Load
Exhaust
700F
430F Discharge
450F Discharge
1,250F Inlet
This particular rotor consists of an overhung hot gas expander wheel, a pair
of midspan compressor wheels, and three stages of overhung steam turbine
wheels9 as described in Fig. 5-10. A series of axial through bolts are used to connect the expander stub shaft through the compressor wheels, and into the turbine stub shaft. This type of assembly is similar to many gas turbine rotors.
However, in this machine, the rotor must be built concurrently with the inner
casing. Specifically, the horizontally split internal bundle is assembled with the
titanium-aluminum compressor wheels, stub shafts, plus bearings and seals. The
end casings are attached, the expander wheel is bolted into position, and the turbine stages are attached with another set of through bolts.
The eight rotor segments are joined with Curvic couplings, identified as
#1 through #7 on Fig. 5-10. Even with properly ground and tight fitting Curvics, there is potential for relative movement of rotor elements. Although each
of the rotor segments are component balanced, any minor shift between elements
will produce a synchronous force. Since this unit operates at 18,500 RPM, a few
grams of unbalance, or a Mil or two of eccentricity will result in excessive shaft
vibration, and strong potential for machine damage. Furthermore, the distribution of operating temperatures noted on Fig. 5-10 reveals the complexity of the
thermal effects that must be tolerated by this unit. The 1,250F expander inlet is
followed by compressor discharge temperatures in excess of 430F. The steam
turbine operates with a 700F inlet, and a 160F exhaust.
By any definition, this must be considered as a complicated and difficult
rotor. On the positive side, this machine is a compact design that yields a high
thermal efficiency. Hence, when the unit is properly assembled, and balanced, it
is very cost-effective to operate.
Journal 4.500"
5 Pads - LOP
465# Static Load
Ambient
Air Suction
220F Suction
Turbine
Compressor
Stub Shaft
Curvic #4
1st Stage
Compressor
Curvic #3
2nd Stage
Stub Shaft
Curvic #2
Expander
Wheel
Curvic #1
Exhaust
160F
Steam Inlet
700F
Expander
Thrust
Faces
1st Stage Turbine

2nd Stage Turbine
3rd Stage Turbine
Curvic #5
Curvic #6
Curvic #7
Balance Plane #1
20 Axial Holes
Balance Plane #2
20 Radial Holes
1Y
30
Balance Plane #3
20 Radial Holes
30
60
1X
Rotation and Angular
Coordinates Viewed
from the Expander
CCW
Rotation
2Y
Balance Plane #4
30 Axial Holes
60
2X
Rotor Weight = 910 #

Rotor Length = 77.00"
Bearing Centers = 45.68"
Fig. 510 Combined Expander-Air Compressor-Steam Turbine Rotor Configuration

9 Robert
C. Eisenmann, Some realities of field balancing, Orbit, Vol.18, No. 2 (June 1997), pp 12-17.
218
Chapter-5
A double overhung rotor with an appreciable midspan mass has the potential for multiple resonances with either forward or reverse modes. In order to
better understand the behavior of this machine, various historical data sets were
reviewed. It was noted that reverse orbits appeared around 7,000, and 17,000
RPM. Field balancing activities on this machine were generally successful when
a two step correction was used. The first step consisted of an intermediate balance based on transient data as the machine passed through 14,000 RPM. This
initial balance was accomplished using the outboard planes #1 and #4. This was
followed by a trim at 18,500 RPM on the inboard planes #2 and #3 located next to
the compressor wheels. It was evident that if the rotor was not adequately balanced at 14,000 RPM, it probably would not run at 18,500 RPM.
Further examination of historical data revealed that vibration severity
changed in accordance with the machinery operational state. For instance, the
peak vibration amplitudes occur at a rotor critical that appears between 7,600
and 8,100 RPM. This resonance displays the following variable characteristics:
Cold Startup to 14,500 RPM Peak Response of 2.0 to 5.0 Mils,p-p
Warm Coastdown from 14,500 RPM Peak Response of 4.0 to 5.0 Mils,p-p
Hot Crashdown from 18,500 RPM Peak Response of 6.0 to 8.0 Mils,p-p
These amplitude variations are combined with changes in the amplification factor through the resonance (potential change in damping). Clearly, this information must be supplemented by an examination of the variable speed vibration
data plus an understanding of the rotor critical speeds, and mode shapes.
Phase Lag (Degrees)
200
Expander Bearing
Probe #1Y
250
300
350
40
400
Turbine Bearing
Probe #2Y
Displacement (Mils,p-p)
90
450
4,000
5.0
Fig. 511 Bode Plot Of

Shaft Y-Axis Proximity
Probes During A Typical
Machine Startup
6,000
8,000 10,000 12,000 14,000 16,000 18,000

Mode @
Full Speed @
7,800 RPM
18,500 RPM
4.0
Expander Bearing
Probe #1Y
3.0
Process Hold @
14,500 RPM
2.0
1.0
Turbine Bearing
Probe #2Y
0.0
4,000
6,000
8,000
10,000 12,000 14,000 16,000 18,000
Rotational Speed (Revolutions/Minute)
219
A typical startup Bode plot of the Y-Axis response from each measurement
plane is shown in Fig. 5-11. Both plots are corrected for slow roll runout at 1,000
RPM, and the resultant data is representative of the true dynamic shaft motion
at each lateral measurement plane. The major resonance appears at 7,800 RPM.
A process hold point occurs at 14,500 RPM, and the unit displays various amplitude and phase excursions at this speed. Some of this behavior is logically due to
the heating of the rotor and casing, plus variations in settle out of the operating
system (i.e., pressures, temperatures, flow rates, and molecular weights).
The Bode also exhibits additional vector changes between 14,500 and
18,500 RPM. Some of these changes are due to the influence of a backward mode
around 17,000 RPM. Other changes appear as the machine approaches the normal operating speed of 18,500 RPM. This higher speed data is difficult to fully
comprehend in the Bode plot, but it becomes more definitive when replotted in
the polar format of Fig. 5-12.
asing Speed and Angle
Incre
s
Probes
#1Y & 2Y
30
60
CCW
Ro
tat
ion
Turbine Bearing
Probe #2Y
330
Exp
a
90
nde
300
Turbine
0.0
1.0
2.0
3.0
4.0
120
5.0
Mode @
7,800 RPM
Expander Bearing
Probe #1Y
270
Fig. 512 Polar Plot Of

Shaft Y-Axis Proximity
Probes During A Typical
Machine Startup
240
150
180
210
The point of major interest on Fig. 5-12 is that at full speed, the turbine end
shaft is moving towards the 9 oclock direction, and the shaft at the expander is
heading towards 4 oclock. This behavior indicates a couple, and the presence of
some type of pivotal mode occurring at a frequency above the normal running
speed of 18,500 RPM. In many cases, this type of response would not be unusual.
However, for this unit, the machinery files had no indication of a resonance
around the operating speed. Due to the measured response of midspan balance
weights (planes #2 & #3) at 18,500 RPM, it was clear that the vibration data was
correct. This also implies that the historical undamped mode shapes were not
fully representative of actual machinery behavior.
As previously noted, there are only two lateral vibration measurement
220
Chapter-5
planes along the entire length of this rotor. Since there were no other feasible
locations for shaft probes, additional measurement options were eliminated.
Hence, the only viable approach resided with a proper analytical model of this
rotor system. A 65 station damped model was constructed. This computer model
included bearing stiffness and damping that varied with speed, plus flexible
bearing supports. Damped natural frequencies, direction of each mode, and the
log decrement for each mode are summarized in Table 5-4.
Table 54 Summary Of Calculated Damped Natural Frequencies
Damped
Frequency
Mode
Direction
Log
Decrement
Stiff Shaft - Pivotal
5,040 RPM
Backward
1.61
Stiff Shaft - Pivotal
5,470 RPM
Forward
1.39
Stiff Shaft - Translational
5,910 RPM
Forward
1.92
Stiff Shaft - Translational
6,310 RPM
Forward
1.47
Shaft Bending - 2 Nodes
7,080 RPM
Backward
0.166
7,840 RPM
Forward
0.0753
17,710 RPM
Backward
0.263
21,930 RPM
Forward
0.263
Mode Description
The first four modes are stiff shaft pivotal and translation shapes with high
log decrements. These modes did not appear in the vibration data due to the high
damping for each mode. A backward mode was detected at 7,080 RPM. This
mode was not visible in the startup plots, but it briefly appears in some of the hot
coastdown data. The most active forward mode within the operating speed range
occurs at a damped frequency of 7,840 RPM, and the calculated mode shapes for
this resonance are presented in Fig. 5-13. From this diagram, it is noted that the
Fig. 513 Calculated

Damped Mode Shapes Of
Main Rotor Resonance At
Nominally 7,800 RPM
221
normalized deflections at both bearings are quite small. This indicates minimal
motion of the journals within their respective bearings. With small relative
motion, the velocity is low, and bearing damping is minimal. This behavior is
reflected in the low 0.0753 log decrement for this mode.
The validity of the analytical model is supported by correlation of the computed resonant frequency of 7,840 RPM (from Fig. 5-13), with the measured resonance of 7,800 RPM (Fig. 5-11). It is also clear from Fig. 5-13, that the rotor
balance response at this resonance can be effectively controlled by corrections at
the modally effective end planes #1 and #4.
With increasing speed, the damped analytical model reveals another backward mode at 17,710 RPM. The vertical and horizontal shapes for this reverse
mode are presented in Fig. 5-14. This pivotal mode is often visible as reversed
Fig. 514 Calculated

Damped Mode Shapes Of
Backward Resonance At
Nominally 17,710 RPM
orbits on the transient vibration data. Immediately above the normal operating
speed of 18,500 RPM, a damped mode was calculated at a frequency of 21,940
RPM as in Fig. 5-15. This forward mode has the same deflection characteristics
as the previously backward mode at 17,710 RPM. In both cases, the inboard balance planes #2 and #3 are the most modally effective correction planes for this
speed domain. It is concluded that weight corrections adjacent to these compressor wheels (planes #2 and #3) should be out of phase. This is due to the fact that
a nodal point exists at the middle of the rotor. The validity of this conclusion was
field tested on the machine. The installation of a pair of weights at the middle
planes at the same angle resulted in excessive vibration. However, a couple shot
proved to be smooth, and supportive of the analytical mode shape at high speed.
Finally, the existence of a pivotal resonance at slightly above running speed
was previously noted on the polar plot, Fig. 5-12. The damped mode shape presented in Fig. 5-14 supports this observation. Once again, the vibration measurements, and the analytical tools are combined to explain the behavior of a
complex machine. Although the variable behavior through the main critical
speed at 7,800 RPM is still not totally clear, it is speculated that a loosening or
relaxation of the segmented rotor occurs with elevated temperatures.
222
Chapter-5
Fig. 515 Calculated

Damped Mode Shapes
Of Rotor Resonance
Occurring Above The
Normal Operating Speed
This work provided an improved understanding of the shaft response and

damped mode shapes, plus a better appreciation of the process influence. Armed
with this information, the rotor was balanced at the intermediate speed by using
transient data acquired during cold startups at 14,000 RPM. This intermediate
step was a two plane balance with weight corrections at the outboard planes #1
and #4. This allowed the rotor to run at full speed of 18,500 RPM, and a final two
plane trim balance was performed on the interior planes #2 and #3 after a full
heat soak. The synchronous shaft vibration amplitudes were significantly
reduced. The magnitude of the final running speed vibration vectors ranged from
5 to 8% of the diametrical bearing clearance. The suitability of this balance state
is demonstrated by an extended process run on this machine. Additional details
on the unbalance response of this machine are presented in case history 36.
FORCED RESPONSE CALCULATIONS

The programs used to calculate undamped critical speeds, stability, plus
damped critical speeds, all display the final results as non-dimensional amplitudes. This is adequate for determining mode shape geometry, and identifying
the stations of maximum deflection. However, these basic concepts must be significantly extended to duplicate actual rotor behavior.
This desirable simulation of rotor motion is addressed by a forced response
analysis of a damped rotor system. Various forcing functions such as mass unbalance, skewed disks, or shaft bows are allowed in this type of analysis. In addition
to previous rotor modeling capabilities, the configuration for a forced response
analysis typically includes rigid plus flexible disks. This type of program accepts
constant coefficients for bearings and supports, or coefficients that vary as a
function of speed. Usually, the most accurate results are obtained by employing
fifth or sixth degree polynomial functions that describe the stiffness and damping coefficients for the bearing oil film, and the housing as a function of rotational speed. The program computes Bode and polar plots, elliptical orbits, two
Forced Response Calculations
223
and three-dimensional mode shapes, plus bearing forces. Some programs, such
as UNBAL by Gunter10 use a Complex Matrix Transfer method, and other programs such as DYROBES by Chen, Gunter, and Gunter11 are based upon Finite
Element Analysis (FEA) numerical methods.
Since support coefficients may be calculated as a function of rotational
speed, shaft displacement response vectors may be computed with minimal discontinuity. One of the obvious applications for this information would be the
development of synchronous 1X vectors in Bode plots as shown in Fig. 5-16. The
Fig. 516 Calculated Gas

Turbine Bode Plot
bottom half of the Bode plot displays 1X vibration amplitudes in Mils,p-p as a

function of rotative speed in RPM. The data in the top half of Fig. 5-16 depicts
the phase lag in Degrees. The same characteristics are shown in the polar plot of
1X vectors in Fig. 5-17. In both cases, the data is observed from a true vertical or
horizontal perspective. The angular starting point for the phase angles begins at
the probe location, and the convention follows standard phase lag logic with the
angles progressing against rotation.
These analytical data presentation are designed to be analogous to the
Bode and polar plots measured by proximity probe systems. The same 1X synchronous vectorial data is presented on both types of plots. The Bode plot displays excellent visibility of amplitude and phase changes with respect to speed,
and the polar plot enhances the variations with respect to phase. In addition, the
computer solution is not limited to the physical restrictions imposed on the physical installation of the proximity probes. In fact, the analytical model allows the
generation of Bode and polar plots at any rotor station over any speed domain.
Since the entire rotor motion has been computed at numerous speeds, it is
possible to construct both two and three-dimensional mode shapes of the scaled
10 Edgar J. Gunter, UNBAL, Unbalance Response of A Flexible Rotor - HP Version 4, Computer Program in Hewlett Packard Basic by Rodyn Vibration, Inc., Charlottesville, Virginia, July,
1988, modified by Robert C. Eisenmann, Machinery Diagnostics, Inc., Minden, Nevada,1992.
11 W.J. Chen, E. J. Gunter, and W. E. Gunter, DYROBES, Dynamics of Rotor Bearing Systems,
Version 4.21, Computer Program in MS-DOS by Rodyn Vibration, Inc., Charlottesville, VA, 1995.
224
Chapter-5

Turbine Polar Plot
rotor behavior. For example, Fig. 5-18 depicts a two-dimensional rotor mode
shape superimposed upon an outline of the gas turbine rotor. The solid lines represent the predicted vertical vibration, and the dotted lines depict the horizontal
shaft motion. In all cases, scaling is provided via the left hand axis. This type of
scaled mode shape allows the comparison of anticipated displacement amplitudes with the actual machine clearances. On some machines, this type of information may not be particularly useful. However, on industrial turbines with
close tip clearances on the axial blades, this type of displacement data along the
rotor may be extraordinarily important.
The calculated shaft mode shape may also be viewed as a three-dimensional plot as shown in Fig. 5-19. In many cases this type of display is visually
more informative than the two-dimensional plot. This three-dimensional plot is
Fig. 518 Calculated TwoDimensional Plot for Horizontal And Vertical Gas Turbine Rotor Mode Shapes
225
Fig. 519 Calculated

Three-Dimensional Gas
Turbine Mode Shape At
Normal Operating Speed
composed of shaft orbits at various locations, and it is scaled from a maximum

vector that is listed on each diagram. The bearing locations are identified, and
the speed is listed. This presentation provides an additional level of visibility to
the calculated rotor mode shapes. The entire display may be rotated to observe or
enhance the characteristics of a particular plot. The determination of calculated
amplitude at specific rotor stations is easier with the two-dimensional plot of Fig.
5-18. However, the three-dimensional diagram of Fig. 5-19 does provide a better
physical rendition of the shaft modal behavior.
The shaft orbits at any location and any rotative speed may also be
extracted from the calculations and presented separately. Fig. 5-20 depicts a typical pair of calculated orbits from opposite ends of the turbine. The orbits are

Turbine Shaft Orbits At
Proximity Probe Measurement Stations
226
Chapter-5
viewed from the inlet end of the turbine, and the calculated orbits are oriented to
be consistent with the measured shaft vibration data. The specific data used for
Fig. 5-20 is representative of the anticipated shaft vibration at the actual probe
locations. By changing the forcing function, such as various levels of unbalance
at different locations within the turbine, the affect upon the overall mode shape,
and resultant loads plus shaft vibration at the journal bearings may be calculated. Similarly, the impact of skewed wheels or bent rotors may be examined on
paper before the machine is ever built.
Overall, it is evident that the computation of anticipated vibratory behavior
along the length of a rotor provides useful information regarding the behavior of
the machinery. If it can be demonstrated that this computational information is
correct, and consistent with shaft vibration measurements, then a significant
tool is available for the machinery designer as well as the diagnostician.
Case History 13: Gas Turbine Response Correlation
At the conclusion of a set of analytical calculations, the issue of verification
of the results must be addressed. This is not an easy topic since comparison of
analytical computations with field vibration measurements is seldom performed.
As such, specific items of comparison are rarely defined, and tradition evaluations are often filled with generalities. Within the context of this chapter, it
seems appropriate to perform a comparison on the basis of both qualitative and
quantitative parameters. Specifically, synchronous characteristics of a single
shaft gas turbine will be reviewed, and definable items during an unbalance
response test will be correlated.
The 40,000 horsepower machine under discussion contains a 22,500 pound
rotor that normally operates between 5,000 and 5,300 RPM. The unit is equipped
with elliptical journal bearings, and a double acting Kingsbury type thrust bearing. This rotor contains seventeen stages in the axial flow air compressor, and
two turbine stages as depicted in Fig. 5-21.
Fig. 521 Gas Turbine

Rotor Configuration
In nearly all situations, the measured shaft vibration is elliptical, with the
horizontal motion exceeding the vertical. In many cases, the orbit is tilted in the
direction of rotation. At normal speeds, the phase angles between inboard and
outboard orbits are almost identical. The same general behavior is noted in the
analytical computations. For example, the orbit plots presented in Fig. 5-20 were
extracted from a model of this machine. Hence, the general motion described by
227
the analytical model reflects the measured field vibratory characteristics.

Transient speed characteristics are often difficult to understand due to the
potential variation, and distribution of residual unbalance across the length of
the rotor. For example, a mid span weight will drive the first critical, but will
have minimal influence upon the pivotal mode. By the same token, a couple
unbalance may be sufficient to excite the pivotal second critical, but the first
translational mode may experience minimal excitation. Thus, the modal weight
distribution will influence the critical speeds stimulated, and the amount of measured excitation. Although it is difficult to duplicate amplitude response through
a series of resonant frequencies, it is reasonable to compare the measured critical speeds with the calculated natural frequencies. For instance, Table 5-5 compares the results for the first three critical speeds of this turbine rotor. The top
row of calculated values displays the damped critical speeds with flexible supports. This analysis identifies a split horizontal (H) and vertical (V) critical for
the first and second modes. The second row of calculated critical speeds were
extracted from a forced response analysis that includes an improved definition of
bearing and support characteristics. Finally, the bottom row summarizes the
range of measured criticals for several different machines during multiple startup and coastdown field vibration data sets.
Table 55 Comparison Of Calculated Versus Measured Critical Speeds
1st Mode
Translational
2nd Mode
Pivotal
3rd Mode
Bending
(RPM)
(RPM)
(RPM)
Calculated Damped
1,100 (H)
1,400 (V)
1,830 (H)
2,250 (V)
5,730
Calculated Forced
1,100 (H)
1,300 (V)
1,800 (H)
2,360 (V)
5,600
Field Measured
1,000 to
1,450
1,900 to
2,300
5,600+
Origin of
Critical Speed
Note that the first critical displays excellent agreement between the
damped natural frequencies, the forced response criticals, and the measured critical speeds. Similarly, the pivotal second mode also shows excellent agreement
between the calculated and measured resonant frequencies. The bending third
critical is visible in the calculations, but is somewhat elusive in the field measurements. Since this mode is above the normal operating speed range, it can
only be reached during over speed runs. These runs are usually of rather short
duration, and the resonance generally has minimal time to respond. Overall, this
agreement between the calculated and measured critical speeds provides
increased confidence in the validity of the computations.
Another check on the accuracy of the model may be performed by installing
an easily definable excitation on a real machine, and adding the same excitation
to the model. A direct comparison of measured versus calculated vibration
228
Chapter-5
response characteristics should provide a suitable test of the model. For this test,
consider the addition of unbalance calibration weights to each end of the turbine.
These weights would alter the 1X synchronous response, and the results
should be visible in the vibration measurements, plus the analytical computations. For the purposes of this response test, an unbalance calibration weight of
77 Gram-Inches at 230 was added to the inlet coupling. The centrifugal force
from this weight at 5,100 RPM was 125 Pounds (0.6% of the rotor weight). At the
exhaust coupling, an unbalance of 234 Gram-Inches was attached at 275. This
weight produced a centrifugal force of 381 Pounds at 5,100 RPM (1.7% of rotor
weight). Since the rotor residual unbalance was low, the vertical shaft vibration
amplitudes were also small. Hence, the most meaningful data was extracted
from the horizontal proximity probes.
Sequentially, an initial data set was obtained at 5,100 RPM without any
extra unbalance. Next the machine was shutdown, the 77 Gram-Inch weight was
installed at the inlet, and a second data set acquired. The turbine was again
shutdown, and the inlet weight was removed. Next, the 234 Gram-Inch exhaust
end weight was added, and a final data set was acquired at 5,100 RPM. The 1X
vectors from the horizontal probes were runout compensated, and the results are
summarized in Table 5-6.
Table 56 Measured X-Axis Vibration Response Vectors With Unbalance Weights
Weight Condition
Inlet Bearing #1
Exhaust Bearing #2
No Weight Installed
0.85 Mils,p-p @ 32
1.01 Mils,p-p @ 346
Weight at Inlet End
1.04 Mils,p-p @ 15
1.38 Mils,p-p @ 330
Weight at Exhaust End
1.06 Mils,p-p @ 350
1.65 Mils,p-p @ 337
It is obvious that the small weight installed at the inlet end of the turbine
produced only minor changes, whereas the exhaust end weight resulted in a significantly larger change in shaft vibration. Since the initial synchronous 1X vectors are quite small, a comparable analytical case was developed with minimal
shaft bow, and low residual unbalance. Specifically, a midspan shaft sag of 0.2
Mils (0.4 Mils TIR), was combined with a residual unbalance at the first compressor stage of 100 Gram-Inches. Another 100 Gram-Inch residual was located
at the second stage turbine wheel. Calculations were performed at 5,100 RPM
Table 57 Calculated Horizontal Vibration Response Vectors With Unbalance Weights
Weight Condition
Inlet Bearing #1
Exhaust Bearing #2
No Weight Installed
0.93 Mils,p-p @ 281
1.09 Mils,p-p @ 290
Weight at Inlet End
1.10 Mils,p-p @ 260
1.26 Mils,p-p @ 261
Weight at Exhaust End
1.56 Mils,p-p @ 265
1.67 Mils,p-p @ 274
229
with the initial shaft bow, and the two residual unbalance locations. Two additional cases were run with the previously identified coupling unbalance weights.
The computed 1X vibration vectors from these runs are presented in Table 5-7.
The initial rotor bow and residual unbalance vectors were selected to match
the initial measured shaft vibration vector magnitudes. Since two sets of unbalance weights were used, both the direct and the cross-coupled balance response
vectors may be calculated. The specific equations for these calculations are listed
in chapter 11 of this text. For a two plane correction, equations (11-13), through
(11-16) may be used. For example, the measured shaft vibration data at the turbine exhaust bearing may be used to calculate the balance sensitivity vectors
from equation (11-16) as follows:
S 22
W2
234 Gram Inches 275
= ---------------------- = ----------------------------------------------------------------------------------------------------
B 22 A 2
1.65 Mils p p 337 1.01 Mils p p 346
234 Gram Inches 275

S 22 = ------------------------------------------------------- = 349 Gram-Inches/Mil 312
0.67 Mils p p 323
If the same calculations are performed for each set of primary and cross
coefficients, the measured versus calculated balance sensitivity vectors may be
generated as shown in Table 5-8. Note that the calculated sensitivity angles were
adjusted by 45 to correct for the true horizontal orientation of the analytical calculations versus the +45 location of the proximity probe. Thus, the tabulated
vectors in Table 5-8 are directly comparable in terms of angular position.
The similarity between measured and calculated sensitivity vectors in
Table 5-8 lends further credibility to the validity of the analytical calculations.
Certainly the analytically derived balance sensitivity vectors are not of sufficient
accuracy to perform a refined field trim balance. However, they exhibit magnitudes that reflect the vibration response measurements, with reasonably consistent vector angles. Again, it is concluded that the analytical model does an
excellent job of simulating the field dynamic behavior of the gas turbine rotor.
Table 58 Comparison Of Measured Versus Calculated Balance Sensitivity Vectors
Vector Identification
Measured Sensitivity
Calculated Sensitivity
Inlet Bearing - S11
229 Gram-Inches/Mil,p-p @ 263
Inlet Bearing - S12
Exhaust Bearing - S21
Exhaust Bearing - S22
230
Chapter-5
Case History 14: Charge Gas Compressor with Internal Fouling

The centrifugal compressor depicted in Fig. 5-22 operates in cracked gas
service. A low stage double flow compressor is coupled to the discharge end of
this machine, and a high stage compressor is coupled at the thrust end of the
subject compressor. This train is steam turbine driven at a maximum operating
speed of 5,400 RPM. As noted, the rotor weighs 3,520 pounds, and it has a span
of 107 inches between bearing centers. This compressor contains six impellers,
and they are equally divided between the 2nd and 3rd process stages.
2nd Stage
3rd Stage
Suct.
Disch. Suct.
Disch.
Axials
Compressor
2nd & 3rd Stages
Vert.
Vert.
15
20
Rotor Weight = 3,520#

Bearing Span=107"
6 Impellers
75
90
Fig. 522 Charge Gas

Case Configuration
Horiz.
Low Horizontal &

Vertical Vibration
Horiz.
High Horizontal
Vibration
The compressor had been operating smoothly for an extended period of time
when the horizontal vibration at the discharge end began to increase. The trend
plot in Fig. 5-23 documents the vibration change over a four month period. At the
beginning of this data, the machine displayed low and acceptable vibration
amplitudes from all radial probes. A power outage in February resulted in an
increase of vibration amplitudes at the discharge end. Approximately one week
later, a problem with a seal pot float mechanism occurred, and vibration levels
increased again. The amplitudes remained fairly constant throughout March,
and then began a gradual downward trend towards the end of May.
It should be mentioned that these data points were acquired manually with
a portable data collector on a weekly route. Changes or variations between these
periodic samples are not visible. Hence, the transition between the low vibration
condition on or about May 28, and the 6.0 Mil,p-p value displayed on June 5 was
unknown. Furthermore, the constituent parameters of rotational speed vectors,
and radial position data was not available. The high vibration amplitude of 6.0
231
Fig. 523 Charge Gas

Compressor Vibration
Amplitude - Four Month
Trend Plot
Mils,p-p on June 5th continued to increase until the horizontal probe exhibited an
unfiltered amplitude of 6.9 Mils,p-p. This behavior was documented in the orbit
and time base plots in Fig. 5-24. Simultaneously, the suction end displayed low
vibration amplitudes (1.3 Mils,p-p), and this data is shown in Fig. 5-25. It was
clear that the discharge journal was moving horizontally across the entire bearing clearance. That is, the 7 Mil vibration, plus 1 or 2 Mils for the oil film thickness is equivalent to the total diametrical bearing clearance of nominally 9 Mils.
The 1X vibration was reduced by unloading the compressor to allow operation at a lower speed. A further drop in vibration was achieved by reducing the
oil supply temperature 7F to increase the damping. This temperature reduction
was accomplished by adding cold firewater to the water side of the oil cooler.
Since this is a dirty gas service, the issue of coke buildup should always be considered. In this case, it was understood that a drop in efficiency had occurred
during the past few months, but the specific decrease was not quantified. In retrospect, the plant personnel performed machinery efficiency calculations based
on a heat and material balance. This was a poor method to determine compressor
efficiency, and it turned out to be extraordinarily inaccurate. The only realistic
approach to determine process machinery efficiency is to begin with an accurate
measurement of the input shaft torque as discussed in chapter 6 of this text.
Wash oil rates were increased, with no measurable improvement. Based
upon the available evidence, it was initially concluded that the discharge bearing
was damaged. In addition, the suction end historical data was inconsistent. Specifically, the orbits in Figs. 5-24 and 5-25 describe a pivotal behavior across the
compressor. The suction end phase had changed several times, and motion of this
compressor was considered to be abnormal. Finally, it was agreed to shutdown
the machinery, and prepare for a rotor, bearing, and seal change.
Following an orderly shutdown, the subsequent disassembly and inspection
of the compressor resulted in several surprises. First, the discharge journal bearing was not damaged. In fact, the disassembly clearances were similar to the previous installation clearances. Second, the suction end bearing displayed babbitt
232
Chapter-5
Fig. 524 Compressor Discharge End

Bearing Shaft Radial Vibration
Fig. 525 Compressor Suction End Bearing Shaft Radial Vibration
damage on the bottom pads. Third, the compressor had a major accumulation of
coke on the stationary, and the rotating elements. Inlet guide vanes, diaphragms,
and return bends all exhibited various levels of coke deposits. In addition, the
last stage wheel revealed major coke clusters at random locations within the
impeller.
The compressor rotor contains three impellers for each process stage (2nd
and 3rd). In both sections of the compressor, the inlet wheel for the respective
stage was reasonably clean, and coke buildup increased progressively on the
next two wheels. This is typical for a cracked gas machine to display increasing
coke deposits as the heat of compression increases across the wheels that form
the particular process stage. However, the amount of buildup on the last wheel in
each process stage was substantial. Further examination of the casing revealed
that most of the interstage labyrinths, and the balance piston labyrinths were
completely filled with coke. At six locations, the mating surfaces on the rotor
were highly polished, and the evidence of close contact between the rotor and the
filled-in labyrinths was clear and unmistakable.
The condition of three of these surfaces is documented in Fig. 5-26. This
photograph of the third process Stage shows the relatively clean inlet wheel on
the left, and the heavily coked discharge end wheel on the right side. The polished shaft surfaces on the rotor are coincident with the coke filled interstage
labyrinths. On the back side of the last impeller, the rotor balance piston resides.
Although photographic evidence of this element is not as clearly defined, the balance piston displayed most of the same characteristics as the coke filled interstage labyrinths. The physical interpretation of this unique mechanical condition
was hypothesized as a machine that was operating with a series of internal bearings. Specifically, the two external oil film tilt pad bearings were supplemented
by six internal dry bearings. Five of these internal bearings were associated with
233
Fig. 526 Compressor

Third Process Stage With
Internal Coke Deposits
Producing Midspan
Pseudo Bearings
interstage labyrinths, and the sixth was at the discharge end balance piston.
The hypothesis of the development of six new internal bearings was examined in greater detail to determine if this could be responsible for the compressor
high vibration problems. The only viable method to approach this problem would
be with an analytical simulation of the machinery. The arrangement of shaft,
impellers, spacers, and couplings for a normal rotor is depicted in Fig. 5-27. This
machinery sketch identifies the proximity probe locations, and the radial journal
bearings at each end of the rotor. Stiffness and damping coefficients for the oil
film portion of these tilt pad bearings were computed. At a speed of 5,300 RPM,
the calculated horizontal oil film stiffness Kxx was 350,000 Pounds/Inch. The vertical stiffness Kyy was computed to be 2,050,000 Pounds/Inch. The calculated horizontal oil film damping Cxx was 1,100 Pounds-Seconds/Inch. Finally, the vertical
damping Cyy was 3,000 Pounds-Seconds/Inch. Since these are tilt pad bearings,
cross-coupling coefficients do not exist. The journal and thrust bearing housing
weight was approximately 200 Pounds, and the horizontal and vertical support
stiffness (Ksxx & Ksyy) for this housing were estimated at 2,000,000 Pounds/Inch.
Housing damping was calculated at 10% of the critical damping to be 200
Pounds-Seconds/Inch for the suction end bearing housing (Csxx & Csyy) as per
equation (4-17). The discharge end housing contains only a journal bearing, and
the weight of this housing was estimated at 100 pounds. The vertical stiffness of
this housing was set at 2,000,000 Pounds/Inch, and the horizontal stiffness was
slightly reduced to 1,500,000 Pounds/Inch. The estimated vertical and horizontal
damping values were proportionally reduced in accordance with the changes in
stiffness and housing weight.
The normal model included a residual unbalance of 30 Gram-Inches at 175
on the suction end, and another 30 Gram-Inches at 145 on the discharge end.
This total residual unbalance was set to be somewhat less than the normal bal-
234
Chapter-5
Fig. 527 Normal Compressor Rotor Configuration With Normal Bearings

And Typical Residual
Unbalance Levels
ance tolerance (113.4W/N) of 74 Gram-Inches for the entire rotor. This initial
model allowed examination of the normal synchronous vibration response
between 500 and 5,500 RPM. The computed response at operating speed provided an acceptable duplication of normal machine behavior. In addition, the
transient calculations accurately predicted the first critical speed region centered at 2,500 RPM. Thus, the initial model (Fig. 5-27) successfully duplicated
the historical machinery behavior. It was now reasonable to extend this model to
the abnormal condition of a heavily coked compressor as shown in Fig. 5-26.
The rotor removed from the compressor was check balanced, and the residual unbalance determined at each end of the rotor. At the suction, the residual
was 488 Gram-Inches at 218. A much higher unbalance was discovered at the
discharge end of the rotor with a measured 2,074 Gram-Inches at 201. This synchronous excitation data was loaded into the model in conjunction with a 0.25
Mil midspan rotor sag. The support condition for the abnormal case required a
minor modification of the existing bearings, plus the addition of the new internal
bearings at the filled laby locations. The previous bearing housing characteristics
were held constant. Similarly, the tilt pad bearing oil film coefficients at the discharge end were retained without modification. However, the suction end journal
bearing coefficients were modified to reflect the demonstrated higher loads at
this location. Horizontal stiffness Kxx at this location was increased to 1,500,000
Pounds/Inch, and the vertical stiffness was held at 2,050,000 Pounds/Inch.
Finally, the oil damping at the suction bearing was held constant.
Internal compressor bearings were placed at each of the locations where the
labyrinths were filled with coke, and there was obvious physical evidence of close
clearance contact between the shaft and the laby areas. These internal bearings
are identified as Brg. #2 through Brg. #7. The normal rotor journal bearings are
shown as Brg. #1, and Brg. #8 on this model. The photograph in Fig. 5-26 shows
the three internal bearings associated with the 3rd process stage as Brg. #5, Brg.
#6, plus Brg. #7 at the balance piston. These internal bearing locations are identical to the locations on the model diagram presented in Fig. 5-28.
Computation of internal bearing coefficients was difficult due to the various
unknowns associated with the internal behavior of this unit. Using short bearing
235
Fig. 528 Abnormal Compressor Rotor Configuration With Pseudo Internal

Bearings And Measured
Residual Unbalance
theory, calculations were performed at various clearances, and hydrocarbon gas

viscosities. The resulting minimum stiffness values varied between 85,000 and
325,000 Pounds/Inch. On the high side of the potential stiffness envelope, values
of 5,000,000 to 18,000,000 Pounds/Inch were computed. These maximum support
stiffness would be reduced by the actual structural rigidity of the casing itself.
Hence, the effective internal bearing stiffness would probably fall within the
range of 400,000 to 1,500,000 Pounds/Inch.
In the final assessment, it was clear that a direct computation of support
stiffness would not be attainable. A compromise value of 600,000 Pounds/Inch
was selected for the vertical and horizontal support coefficients at internal bearings #2 through #6. The balance piston displayed less contact than the shaft
labys, and a stiffness of 400,000 Pounds/Inch was used for this location. Crosscoupling coefficients, and all damping coefficients for these internal bearings
were set to zero. Although this represents a simplistic model, the available
mechanical data allows no other realistic alternative.
The forced synchronous response calculations were repeated for this abnormal case of multiple internal bearings, plus high unbalance. The results of these
calculations at 5,300 RPM are presented in Fig. 5-29. The computed shaft vibration is compared with the measured 1X shaft orbits extracted from the previously discussed Figs. 5-24 and 5-25. In both sets of orbits, the same scaling of 2.0
Mils/Division was used, and both orbital sets display a true vertical/horizontal
orientation.
Note that both discharge orbits in Fig. 5-29 are elliptical, and primarily
horizontal. Also note that the horizontal magnitudes are similar, and the
Keyphasor dots for the discharge orbits reside in the same quadrant. Vertical
magnitudes between the measured and computed discharge orbits show some
variation; and the suction end orbits are rotated approximately 90 between the
calculated and measured orbits. Nevertheless, the correlation between the measured and computed vibration response is considered to be acceptable.
Based on this ability to analytically duplicate the field machine behavior, it
is reasonable to conclude that the compressor experienced the physical changes
that were imposed upon the analytical model. Specifically, the high vibration lev-
236
Chapter-5
Fig. 529 Comparison Of Calculated Versus Measured Compressor Shaft Orbits
els at the discharge plus the low suction end vibration amplitudes were attributable to the combined effect of an internal coke buildup on the stationary
internals (manifesting as internal bearings), plus large unbalance due to coke
accumulation on the rotor. This combination of abnormalities resulted in a
heavily loaded suction end bearing with low vibration (and pad damage), combined with a generally unloaded (and undamaged) discharge end journal that
migrated across the available bearing clearance.
Once more, an analytical approach provides an acceptable simulation of a
mechanical abnormality on a centrifugal machine. In this case, the physical evidence was used to develop a model that explained the abnormal behavior
detected by the shaft sensing proximity probes. In all cases, it should be recognized that measurement and calculation technologies are coexistent resources
that can provide significantly improved understanding of mechanical behavior.
Case History 15: Hybrid Approach To A Vertical Mixer
As demonstrated in the last two case histories, analytical solutions may be
effectively combined with field vibration measurements to examine the machinery behavior from two different perspectives. This combination of techniques provides confidence in the individual technologies, plus the accuracy of the final
results. It is clear that a comparison of calculated versus computed lateral vibration behavior makes good engineering sense. However, some physical situations
cannot be properly examined by exclusively using only one technique. It these
situations, it is necessary to combine the computational techniques with the
physical measurements to arrive at a solution. This type of hybrid approach is
not a common practice, but it does provide a way to get the job done with acceptable technical credibility.
As an example of this type of problem, consider the vertical mixer rotor displayed in Fig. 5-30. Charles Jackson would probably classify this assembly as the
proverbial mud ball on a willow stick. The long and slender shaft is supported by
two bearings at the top end, and two mixer wheels are located at the bottom of
the rotor. A 30 inch elevation difference exists between the upper and lower mix-
237
Fig. 530 Physical Configuration Of Vertical Mixer Rotor
ing blades. The distance between bearings is approximately 18 inches, and the
vertical length of unsupported shaft approaches 116 inches. This rotor is driven
by a variable speed motor via a belt and pulley configuration at the top of the
assembly. In operation, the mixer is used in a batch process where the rotor is
totally immersed in the process fluid, and pulley rotational speed is normally
between 900 and 1,200 RPM. The radial bearings are rolling element units, and
a mechanical seal is used to contain the process fluids.
The dual mixer blades have an outer diameter of 20 inches, and an average
thickness of 0.188 inches. Various perforations and raised lips are fabricated into
the blades to provide the necessary agitation action. This blade design was
empirically based, and proven successful over many years of operation. However,
due to process revisions, it would be necessary to install a thicker pair of mixer
blades for future mixtures. The maximum anticipated thickness for the new
blades was 0.488 inches. This blade thickness increase could add an additional
54 pounds to the rotor assembly. Since the initial rotor weight was 615 pounds,
the additional blade weight represented a nominal 9% increase in the assembly
weight. In addition, this extra blade weight represented an appreciable increase
in the overhung mass.
During startup of this mixer with thin blades, it was observed that a critical
speed existed between 250 and 300 RPM. Since this frequency was considerably
below the operating speed range of 900 to 1,200 RPM, there was no interference
between the resonance and normal running speed excitation. However, there was
concern that the heavier mixer blades might have a detrimental influence upon
the rotor critical speeds (especially the higher order modes). There was no information regarding rotor natural frequencies in the machinery files, and there was
limited opportunity for traditional vibration response testing. As displayed in
Fig. 5-30, the entire rotor is suspended from the two top bearings. During operation of the mixer, the only possible vibration measurements must be made from
the exterior of the bearing housing. Obviously, this type of rotor will exhibit a
variety of cantilevered modes, and vibratory motion at the bearings will be minimal under most conditions. Thus, direct casing vibration measurements will not
be beneficial in solving this problem.
The undamped natural frequencies of the mixer rotor could be computed as
discussed earlier in this chapter. Unfortunately, internal shaft diameters were
238
Chapter-5
not known, and the shaft material properties were reasonably undefined. Hence,
a direct calculation of the critical speeds could not be attempted due to a lack of
the fundamental mechanical information on the rotor.
The time honored bump test technique of hit the stationary rotor with a 4x4
timber and measure the vibration response could be used, but this approach
leaves much to be desired. Although one or more natural frequencies would be
excited, there is minimal ability to determine accurate mode shapes for each resonance, and virtually no way to separate out closely spaced or coupled modes.
From many aspects, a realistic engineering solution to this problem might
seem to be unattainable. However, if the question is approached with multiple
tools instead of a single technique, a logical hybrid approach may be developed.
In this particular case, the initial step consisted of accurately measuring the
static mode shape of the non-rotating shaft using an HP-35670A Dynamic Signal
Analyzer plus an accelerometer, and a modally tuned impact hammer. The accelerometer was mounted close to the bottom mixer blade. The force hammer was
used to impact the shaft at twelve different elevations at 10 inch increments up
the length of the shaft. Frequency response functions (FRF) were then acquired
between the accelerometer and each hammer location (acceleration / force). The
data was checked for proper phase shifts, plus acceptable coherence as discussed
in chapters 4 and 6. At this point, the FRF vectors at the various resonances
could then be extracted and used to construct representative mode shapes.
Performing the above tasks manually can be a time consuming process.
Handling a dozen FRFs is not impossible, but it is clear that a complex threedimensional model may prove to be quite challenging. Hence, it is appropriate to
consider methods of automating the field test, plus the associated calculations
and animation of the resultant mode shapes. Historically, this type of work has
been performed with large instrumentation systems operating under computer
control. These types of measurement and data processing systems are complicated to set up and operate. In many cases, the field environment will not tolerate the time or expense associated with large scale modal tests.
A much more attractive approach resides in operating the DSA with software that is dedicated to modal analysis. In this specific case, the DSA was controlled with Hammer-3D12 software that runs directly on the HP-35670A and
eliminates the need for external devices. Within this software, the test element
geometry and transducer array are physically defined. FRFs were acquired
between the accelerometer and each hammer location as previously noted. Following a validity check of the averaged FRFs, curve fitting was applied to each of
the first four resonant frequencies. The individual modes were then assembled,
scaled, and presented as animated mode shapes on the DSA. Since this is a simple and symmetrical rotor, the Hammer-3D software was used in a single plane
mode. The resultant mode shapes from these impact tests were committed to
hard copy, and the first two modes are presented in Fig. 5-31.
As expected, the measured first mode was a pure overhung cantilever mode
12 David Forrest, Hammer-3D Version 2.01, Computer Program in Hewlett Packard Instrument Basic by Seattle Sound and Vibration, inc., Seattle, Washington, 1997.
239
Exposed Mixer Shaft

Directly Below Seal
1
Measured
1st Mode at
234 RPM
Measured
2nd Mode at
1,812 RPM
4
4
5
6
7
6
8
7
10
9
11
10
12
Upper
Mixer Blade
11
12
Bottom Mixer Blade
Fig. 531 FRF Measured Static Mode

Shapes Of Vertical Mixer Rotor
Fig. 532 Calculated Planar Mode

Shapes Of Vertical Mixer Rotor
that appeared at a frequency of 234 RPM. As shown in Fig. 5-31, the second
mode displayed a zero axis crossing between the mixer blades, and it had a measured natural frequency of 1,812 RPM. This measured FRF data was obtained
only on the exposed shaft sections below the shaft seal. There was no opportunity
to acquire any meaningful FRF data in the vicinity of the bearings. Again, this is
static mode shape data with a non-rotating shaft.
The next step consisted of generating an appropriate analytical model to
simulate the measured behavior. This was a difficult task since the specific shaft
material was unknown, and the internal hollow shaft diameters were likewise
unknown. However, the total rotor weight was known to be 615 pounds, and the
external shaft dimensions were easily measured. It was also noted that the top
portion of the mixer shaft underneath the pulleys and bearings was solid. The
hollow portion of the shaft was in three steps with decreasing diameters of 4.5,
4.0, and 3.5 inches. The weight of the pulleys and the mixer blades were measured on a shop scale, and the shaft material density was assumed to be 0.283
pounds per cubic inch. This density of steel was used since the shaft was magnetic, and therefore it was not any type of aluminum or stainless steel.
A simple model of the shaft was then constructed on a Microsoft Excel
spreadsheet. The external shaft dimensions were combined with the known com-
240
Chapter-5
ponent weights, plus the density of steel previously mentioned. It was assumed
that the wall thickness for each of the three sections of hollow shaft were constant. This wall thickness for the hollow sections was then varied until the overall rotor weight matched the total physical weight of 615 pounds. This match
occurred with a wall thickness of 0.5 inches, which seemed to be a reasonable
value for this rotor assembly.
The dimensional rotor data from the spreadsheet was then loaded into the
undamped critical speed program CRITSPD previously referenced in this chapter. In this software, a planar analysis was performed that consisted of setting
the polar inertia terms to zero. Basically this is used to simulate a stationary
non-rotating shaft. The bearing stiffness were then varied between 400,000 and
1,000,000 pounds per inch. As expected, this had little influence upon the calculated mode shapes or resonant frequencies. Certainly, this is a reasonable result
since better than 95% of the strain energy was contained in the shaft, and less
than 5% of the strain energy was in the bearings. Hence, the shaft properties
controlled the natural resonant frequencies, plus the associated mode shapes.
The final piece of unknown data for performing the CRITSPD calculations
was the modulus of elasticity E for the shaft material. Initially, the value for
steel of 30,000,000 Psi was used. This produced planar modes that did not match
the measured FRF results. A series of repetitive runs were made, and the value
of E was incrementally reduced for each run. At a level of 21,000,000 Psi for E,
the calculated planar results closely matched the measured FRF modes. Specifically, the first two computed modes are shown in Fig. 5-32.
The similarities between the measured FRF modes in Fig. 5-31 and the calculated CRITSPD modes in Fig. 5-32 are self-evident. The frequencies for both
first and second modes are consistent, and the comparable mode shapes are virtually identical. The analytical model covers the entire rotor up through the
drive pulleys, whereas the measured static model only addresses the exposed
shaft. The largest deviation occurs in the frequency of the first critical. The measured FRF data provided a value of 234 RPM, and the calculated planar mode
revealed a speed of 276 RPM for this first mode. Although the 42 RPM differential is an appreciable percentage of the resonant frequency, it is still well below
the normal operating speed range.
Since the zero speed planar model matches the static FRF results, it is concluded that the analytical model is an acceptable representation of the mixer
rotor. The next step requires activating the polar moment terms in the CRITSPD
program, and performing a normal synchronous analysis. This run indicated
that the first mode of the rotating shaft would occur at 277 RPM, and the second
critical would increase to 1,841 RPM. The predicted first critical of 277 RPM was
consistent with the plant observations of a resonance between 250 and 300 RPM.
Furthermore, the calculated frequency of the second critical was considerably
above the normal running speed range of 900 to 1,200 RPM.
At this point, the analytical model provided a good representation of the
real machine. This similarity gave confidence to pursue the final step of increasing the thickness of the two mixing blades from 0.188 to 0.488 inches. This provided additional weight to the rotor, plus additional inertia due to the 20 inch
241
Table 59 Summary Of Measured And Calculated Natural Frequencies For Vertical Mixer
Original Thin Mixer Blades
Thick Blades
Rotor
Resonance
Static FRF
Measurement
Planar
Calculation
Synchronous
Calculation
Synchronous
Calculation
1st Mode
234 RPM
276 RPM
277 RPM
244 RPM
2nd Mode
1,812 RPM
1,808 RPM
1,841 RPM
1,705 RPM
diameter of these blades. This change dropped the first mode to 244 RPM, and it
lowered the second critical to 1,705 RPM. Again, these frequencies are considerably removed from the operating speed range, and it is concluded that the additional blade thickness will not adversely influence the natural frequency
characteristics of this vertical mixer.
For comparative purposes, the entire array of measured and calculated natural frequencies of this vertical rotor are summarized in Table 5-9. Additionally,
the calculated mode shapes for the vertical mixer with the thicker mixer disks
are presented in Figs. 5-33 and 5-34 for the first and second modes respectively.
Fig. 533 Calculated Synchronous First

Mode With Thicker Mixer Disks
Fig. 534 Calculated Synchronous Second Mode With Thicker Mixer Disks
242
Chapter-5
In many ways, this vertical mixer case history is a simplistic example of

interleaving measurements and calculations to achieve a realistic engineering
solution. At all times, the machinery diagnostician must be cognizant of the fact
that direct solutions are not always possible, and indirect or hybrid approaches
are sometimes necessary to solve a problem. Furthermore, the accuracy of the
final result does not have to extend to the third decimal point. In many cases, if
you are within 5%, or perhaps 10% of the exact solution, that answer is fully
acceptable within the field environment. You do not want to be inaccurate in your
work, but then again, you do not want to try and attain some unrealistic measure of accuracy or precision.
BIBLIOGRAPHY
1. Chen, W.J., E. J. Gunter, and W. E. Gunter, DYROBES, Dynamics of Rotor Bearing
Systems, Version 4.21, Computer Program in MS-DOS by Rodyn Vibration, Inc.,
Charlottesville, Virginia, 1995.
2. Eisenmann, Robert C., Some realities of field balancing, Orbit, Vol. 18, No. 2 (June
1997), pp. 12-17.
3. Forrest, David, Hammer-3D Version 2.01, Computer Program in Hewlett Packard
Instrument Basic by Seattle Sound and Vibration, inc., Seattle, Washington, 1997.
4. Gunter, Edgar J., ROTSTB, Stability Program by Complex Matrix Transfer Method
- HP Version 3.3, Computer Program in Hewlett Packard Basic by Rodyn Vibration,
Inc., Charlottesville, Virginia, March, 1989, modified by Robert C. Eisenmann,
Machinery Diagnostics, Inc., Minden, Nevada, 1992.
5. Gunter, Edgar J., UNBAL, Unbalance Response of A Flexible Rotor - HP Version 4,
Computer Program in Hewlett Packard Basic by Rodyn Vibration, Inc., Charlottesville, Virginia, July, 1988, modified by Robert C. Eisenmann, Machinery Diagnostics, Inc., Minden, Nevada, 1992.
6. Gunter, E. J. and C. Gareth Gaston, CRITSPD-PC, Version 1.02, Computer program in MS-DOS by Rodyn Vibration, Inc., Charlottesville, Virginia, August, 1987.
7. Lund, J.W., Modal Response of a Flexible Rotor in Fluid Film Bearings, Transactions American Society of Mechanical Engineers, Paper No. 73-DET-98 (1973).
8. Myklestad, N.O., A New Method of Calculating Natural Modes of Uncoupled Bending Vibration of Airplane Wings and Other Types of Beams, Journal of the Aeronautical Sciences, Vol. 11, No. 2 (April 1944), pp. 153-162.
9. Prohl, M.A., A General Method for Calculating Critical Speeds of Flexible Rotors,
Journal of Applied Mechanics, Vol. 12, Transactions of the ASME, Vol. 67 (September 1945), pp. A142-148.
10. Singh, Murari P. and others, SAFE Diagram - A Design and Reliability Tool for
Turbine Blading, Proceedings of the Seventeenth Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University, College Station, Texas (November
1988), pp. 93-101.
11. Thomson, William T., Theory of Vibration with Applications, 4th Edition, Englewood
Cliffs, New Jersey: Prentice-Hall, 1993.
Transducer Characteristics
revious chapters have discussed the fundamental motion characteristics of machinery, including the inter-relationship
between displacement, velocity, acceleration, and frequency. Rotor mode shapes
and the role of mass distribution, support characteristics, stiffness, and damping
have also been reviewed. These physical characteristics have been expanded to
examine some of the analytical computations available. It is understandable that
mathematical modeling techniques provide some powerful tools for predicting
the behavior of rotating equipment. However, it must be recognized that the real
world is always more complicated than the models that are developed to explain
physical phenomena. At best, the mathematical models are approximations of
the actual physical system, and improvement or refinement of the model is often
dependent upon correlations with field observations, and direct machinery measurements.
Over the years, many knowledgeable individuals have stated and restated
that one good measurement is worth a thousand expert opinions. Hence, it is not
only desirable, it is mandatory that machinery measurements be considered as a
major problem solving tool. This topic will be explored from the standpoint of
static measurements, such as position measurements versus machine clearances. In addition, the complex dynamic measurements associated with machinery vibratory behavior will be addressed throughout this text. Within this
chapter, specific consideration will be given to the dynamic measurement transducers, and their associated signal characteristics.
In most cases, mechanical motion cannot be adequately quantified with
human senses. Although the human eye can observe objects vibrating at amplitudes of 10 Mils,p-p, and the ear can detect frequencies of 10,000 Hz, and differential acceleration levels of 0.1 Gs,o-p are normally considered to be unpleasant,
there are substantial differences in perception and threshold levels between
individuals. In order to have some type of measurement quantification and
repeatability, an interface device must be provided between the operating
machinery and the diagnostician. The devices used for this task are electronic
sensors, or transducers. These transducers convert numerous types of mechanical behavior into proportional electronic signals. The transducer outputs are usually converted into voltage sensitive signals that may be recorded and processed
243
244
Chapter-6
with various electronic instruments. Within this chapter, the operational characteristics of the major types of industrial transducers will be examined. In addition, the calibration methods, plus the fundamental advantages and
disadvantages of each type of common industrial sensor will be reviewed.
BASIC SIGNAL ATTRIBUTES

Industrial transducers used for measurement of dynamic characteristics
typically fall into three distinct categories. This includes the shaft sensing proximity probes, the mechanical motion velocity coils, and the solid state piezoelectric devices. Each of these three groups are generally used for measurement of
displacement, velocity, and acceleration respectively. Each of the transducer
types exhibits an array of strengths combined with a set of limitations. At this
point in technology, there is no universal sensor that can be used for all measurements, on all machines, under all conditions. Thus, the machinery diagnostician
must be intimately familiar with the characteristics of each type of transducer,
and apply these devices to their best advantage.
Dynamic transducers measure events that occur in very small time increments. For instance, a thermocouple in process service may be used to measure
temperatures that vary in minutes or even hours. Dynamic transducers used for
vibration measurements must be able to detect phenomena that occur in fractions of a second. The resultant electronic signals are complex traces that contain significant information, and this data is generally quantified in terms of the
following fundamental parameters:
Amplitude (Magnitude or Severity)

Frequency (Rate of Occurrence)
Timing (Phase Relationship)
Shape (Frequency Content)
Position (from Proximity Probes Only)
The amplitude of the dynamic signal is generally proportional to the

severity of the vibratory motion. That is, large amplitudes are directly related to
high levels of vibration, and this is generally associated with the occurrence of a
mechanical problem. Conversely, low vibration amplitudes are typically associated with proper behavior of the machinery, and the absence of mechanical difficulties. In most cases, the degradation of a piece of mechanical equipment is
evident by increasing vibration amplitudes. There are exceptions to this general
rule, and occasionally a machine will exhibit decreasing vibration amplitudes as
mechanical condition degenerates (e.g., case history 50).
However, in most circumstances, amplitude is a direct indicator of vibration
severity, and general machinery condition. The terminology used to define amplitude has originated from different sources, with several different meanings. For
instance, consider Fig. 6-1 of a sine wave in the time domain. In this diagram
three different types of amplitude measurements are identified. The total magni-
Basic Signal Attributes
245
1.0
RMS
0.6
Amplitude
0.4
0.2
Peak to Peak
Zero to Peak
0.8
0.0
-0.2
-0.4
-0.6
-0.8
Fig. 61 Three Types Of

Common Sine Wave
Amplitude Designations
-1.0
0
10
12
14
16
Time (Seconds)
tude of the curve is represented by the Peak to Peak value. As its name implies,
this measurement of amplitude extends from the lowest portion of the dynamic
signal (i.e., bottom peak) to the highest portion of the signal (i.e., top peak). From
Fig. 6-1, the minimum amplitude is -1.0, and the maximum or peak amplitude is
equal to +1.0. Therefore, the total or peak to peak amplitude is equal to the total
signal height of 2.0. Occasionally, historic references will be made to double
amplitude measurements, which are synonymous with peak to peak values.
Shaft vibration measurements are generally expressed as peak to peak
amplitudes. In the USA, units of Mils are used for displacement measurements
(1 Mil = 0.001 Inches). This combination of engineering units and amplitude
selection is abbreviated as Mils,p-p. It is convenient to use peak to peak displacement to relate the severity of shaft motion against the total bearing clearance.
For instance, assume that a shaft radial vibration amplitude is 1.0 Mil,p-p, and
the total bearing clearance is 10 Mils. The vibratory motion is easily determined
to be 10% of the available diametrical clearance. Intuitively, this is an acceptable
and comfortable level. However, if the shaft vibration was 9.0 Mils,p-p, the
motion would be 90% of the available bearing clearance. This magnitude of
vibration with respect to bearing clearance would be considered as unacceptable.
The second common type of amplitude measurement shown in Fig. 6-1 is
the Zero to Peak value. This measurement extends from the middle of the
dynamic signal to the highest portion of the signal (i.e., top peak). From the diagram, the maximum or peak is +1.0, and the zero to peak amplitude is 1.0. Casing vibration measurements are normally expressed as a zero to peak value. In
the USA, the units of Inches per Second (IPS) are used for velocity measurements, and the acceleration of gravity in Gs are used for acceleration. This combination of amplitude measurements and engineering units are generally
abbreviated as IPS,o-p, and Gs,o-p, respectively. Clearly, for a simple voltage signal, the relationship between these peak based amplitude measurements may be
accurately expressed as:
246
Chapter-6
Amplitude peak
to peak
= 2 Amplitude zero
to peak
(6-1)
The third common type of amplitude measurement is the Root Mean

Square (RMS) value. As shown in Fig. 6-1, the RMS amplitude is lower than the
zero to peak value. For a pure sine wave, the actual reduction is equal to 2 2 , or
a numerical value of 0.7071. The following conversions may be used to relate all
three types of common amplitude measurements:
Amplitude rms = 0.7071 Amplitude zero
to peak
(6-2)
Amplitude rms = 0.3536 Amplitude peak
to peak
(6-3)
RMS values are used in Europe for vibration measurements. However, in

the USA this type of magnitude measurement is seldom used for machinery
vibration. The largest utilization of RMS occurs with electrical devices such as
voltmeters. For instance, if a digital multimeter is used to measure AC voltage
from a wall outlet, a reading of 115 volts may be observed. This voltage of 115
Volts,rms, is equivalent to 325 Volts,p-p based upon equation 6-2. Hence, the common household power outlet really has 325 volts of electrical potential difference.
Other types of voltage measurements such as Average amplitudes are occasionally used. However, within this text, vibration amplitudes will be expressed
in accordance with the previous discussion. Specifically, vibration amplitude
measurements within this book will be stated as:
Displacement:
Mils,peak to peak = Mils,p-p
Velocity:
IPS,zero to peak
= IPS,o-p
Acceleration:
Gs,zero to peak
= Gs,o-p
Conversion of vibration units at a fixed frequency may be achieved with

equations (2-17) through (2-22). Using these expressions, one vibration unit may
be converted into another. For instance, a casing velocity measurement in IPS,o-p
may be converted to a casing displacement measurement in Mils,p-p. This type of
conversion is useful when shaft versus casing measurements must be correlated.
In all cases, the frequency of vibration must be included in the conversion calculation. Furthermore, the timing or phase relationship between vibration signals
should also be considered.
The signal frequency is determined by the reciprocal of the period. To use
a consistent example, the previously discussed sine wave is reproduced in Fig. 62. In this diagram, three different time ranges are identified on the curve. Basically, all three ranges are all identical, and they each define one cycle. In one
case, a zero axis crossing is used to identify a complete cycle. In the other two
ranges, the time between two consecutive peaks, or two consecutive valleys are
identified. It does not matter what part of the cycle is used to determine the
period, as long as the measurement point is identical from one cycle to another.
For a repetitive signal, the time required to complete one cycle will remain con-
247
One Cycle
One Cycle
1.0
0.8
0.6
Amplitude
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
Fig. 62 Period Of A Sine

Wave Measured At Three
Different But Repetitive
Locations
-1.0
0
10
Time
12
14
16
(Seconds)
One Cycle
stant. The associated frequency may be determined from equation (2-1), and this
expression is repeated as follows:
1
Frequency = ------------------Period
(6-4)
Period is measured in time units of Seconds or Minutes, and it identifies the

length of time necessary to complete one cycle. If the Period is measured in Seconds, then the reciprocal frequency must carry the engineering units of Cycles
per Second (CPS). This is also know as a Frequency in hertz (Hz). Similarly, if the
Period is measured in Minutes, then the Frequency will carry units of Cycles per
Minute (CPM), or Revolutions per Minute (RPM) if a machine speed is identified.
The relationship between these frequency units are as follows:
Frequency (CPM) = 60 Frequency (CPS or Hz)
(6-5)
The example sine wave shown in Fig. 6-2 has a period of 6.28. If the time
scale is in seconds as noted on this plot, then the fundamental period of the signal is 6.28 seconds. The frequency may be determined from the previous equations (6-4) as follows:
1 Cycle
1
Seconds
Cycles
Frequency = ------------------- = ------------------------------ 60 --------------------- = 9.55 ----------------Minute
Minute
6.28 Seconds
Period
For comparative purposes, now assume that the time scale on Fig. 6-2 is in
milliseconds instead of seconds. One second contains 1,000 milliseconds, hence a
conversion factor of 1,000 exists between the two scales. With this new scale, the
fundamental Period of the sine wave would be 6.28 x 10-3 seconds. The frequency
248
Chapter-6
may again be determined from equation 6-4:

1 Cycle
1
Seconds
Cycles
Frequency = ------------------- = --------------------------------------- 60 --------------------- = 9, 550 ----------------Minute
Minute
0.00628 Seconds
Period
From this example, it is clear that low frequency motion has a long period
(e.g., an earthquake). High frequency vibration is necessarily associated with a
short period (e.g., a turbine rotational speed). This basic relationship is not only
important during the analysis of mechanical behavior, it must also be addressed
during the selection of an appropriate vibration transducer suite, and the associated diagnostic instrumentation.
Before considering the details of any vibration transducer, it is necessary to
discuss the timing between events. This is a vitally important point, since very
few mechanical problems are solved based upon one measurement with a single
transducer. In most instances, multiple transducers are applied, and the signal
outputs are examined for relative amplitudes, frequency content, and the timing
or phase between signals.
This concept of timing may be expressed in various ways, and it is often
confused by different triggering schemes. In the simplest format, timing is just
the time delay between two signals as depicted in Fig. 6-3. The solid line represents the same sine wave that has been used for the two previous examples. The
dotted line describes another sine wave that is offset in time by one quarter of a
cycle from the solid line. On a rotating machine, one complete cycle is equal to
one revolution, or 360 degrees. Hence, one cycle is 360, and a quarter of a cycle
is 90 as shown in Fig. 6-3.
The relative timing between the two sine waves may be described in two
different ways. With time progressing from left to right, it may be properly stated
that the solid curve leads the dotted curve by 90. It may also be stated that the
dotted curve lags the solid curve in time by 90. Both statements are correct, and
it can be very confusing if the same parameter is described in more than one way.
One Cycle
360
90
1.0
0.8
0.6
Amplitude
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
Fig. 63 Phase Or Timing

Differential Between Two
Sine Waves
-1.0
0
10
Time (Seconds)
12
14
16
249
To alleviate any potential of misinterpretation, the following six rules for phase
angles shall be consistently applied throughout this text:
1. All phase measurements require a minimum of two signals. One signal will
be the dynamic motion signal (e.g., vibration), and the other signal must be
the trigger signal (e.g., Keyphasor).
2. Phase measurements require that both the dynamic motion signal and the
trigger occur at the same frequency.
3. Phase measurements must be made within the same time cycle.
4. The positive peak of the filtered motion signal (e.g., vibration) shall always
be considered as the zero degree (0) reference point for that transducer.
5. Looking backwards in time, the phase angle shall be the angular distance
between the positive peak of the motion signal, and the first trigger point.
6. Phase angles shall be expressed as degrees of phase lag. The term lag
implies a negative angle, i.e., against rotation, and backwards against time.
These rules apply for all measured vibration, pressure pulsation, force signals, etc. Vectors derived from the measured dynamic signals shall maintain the
same rules, and all analytical calculations shall be configured to conform to similar criteria. Finally, the location of balance weights, and other references to specific angular locations on a rotor shall be consistent with the above rules.
To illustrate this phase convention, Fig. 6-4 was prepared. This sketch
describes the reference sine wave, combined with two different timing marks
(dots). The depicted events are filtered at the same frequency, and the peak of the
motion signal is identified as the zero degree (0) reference point. If the trigger
point was coincident with this positive peak of the vibration signal, the signal
phase angle would be 0. If the trigger signal occurred one quarter of a cycle earlier, the resultant signal phase angle would be a 90 lag.
One Cycle
360
270
90
1.0
0.8
Reference
at 0
0.6
Amplitude
0.4
0.2
0.0
-0.2
-0.4
-0.8
Fig. 64 Sine Wave Combined With Two Different

Phase Reference Points
Trigger
Point
at 90
Trigger
Point
at 270
-0.6
-1.0
0
10
Time (Seconds)
12
14
16
250
Chapter-6
The second trigger point on Fig. 6-4 is located at three quarters of a cycle
before the peak of the motion (or vibration) signal. The phase angle for three
quarters of a cycle (3/4 times 360) is equal to 270, and this trigger point is also
identified. In a similar manner, all phase angles will be determined, and
reported in consistent units of degrees of phase lag.
One of the popular misconceptions on phase measurements is that the
phase angle is referenced or determined from the angular location of the timing
or Keyphasor transducer. It must be recognized that the timing probe is used to
provide an angular reference point between the rotating system and the stationary mechanical system. As discussed throughout this text, the timing probe identifies the angular or rotative position of the rotor at the exact instant in time
when the trigger pulse is produced. The actual phase measurements are measured against rotation from the angular location of the vibration transducer. To
state it another way, the angular position of the vibration probe always represents zero degrees (0), irrespective of the timing probe location.
The trigger point in Fig. 6-4 is shown as a single dot. In actuality, the dot
originates from analog oscilloscope utilization techniques where a pulse shaped
timing signal is connected to the scope Z-Axis input. A negative going pulse input
would produce a blank followed by a bright spot as shown in Fig. 6-5. The Z-Axis
oscilloscope input of the timing signal is also referred to as blanking. For most
oscilloscopes, this blank-bright sequence is superimposed upon the main
dynamic motion signal. The portion of the trigger signal with a negative (downhill) slope produces a void or blank spot on the main motion signal. Conversely,
the portion of the trigger signal with a positive (uphill) slope produces an intensified bright spot upon the motion signal. If a digital oscilloscope is used instead of
an analog scope, the dot intensification does not occur, and only a blank spot
Trigger Pulse
Negative Slope
Produces
Blank Spot
Motion Curve
Positive Slope
Produces
Bright Spot
Positive Slope
Produces
Bright Spot
Motion Curve
Time
Fig. 65 Negative Trigger Pulse Signal
And Associated Blank-Bright Sequence
Negative Slope
Produces
Blank Spot
Trigger Pulse
Time
Fig. 66 Positive Trigger Pulse Signal And
Associated Bright-Blank Sequence
251
appears on the display. The width of the digital scope blank spot is equal to the
width of the timing pulse.
A negative going trigger pulse is typical for a proximity probe observing a
notch or a hole drilled into a shaft (e.g., Fig. 6-5). For a probe triggering off a projection such a shaft key, or for an optical transducer, a positive going trigger signal will be generated. This type of positive pulse is shown in Fig. 6-6. It is noted
that the positive slope of the trigger still produces a bright spot, and the negative
slope still shows up as a blank spot on the dynamic motion curve. The sequence
of these events are now reversed to be bright then blank due to the directional
characteristics of the timing pulse. Since all oscilloscopes have a sweep that displays signals from left to right on the screen, the actual sequence for any type of
pulse on any oscilloscope can be easily determined by visual observation of the
signals in the time domain. Again, this type of blank-bright display only applies
to an analog oscilloscope such as a Tektronix 5110. A digital scope with blanking
such as the HP-54600B will only display the blank spot.
The synchronizing trigger on an oscilloscope, plus the trigger on a tracking
filter or DSA generally require the user to select a positive or a negative slope.
Often these devices are set to trigger at about 50% of the slope (halfway between
the upper and lower voltage). On many instruments, it is also possible to manually set the trigger at any other part of the pulse slope. In all cases, it is highly
recommended that the input Keyphasor signal be observed in the time domain,
and the trigger point verified. On some signals it may be necessary to manually
adjust the instrument trigger point to obtain a consistent trigger signal. From a
mechanical interface standpoint, the physical trigger point is discussed in further detail in chapters 7, 8 and 11 of this text.
Many computerized data acquisition and processing systems have mimicked this analog oscilloscope logic into the computer output displays. Hence, the
blank-bright or bright-blank analog timing convention remains as an integral
part of the data presentation. This ability to precisely measure the timing
between events proves to be very useful for incidents occurring at the same frequency. However, most industrial machines display more than a single discrete
running speed frequency.
2.0
2.0
1X Only
1X + 2X
1.5
1.0
1.0
0.5
0.5
Amplitude
Amplitude
1.5
0.0
-0.5
0.0
-0.5
-1.0
-1.0
-1.5
-1.5
-2.0
-2.0
0
10
12
14
Time (Seconds)
Fig. 67 Fundamental Sine Wave
16
10
12
14
16
Time (Seconds)
Fig. 68 Fundamental Plus 2nd Harmonic
252
Chapter-6
2.0
2.0
1X + 3X
1X + 4X
1.5
1.0
1.0
0.5
0.5
Amplitude
Amplitude
1.5
0.0
-0.5
0.0
-0.5
-1.0
-1.0
-1.5
-1.5
-2.0
-2.0
0
10
12
14
16
Time (Seconds)
Fig. 69 Fundamental Plus 3rd Harmonic
10
12
14
16
Time (Seconds)
Fig. 610 Fundamental Plus 4th Harmonic
For instance, consider the time domain plots in Figs. 6-7 through 6-10. In
each plot, the fundamental 1X amplitude, frequency, and timing are identical to
the sine wave originally used in Fig. 6-1. However, it is apparent that the overall
amplitudes, and the shape or general appearance of Figs. 6-8 through 6-10 have
significantly changed from the initial sine wave. These changes are due to the
addition of higher order harmonics to the fundamental. In all three cases, the
additional harmonic components are set to be even multiples of the fundamental,
and the amplitudes are equal in magnitude to the fundamental. The resultant
time domain curves each have distinctive shapes or patterns that may be further
altered by changing the respective amplitudes and/or the timing relationships of
the higher order harmonics. The inclusion of non-synchronous frequencies will
produce additional pattern variations. Specifically, case histories on signal summation, amplitude modulation, and frequency modulation are presented in chapter 7.
The complexity of time domain signals manifests as a wide variety of potential shapes and patterns. It is easy to visually recognize some of the simple signal combinations, but the more complex signals may not be comprehensible in
the time domain. For these situations, additional signal processing techniques
and instruments are required.
Various types of filters and frequency analyzers may be employed to dissect
these time domain signals into quantifiable and manageable portions. These
electronic devices and their specific attributes are reviewed in chapter 8. However, prior to any further discussion of signal manipulation, it would be desirable
to examine the characteristics of the dynamic transducers that are used for the
majority of the industrial measurements on process machinery.
Proximity Displacement Probes
253
PROXIMITY DISPLACEMENT PROBES

Shaft sensing proximity probes are used to obtain relative displacement
measurements of rotating or reciprocating shaft surfaces. These are non-contacting transducers that are mounted on a reasonably stationary mechanical structure (e.g., bearing housing). From the mounting point they observe the static and
dynamic displacement behavior of the moving machinery element. Proximity
probes are supplied in a wide variety of sizes, shapes, and configurations. For
example, the photograph in Fig. 6-11 depicts four different sizes and configurations of Bently Nevada Corporation proximity probes.
Fig. 611 Typical Configurations Of Proximity Probes

Manufactured by Bently
Nevada Corporation
The left hand probe in this photo consists of a fully threaded stainless steel
body with an 8 mm Ryton probe tip. The protective probe tip may also be constructed of durable high performance plastics such as polyphenylene sulfide
(PPS) that is capable of withstanding harsh physical and/or chemical environments. For less severe applications, the two middle probes in Fig. 6-11 contain
fiberglass tips of 8 mm and 5 mm diameters mounted in fully threaded bodies.
Finally, a different body configuration is shown in the probe at the right side of
the photograph. This is a reverse mount probe that screws into a mounting
assembly (stinger) that can be cut to length to accommodate a variety of installation requirements.
A probe tip diameter of 5 mm (0.0197 inches) is typically mounted in a
stainless steel body with an external thread of 1/4x28 UNF. Similarly, the standard 8 mm (0.315 inches) diameter probe tip is usually mounted in stainless
steel body with external 3/8x24 UNF threads. Smaller diameter probe tips (e.g.,
1/8) with lower dynamic ranges, and larger diameters (e.g., 2) for extended
range measurements are also commercially available. However, the vast majority of the industrial applications use 5 mm and 8 mm probes.
Regardless of the physical configuration, all eddy current proximity probes
consist of the same basic components. For example, consider the sketch of a typical probe assembly in Fig. 6-12. In this diagram, a flat wound coil located close to
the probe tip is connected by two wires to a coaxial cable that runs between the
probe and Proximitor. This coaxial cable must be electrically tuned to a specific
length in order to maintain the proper impedance between the probe and the
254
Chapter-6
Coaxial Cable to
Proximitor
Probe Body
Protective Tip
Fig. 612 Typical Components of A Proximity Probe
Coil
Proximitor. If the interconnecting cable length is altered from the correct

value, transducer calibration will be influenced.
All external wiring connections are terminated at the Proximitor. Typically this is a fully shielded three conductor cable. The cable shield is normally
grounded at the monitor rack, and the field end of the shield is allowed to float.
This prevents the development of ground loops in the transducer wiring. Within
the three wire cable, one wire is used for the output signal (white), the second
wire provides a common ground (black), and the third wire is the power input
(red). A power input of -24 volts DC is applied to the Proximitor from a monitor,
or a regulated DC power supply. The Prox contains an internal oscillator that
converts some of the input energy into a radio frequency signal in the megahertz
range. This high frequency signal is directed to the probe coil via the coaxial
cable. The Proximitor is also know generically as an Oscillator-Demodulator.
The flat pancake coil at the tip of the probe broadcasts this radio frequency
signal into the surrounding area as a magnetic field. If a conductive material
does not intercept the magnetic field, there is no power loss in the radio frequency signal. However, if a conductive substance intercepts the magnetic field,
eddy currents are generated on the surface of the material, and power is drained
from the radio frequency signal. As the conductive material approaches the
probe tip, additional power is consumed by the eddy currents on the surface of
the conductor. When the probe is in contact with the conductive material, the
majority of the power radiated by the probe tip is absorbed by the material. As
the power loss varies, the output signal from the Proximitor also exhibits a
change in voltage. In all cases, a small gap produces a small output voltage, and
a large gap results in a large output voltage from the Proximitor.
The relationship between distance and output voltage is achieved by a combination of electronic circuits within the Proximitor. For instance, a demodulator removes the high frequency carrier signal, and a linearization circuitry
provides a reasonably flat curve over a typical range of 80 to 100 Mils (0.080 to
0.100 Inches). These signal output characteristics may be easily checked by per-
255
forming a simple distance versus voltage calibration. This type of calibration is

generally performed with a spindle micrometer device as shown in Fig. 6-13. In
this configuration, the proximity probe remains in a fixed position, and the spindle micrometer is used to accurately move a circular target back and fourth
against the probe. Other fixture designs may be used that consist of a stationary
target combined with a moveable probe mount. In either case, the probe to target
distance must be both adjustable, and accurately measurable.
Fig. 613 Proximity Probe

Static Calibration Fixture Spindle Micrometer From
Bently Nevada Corp. TK-3
As previously mentioned, when the probe tip is in physical contact with the
observed conductive surface, the transducer output is at a minimum voltage
level. As the distance between the probe and the target material increases, the
output voltage increases in a proportional manner. A typical calibration procedure requires the tabulation of output voltage versus the physical distance
between the probe and the target. For accuracy, these readings are generally performed at 5 Mil (0.005 Inch) increments, and the results are normally presented
in a graphical format such as Fig. 6-14.
Within this calibration plot, the vertical axis displays the Proximitor DC
output voltage. Since a minus 24 volt DC power supply was supplied to the Proximitor, the output voltages are also negative voltages. It is understandable that
the final output voltage remains less than the power input level. The horizontal
axis in this plot presents the physical distance between the probe tip and the target surface in Mils. For this particular calibration plot, a range of 10 to 110 Mils
was spanned. The actual coordinates for each plotted point are listed next to the
curve. This information not only provides good documentation of the calibration
curve, it also allows the easy computation of transducer sensitivity. In all cases,
proximity probe sensitivity is defined as the slope of the calibration curve as in
the following calibration expression:
Differential Voltage
Proximity Probe Sensitivity = ------------------------------------------------------------Differential Gap
(6-6)
It is presumed that the linear portion of the calibration curve is a straight

line. If deviations from a straight line are encountered, there is probably something wrong with the transducer system or the calibration setup. Once a proper
straight line is established for the linear portion of the curve, the transducer sensitivity may be determined by computing the slope of the line. If voltage output
256
Chapter-6
22.0
(110;21.25)
(105;20.5)
Proximitor Static Output Gap Voltage (Negative Volts DC)
20.0
(100;19.71)
(95;18.88)
18.0
(90;17.99)
(85;17.05)
(80;16.1)
16.0
(75;15.13)
(70;14.12)
14.0
(65;13.12)
(60;12.08)
12.0
(55;11.1)
(50;10.11)
10.0
(45;9.1)
(40;8.09)
8.0
(35;7.08)
6.0
(30;6.04)
(25;5.04)
4.0
(20;4.03)
(15;3.04)
2.0(10;2.03)J
0.0
0
10
20
30
40
50
60
70
80
90
100
110
120
Distance Between Probe Tip and Target (Mils)
Fig. 614 Proximity Probe Static Calibration Curve On 4140 Target Material
values for 20 and 100 Mils are extracted from the calibration plot in Fig. 6-14,
the probe sensitivity may be determined as follows:
15.68 Volts
19.71 4.03 Volts
milliVolts
milliVolts
Sensitivity = ------------------------------------------- = -------------------------- 1, 000 --------------------- = 196 ------------------------Mil
Volt
80 Mils
100 20 Mils
The normal voltage sensitivity for this class of transducer is 200 millivolts/
Mil 5%. Thus, values between 190 and 210 millivolts/Mil are acceptable, and
the calculated slope of 196 millivolts/Mil falls within this acceptance range. In
many cases, the proximity probe calibration is simply used to verify compliance
with the standard sensitivity used for system calibration, trending and analysis.
However, in situations requiring additional measurement accuracy, the actual
curve slope will be used. For numerical computations it is common to perform a
curve fit of the calibration curve. For instance, a linear curve fit on the probe calibration plot in Fig. 6-14 produces the following expression for determining out-
257
put voltage based upon a gap for this particular proximity probe:
Voltage = 0.1957 Gap + 0.2395
This expression may be used for various calculations, or it may be incorporated into computerized monitoring or trending programs. Since the calibration
curve is nominally a straight line, the 0.2395 value is the zero gap intercept
point. Also, the 0.1957 multiplier is the slope of the curve that corresponds to the
manually calculated value of 0.196 Volts per Mil (196 mv/Mil). It is useful to recognize that the reciprocal of the standard sensitivity is 5.0 Mils per Volt. This
format is sometimes easier to remember, and easier to convert units. For example, if a transducer exhibits a 2.0 volt change in DC gap voltage, this is equivalent to a 10.0 Mil shift in gap between the stationary probe and the observed
surface (i.e., 2.0 Volts x 5.0 Mils/Volt = 10.0 Mils).
Since proximity probe requires an electrical coupling to the observed surface, it is reasonable to conclude that variations in the target will influence the
transduction. In actuality, anything that distorts the crystal lattice of the conductor will influence probe sensitivity. Changes in electrical resistivity, magnetic
permeability, residual magnetism, or localized stress concentrations will appear
as noise on the signal. Dependent on their magnitude, they may also change the
scale factor. Material changes will have a significant effect upon the sensitivity.
In the preceding discussion, sensitivity was assumed to be 200 mv/Mil. If this
was calibrated on 4140 steel, the same transducer system would exhibit output
sensitivities with other common metals as summarized in Table 6-1.
Table 61 Variation Of Proximity Probe Voltage Sensitivity With Different Metals
Material
Probe Voltage Sensitivity
Copper
380 millivolts/Mil
Aluminum
370 millivolts/Mil
Brass
330 millivolts/Mil
Tungsten Carbide
290 millivolts/Mil
Stainless Steel
250 millivolts/Mil
Steel 4140 or 4340
200 millivolts/Mil
These variations in signal voltage sensitivity may be used directly for computation of displacement. In addition, any monitors, computers, or analytical
instrumentation connected to the probe output may be programed with the precise sensitivity for the observed material. Alternatively, the Proximitor may be
electrically tuned to the new material, and the output scale factor adjusted to
obtain a consistent output of 200 mv/Mil. Either approach may be successfully
applied. However, it is highly recommended that a uniform and consistent
approach within the entire operating complex be established and maintained.
Regardless of the specific calibration for a transducer, it is clear that the
258
Chapter-6
proximity probe provides a direct means for converting distances into electronic
voltages. As such, this type of transducer has been referred to as an electronic
micrometer. This characteristic is widely used in many industrial applications.
For instance, measurement of average axial (thrust) position of rotating elements with respect to their stationary casings has evolved into a fundamental
machinery protection measurement.
For thrust position measurements, the machine is assembled with knowledge of the forward and reverse rub points of the rotor against the stationary elements. Thrust bearing position is shimmed to obtain a specific axial clearance
within the machine (e.g., 1st stage nozzle clearance in a steam turbine). Float
shims are then ground to maintain the correct float of the thrust collar(s) within
the bearing. As illustrated in Figs. 4-21 and 4-22, the physical configuration of
the thrust bearing, the actual probe location, plus the direction of normal thrust
loads must be considered during setup of a thrust probe installation.
When the mechanical clearances are properly established and verified, the
thrust (axial) proximity probes are installed. Typically, two probes are mounted
at each thrust bearing to provide full redundancy. These thrust probes are usually connected to a dual voting thrust position monitor. In order to have full confidence in the thrust monitor readings, it is necessary to have physical
verification of the mechanical thrust setup, and full correlation throughout the
system. This is achieved by acquiring and comparing several different measurements. Normally, a dial indicator is positioned at the end of the shaft, and zeroed
with the rotor sitting hard against the active shoes. A digital multimeter (voltmeter) is connected to the Proximitor output to allow measurement of the output voltage. Normally, a second digital multimeter is connected to the monitor
input to verify that excessive line losses have not occurred between the Proximitor output and the monitor input. In addition, the meter reading from the
thrust position monitor must be obtained. If a computerized trending system is
installed, the thrust position indication, plus the digitized gap voltage readings
must be compared against the actual transducer output values.
With the various readings established, the rotor is physically bumped or
moved back and forth between the active and the inactive thrust shoes. Depending on the philosophy of the particular operating company, the zero point on the
thrust monitor may be set for the active thrust shoes, or the middle of the float
zone. Initially, this is an arbitrary point, but once established, it must be maintained for all future measurements on the machine. Furthermore, all machines
within a plant complex should be set up with the same reference scheme. It is
unnecessarily confusing to all parties to have some machines with a zero thrust
position at the center of the float zone, and other units with zero equal to the
rotor positioned hard against the active thrust shoes. Pick one scheme, and stay
with it for all of the machines within an operating complex.
During a routine setup, the probes are initially set to a gap voltage at the
middle of the curve (e.g., approximately -12 volts DC). The rotor is then manually
thrust back and forth between the active and inactive thrust shoes. Based upon
the proximity probe calibration curve, the desired zero point, and the actual
thrust bearing float zone the probes are then reset to realistic gap voltages.
259
Once the probes are set and locked into position, the rotor is bumped back and
forth at least three times to compare and verify all readings. Normally, it is good
practice to obtain a set of soft readings when the rotor stops at the thrust shoes.
Next, a heavy pull on the pry bar will usually result in additional movement to a
hard thrust position. The difference between the soft and the hard thrust is due
to compression of any springiness in the thrust assembly. With full thrust loads,
the machine should typically run in the hard thrust positions.
When all of the readings are consistent, the rotor is rolled 180 (half a
turn), and three more sets of readings are obtained. Each group of three readings
must indicate consistent hard thrust points. In addition, the 180 shaft rotation
should produce similar results. If the values are significantly different, the
thrust bearing should be disassembled and carefully reexamined. Variations can
be due to anything from a steel burr, to a cocked thrust collar, to a warped bearing housing, or a mismatch between upper and lower halves. In any case, the
reasons for the thrust inconsistencies must be identified, and corrected before
22.0
(110;21.25)
Meter
Scale
(105;20.5)
Proximitor Static Output Gap Voltage (Negative Volts DC)
20.0
Normal
Increasing
Gap Voltage
18.0
(100;19.71)
Normal
(95;18.88)
+40 Trip
(90;17.99)
(85;17.05)
+30 Alert
(80;16.1)
16.0
(75;15.13)
(70;14.12)
14.0
at +17 Mils
(60;12.08)
Rotor North
-13.66 Volts
(65;13.12)
12.0
Active Shoes
+20
3,380 RPM
-13.2 Volts
at +16 Mils
+10
(55;11.1)
(50;10.11)
10.0
Inactive Shoes
-9.41 Volts
at -4 Mils
(40;8.09)
8.0
Rotor South
(45;9.1)
-10
(35;7.08)
6.0
-20 Alert
(30;6.04)
(25;5.04)
4.0
-30 Trip
(20;4.03)
(15;3.04)
2.0(10;2.03)J
Mechanical
Float
-40
21 Mils
Counter
0.0
0
10
20
30
40
50
60
70
80
90
100
110
120
Fig. 615 Steam Turbine Proximity Probe Thrust Position Calibration Curve
260
Chapter-6
final assembly, and startup of the machine.

At the conclusion of the field setup, the acquired information should be documented in a usable format. Generally, the probe calibration curve should be
used to summarize the thrust setup. For example, consider the chart presented
in Fig. 6-15 depicting the thrust calibration on a large steam turbine. This curve
describes one of two thrust channels. The characteristics exhibited by the companion thrust channel display very similar behavior to this example curve. In
this machine, the axial probe installation results in a gap voltage characteristic
that increases with normal turbine thrust position. Hence, increasing gap voltages are identified as Normal, and reverse thrust positions are identified as
Counter on the graph. The mechanical float zone measured with a dial indicator
is shown as 21 Mils. With the rotor positioned hard on the active shoes, the output gap voltage is shown as -13.66 volts DC. As the rotor is bumped back hard
against the inactive thrust shoes, the gap voltage decreases to -9.41 volts DC.
Based upon a normal sensitivity of 200 millivolts per Mil (or 5.0 Mils per Volt),
the thrust float measured by the probe voltage is determined as follows:
Thrust Float probe
volts
Mils
= ( 13.66 9.41 Volts ) 5.0 --------- = 4.25 5.0 = 21.2 Mils
Volt
This probe gap differential agrees with the mechanical float of 21 Mils. In
addition, the thrust monitor meter scale shown to the right of the graph displays
the following variation across the float zone:
Thrust Float monitor = ( +17 Mils ) ( 4 Mils ) = 21 Mils
Once more, proper agreement is achieved between the mechanical float,
probe voltage float, and the indicated monitor float zone. Hence, there is good
confidence in the accuracy of this thrust position measurement, and the relationship between the electronics and the actual mechanical system. The calibration
plot in Fig. 6-15 also indicates the Normal and Counter setpoints for the thrust
Alert and Trip. In all cases, the Alert or Alarm point should set well within the
available babbitt thickness of the thrust bearing. The second setpoint of Trip or
Danger should be set with the idea of saving the rotor. This may allow damage or
even destruction to the thrust bearing. Certainly the Trip point must be less
than the forward and reverse axial rub points to allow the machine to coastdown
without rotor contact to the stationary casing.
The running position of this turbine at 3,681 RPM is shown on the right
hand side of Fig. 6-15. This steady state operating position at full process load is
reasonably close to the hard position of the rotor on the active shoes. This is the
normally trended position measurement, and it should be recognized that the
actual reading will be dependent upon many other elements. Consideration
should always be given to items such as the actual thrust load, compressibility of
the thrust shoes, thrust balancing scheme for the machine, condition of interstage labyrinths, probe temperature sensitivity, ambient conditions, etc.
One final point on Fig. 6-15 should be mentioned. Note that the normal trip
setpoint is located at +40 Mils, and the counter trip is at -30 Mils. Actually, the
261
thrust float zone is set about 8 to 10 Mils high on the calibration curve. The
available transducer operating range would be better utilized if the probe cold
gap voltage had been set in the vicinity of -12.0 volts DC.
These position measurements with proximity probes are not limited to the
axial direction. In fact, radial measurements of journal position are very important during the diagnosis of any piece of mechanical equipment. Machinery axial
measurements are limited to one dimension (forward and reverse). However,
radial or lateral position changes can occur in two dimensions (vertical and horizontal). To accommodate this additional degree of freedom, radial position measurements must be performed in two directions. This requirement drives the
need for two orthogonal, or perpendicular proximity probes for each journal bearing. Orientation of these probes does not matter from a motion detection standpoint, but it is extraordinarily significant for proper determination of the actual
lateral movement, and shaft position change.
Traditional transducer identification often attempts to relate the angular
orientation to a true vertical, or a true horizontal direction. The vertical direction
is often termed the Y-Axis, and the horizontal direction is called the X-Axis. As
shown in the upper left hand shaft A of Fig. 6-16, this is an accurate description
90
45
45
90
45
D
45
Y
Y
X
45
H
45
90
90
45
45
Fig. 616 Radial Proximity Probe Angular Identification
when the vertical probe is located directly above the shaft at 12 oclock, and the
horizontal pickup is on the right side of the shaft at the 3 oclock position. If gap
voltages from these probes are measured with a digital multimeter, the differential changes can be directly converted to plus and minus position changes in the
true vertical and horizontal directions. Similarly, if the probe outputs are wired
directly to a DC coupled oscilloscope, the X-Y movement of the dot on the CRT
will accurately describe changes in radial shaft centerline position.
Unfortunately, proximity probes are seldom located at true vertical and
horizontal locations. In many instances, the probes are offset at 45 from the
262
Chapter-6
vertical centerline as shown in the B shaft diagram in Fig. 6-16. This configuration avoids the horizontal bearing splitline, and it provides top access for oil supply lines, thermocouples, seal oil piping, etc. This configuration has been adopted
as a standard orientation by many organizations such as the American Petroleum Institute1. Fig. 6-16 also displays six additional potential configurations for
radial proximity probes (diagrams C through H). These sketches were generated
by indexing the original probe configuration by 45 increments around the shaft,
and maintaining a fixed relationship between the X-Axis and Y-Axis probes.
Obviously, the orthogonal probes may be installed at any angle, but the proper
identification of X and Y is mandatory for correct physical interpretation of the
resultant transducer data.
In many cases, the probes may be installed above and below the horizontal
splitline. This is common on industrial gear boxes where the transducers are
mounted towards the outside of the box. Thus, the bull gear probes may be
mounted at 45 from the horizontal splitline on one side of the box. The pinion
probes are mounted at 45 from the opposite horizontal splitline on the other
side of the gear box. This type of variation in transducer mounting locations is
often necessitated by the physical construction of the machine.
An example of this type of gear box proximity probe installation is shown in
Fig. 6-17. This is a down mesh, speed reducing, double helical gear box. The
input pinion shows the Y-Axis probes mounted above the horizontal splitline;
whereas the output bull gear displays the X-Axis probes mounted above the horizontal splitline. It has been argued that this difference in probe orientation is
due to the difference in rotation direction between the two shafts. In fact, this is
8X
7X
45
45
45
45
CW
8Y
7Y
Bull Gear
157 Teeth
5,235 Lbs.
T7a
T7b
Output
Double
Helical
GEAR BOX
Input
Fig. 617 Radial And
Thrust Position Proximity
Probes Mounted On A
Speed Decreasing Double Helical Gear Box
5Y
CCW
45
45
5X
Pinion
37 Teeth
733 Lbs.
6Y
45
45
6X
1 Vibration, Axial Position, and Bearing Temperature Monitoring Systems API Standard
670, Third Edition, American Petroleum Institute, (Washington, D.C.:American Petroleum Institute, November 1993).
263
not the case, and the direction of shaft rotation has nothing to do with correct
identification of radial probe angular orientation.
In all cases, the diagnostician must maintain a consistent transducer orientation that will interface with traditional analog instruments, and provide
proper signal polarity (i.e., direction). Since most computer-based instruments
and analysis systems are patterned after their analog predecessors, the same
transducer orientation rules normally apply. The diagram shown in Fig. 6-18
depicts the standard probe orientation scheme that will work under all condiVertical
Y-Axis
Left
Fig. 618 Identification

Procedure For All Radial
Vibration Transducers
Horizontal
X-Axis
Right
View from the Machinery Drive End
tions for clockwise and counterclockwise shaft rotation. First, the individual
should assume a viewing position from the drive end of the machinery train.
Next, the individual should view the probes from the center of the shaft. Then,
when looking directly between the two radial probes (i.e., put your nose between
the probes), the left hand probe will always be the Vertical or the Y-Axis. The
right hand probe will always be the Horizontal, or the X-Axis transducer.
Regardless of the angular position of the orthogonal probes this simple procedure provides the correct identification. This technique was used to identify
transducers within the array of eight different probe orientations in Fig. 6-16. In
addition, the gear box probes shown in Fig. 6-17 were correctly identified using
this same method. For consistency, this technique should also be used for the
identification of casing transducers such as velocity coils and accelerometers.
In practice, the Vertical, Y-Axis, Left Hand transducer is always connected
to the Vertical or Y-Axis deflection of analytical instrumentation such as an oscilloscope. Similarly, the Horizontal, X-Axis, Right hand probe is connected to the
Horizontal or X-Axis of the oscilloscope.
In order to observe the final data with respect to a true vertical and horizontal coordinate system, the data must be rotated by the angular offset of the
probes from the desired coordinates. This can be accomplished by an angular
rotation of the graphical display. The amount of rotation will be equal to the
angle between the true vertical centerline, and the centerline for the Y-Axis
probes. For instance, in Fig. 6-18, the Y-Axis probe is 45 counterclockwise from
the true vertical centerline. Data from this probe arrangement would be rotated
45 counterclockwise to provide the correct physical viewing of the data.
264
Chapter-6
sc
illo
In
pu
t
sc
u
Tr
Probe X-Axis
ta
on
iz
or t)
H gh
e (Ri
l
ca
lo
rY
True Vertical
(Up)
ob
e
XAx
is
ob
YA
xi
iz
on
ta
lo
rX
In
pu
Pr
Pr
op
or
Ve
rti
op
e
illo
sc
sc
O
Oscilloscope Horizontal or X Input
Tr
u
e
(U Ver
p) tic
al
Probe Y-Axis
Oscilloscope Vertical or Y Input
True Horizontal
(Right)
Fig. 619 Angular Vibration Transducer Position And Oscilloscope Signal Orientation
This concept may not be intuitively obvious, and an additional perspective

is provided in Fig. 6-19. In this diagram, the left sketch depicts the normal orientation of an oscilloscope screen. As shown, the Y-Axis transducer is connected to
the scope vertical voltage amplifier. Variations in voltage at this input will result
in vertical deflections of the trace. If the scope is operated in an X-Y mode, and
the X-Axis transducer is directed to the horizontal amplifier, any voltage change
will produce horizontal deflections of the trace. Furthermore, assume that the
probes are physically oriented at 45 from true vertical as depicted in Fig. 6-18.
In this orientation, it is clear that true vertical is located halfway between the
probes. Obviously, in the left sketch of Fig. 6-19, the true vertical (up) direction
must also be halfway between the oscilloscope vertical and the horizontal axis.
By simple deduction, it is also reasonable to conclude that the true horizontal
(right) direction is an axis that points down towards the lower right hand corner
of the grid.
If the oscilloscope is physically rotated 45 counterclockwise, and the Vertical or Y-Axis of the scope is placed coincident with the Y-Axis probe, the sketch
shown at the right side of Fig. 6-19 will evolve. In this arrangement, the principal oscilloscope axes line up with the respective transducers, and a proper Up
Down and LeftRight view is obtained. Although this might seem like a trivial
exercise, it is vitally important to maintain proper direction and orientation of
all of the machinery response measurements. This applies to dynamic motion
(i.e., vibration) as well as static shaft centerline position data.
Proper manipulation of static data acquired with a Digital Multimeter
(voltmeter) requires that the installed probe angle be included within the calculations. For instance, consider a bearing with proximity probes installed at 45
from the true vertical centerline as shown in Fig. 6-20. Further assume that the
probe gap voltages were obtained when the machine was stopped, and again at
265
90
Shaft Centerline
Shift Vector
135
4.
ils
ils
2
3.
45
5.44 Mil
s@
81
45
13
180
Fig. 620 Vector Calculation Of Radial Shaft Centerline Position Change
normal operating speed. If the left hand Y-Axis probe displayed a cold gap at stop
of -9.58 volts DC, and -8.94 volts DC at full speed, the overall change was +0.64
volts toward the probe. Similarly, if the right hand, X-Axis probe had a cold gap
of -9.44 volts DC, and a hot running gap of -8.56 volts DC, the change would be
+0.88 volts toward the probe. The positive sign associated with both of the differential voltages indicates that shaft displacement was towards the probes. This is
also evident by the fact that the full speed gap voltages decreased from the zero
speed values indicating that the shaft moved closer to both of the probes.
Based upon a sensitivity of 200 millivolts per Mil (or 5.0 Mils per Volt), the
Y-Axis distance change is calculated by: 0.64 Volts x 5.0 Mils/Volt = 3.2 Mils. In
the orthogonal X-Axis, the distance change with respect to the probe is determined in the same manner: 0.88 Volts x 5.0 Mils/Volt = 4.4 Mils. These calculated
X-Y changes in shaft position are individual vector quantities where the angles
are governed by the physical orientation of the probes. If the true right horizontal axis is designated as 0, and the top vertical axis is identified as 90; then the
X probe would be located at 45, and the Y transducer would be positioned at an
angle of 135. Combining these angles with the previously calculated magnitudes, the following shift vectors with respect to each probe may be defined:
Y Probe Vector = Y = A = 3.2 Mils 135
X Probe Vector = X = B = 4.4 Mils 45
These two vector quantities may be summed to determine the centerline
shift of the shaft. This vector summation may be performed on a pocket calculator suitable for vector math (e.g., HP-48SX), or the necessary result may be
266
Chapter-6
obtained by using the vector addition equation structure previously discussed in

chapter 2 of this text. This vector summation may be performed in the true horizontal direction by using equation 2-31 as follows:
Horiz add = A cos + B cos
Horiz add = 3.2 cos 135 + 4.4 cos 45
Horiz add = 3.2 ( 0.707 ) + 4.4 0.707 = ( 2.263 ) + 3.111 = 0.848 Mils
Similarly, a vector summation may be performed in the true vertical direction by applying equation 2-32 in the following manner:
Vert add = A sin + B sin
Vert add = 3.2 sin 135 + 4.4 sin 45
Vert add = 3.2 0.707 + 4.4 0.707 = 2.263 + 3.111 = 5.374 Mils
The true vertical and horizontal coordinates may now be plotted on graph
paper, or they may be converted to polar coordinates. The magnitude of the centerline vector shift is calculated from equation 2-33 as:
Shift add =
Shift add =
( Horiz add ) + ( Vert add )

2
( 0.848 ) + ( 5.374 ) =
29.599 = 5.44 Mils
The vector angle associated with this magnitude shift may now be computed with the arctangent equation 2-34 from chapter 2 as follows:
Vert add
add = atan --------------------------
Horiz add
5.374
add = atan ------------- = atan { 6.337 } = 81
0.848
Clearly, the changes in X-Y proximity probe gap voltages allows the determination of a shaft centerline shift. This information may be presented as a Cartesian coordinate position of 0.848 Mils horizontally to the right, and 5.374 Mils
vertically upward. This same change in position may be expressed as a vector
quantity of 5.44 Mils at an angle of 81. The vector summation may also be performed graphically as shown in Fig. 6-20. This diagram also describes the physical representation of these shaft centerline shift calculations. It is apparent that
the final results are identical no matter what calculation or plotting technique is
applied. The diagnostician should apply the most appropriate method based
upon the accuracy required, and the resources available for the calculation.
Knowledge of the radial shaft centerline position is an important diagnostic
tool. Both the magnitude of the centerline shift and the associated angle are significant in evaluation of the machinery behavior. For instance, the previous
267
example revealed a shift vector of 5.44 Mils at 81. If this occurred on a machine
with a load-on-pad tilt pad bearing, and a 15.0 Mil diametrical clearance, the
results would be indicative of normal operating position. This type of bearing
usually exhibits a steep attitude angle indicative of a vertical rise of the shaft
from the bottom pad. The magnitude of the shift is also reasonable with respect
to the diametrical clearance for this type of radial bearing.
However, the same shift vector of 5.44 Mils at 81 would be quite worrisome
for a machine equipped with plain sleeve bearings and a 10.0 Mil diametrical
clearance. With this mechanical configuration the journal would be positioned
close to the center of the bearing, and there would be a strong potential for instability of the machine. This centered rotation of a cylinder within a cylinder (i.e.,
shaft within a sleeve bearing) will be discussed in further detail in chapter 9.
Up until this point, the discussion has centered around the static measurements that may be obtained with shaft sensing displacement proximity probes.
These transducers also have the capability to detect vibratory motion of the
observed surface. In fact, the majority of the industrial applications for these
probes are based upon their ability to accurately measure relative shaft vibration. The easiest way to understand the transducer operation when observing a
vibrating surface is to consider the behavior around the calibration curve as
depicted in Fig. 6-21.
In this diagram, the distance between the probe tip and the target material
is oscillating in a uniform fashion. As stated in chapter 2, this type of repetitive
movement is referred to as Simple Harmonic Motion (SHM), and it may be conveniently described with a sine or cosine wave. This oscillating motion is transProbe Calibration Curve
Proximitor Output Signal (Volts)
Output AC Voltage
Proportional to Vibration
Fig. 621 Proximity Probe

Voltage Output Signal
With An Oscillating
Observed Surface
Output DC Gap Voltage

Proportional to
Average Position
Mechanical
Oscillation
or Vibration
Average Static Position
268
Chapter-6
lated by the probe calibration curve into an oscillating voltage. More specifically,
this is commonly referred to as an alternating, or an AC voltage.
The time domain characteristics of both curves are identical. The maximum
voltage is coincident with the peak gap distance. Similarly, the minimum output
voltage matches the point of the closest gap. The average distance between the
probe and the target is referred to as the Average Shaft Position, and this is
directly measured as the previously discussed DC Gap Voltage. Hence, the Proximitor signal output consists of two interrelated parameters, the AC voltage
proportional to vibration, and the DC voltage that indicates the average distance
of the oscillating target with respect to the stationary probe.
Both the vibration and position characteristics use the probe calibration
curve to translate the mechanical motion into a voltage output. For a standard
200 mv/Mil system, the gap voltage will change by 1 Volt for every 5 Mils of distance change within the linear portion of the curve. The oscillating or vibratory
movement will be converted into an AC voltage signal at the transducer sensitivity of 200 mv/Mil (i.e., curve slope). Since the fundamental displacement measurement is expressed in terms of peak to peak motion, the vibration units will
likewise be in Mils,p-p.
In most instances, vibration measurements are based upon peak detection,
and conversion via the appropriate scale factor. For instance, if the previous diagram displayed a top peak of the Sine wave at -10.0 volts DC, and a bottom peak
of -8.0 volts DC, the vibration may be determined by:
Vibration peak
Vibration peak
to peak
to peak
= ( 10.0 8.0 Volts ) ( 5.0 Mils/Volt )
= 2.0 Volts, p-p 5.0 Mils/Volt = 10.0 Mils, p-p
The same result can be achieved if the peak to peak voltage amplitude is
extracted from an oscilloscope and multiplied by the scale factor in Mils/Volt, or
divided by the scale factor in Volts/Mil. Another way to determine the vibration
amplitude is to convert the scope voltage sensitivity to the transducer sensitivity.
For instance, if the proximity probe calibration is 200 mv/Mil, or 0.2 Volts/Mil,
the oscilloscope voltage amplifier may be set at 200 mv/Division or more commonly 0.2 Volts/Division. Dividing the scope amplifier setting by the probe scale
factor yields a conversion factor of 1.0 Mil per Division. Thus, if the resultant
vibration signal covers six vertical divisions on the oscilloscope, the vibration is
determined to be 6.0 Mils,p-p by a direct visual observation.
For a pure sine wave, the conversion equation 6-3 may be applied to compute the peak to peak magnitude of a 0.4245 Volts,rms value obtained from a Digital Multimeter as follows:
Voltage peak
Vibration peak
Voltage rms
0.4245
= -------------------------------- = ---------------- = 1.20 Volts p-p
0.3536
0.3536
peak = 1.20 Volts p-p 5.0 Mils/Volt = 6.0 Mils, p-p
to peak
to
Again the same results are obtained, and it is clear that conversion between
voltage and vibration amplitudes may be accomplished in different, but consis-
269
tent ways. It is also self-evident that proximity probes may be used to measure
average position, and vibration in both radial and the axial directions. This is
necessary from a machinery analysis standpoint, since real process machinery
does translate and vibrate in the lateral and axial directions, and the transducer
suite must be able detect this overall movement.
Another application of proximity probes resides in the realm of providing
timing signals. Typically, these are synchronous, once per revolution pulses that
may be used for accurate speed measurements. They are also employed for phase
measurements and determination of precession when combined with other
probes on the machinery train. These 1X timing pulses are typically referred to
as Keyphasor signals, and the transducers are called Keyphasor probes.
These proximity probe timing sensors are often positioned over a shaft notch or
drilled hole. With this arrangement, the probes produce a negative going pulse
as the shaft indentation passes beneath the timing probe (e.g., Fig. 6-5).
In some instances, a raised surface such as the top of a shaft key is
observed by the Keyphasor probe. In these cases a positive going pulse is generated (e.g., Fig. 6-6). The shape of these timing pulses, and the instrumentation
trigger points are discussed in greater detail earlier in this chapter, and also at
the beginning of chapter 11 on balancing.
The radial, axial, and Keyphasor probes are combined in many different
combinations. In most instances, a pair of mutually perpendicular radial probes
are installed at each journal bearing (X-Y probes). A pair of axial probes are typically mounted at each thrust bearing, and a radial Keyphasor probe will usu-
Casing
Keyphasor
Probe
Keyphasor Probe
X-Y Radial
Probes
Y-Axis
Radial
Probe
X-Axis
Radial
Probe
End Cover
Thrust Probes
Pinion
Extension
Existing
Pinion Shaft
K Notch
Thrust
Probes
Pinion
Bearing
K Notch
Bearing
Casing Bore
Casing
Fig. 622 Side View Of Proximity Probe

Installation On A High Speed Pinion
Fig. 623 End View Of Proximity Probe

Installation On A High Speed Pinion
270
Chapter-6
ally be installed for each shaft speed. A typical installation that involves all three
basic applications for proximity probes is presented in Figs. 6-22 and 6-23.
The machine under consideration is a speed increasing, double helical gear
box. The unit is motor driven at the bull gear input, and the pinion output is coupled to a centrifugal compressor. Figs. 6-22 and 6-23 describe the probe installation at the outboard, or blind end of the pinion. Installation of probes on this unit
was hampered by a short shaft that extended only 1/8 past the radial bearing.
To provide measurable surfaces, shaft extensions for the bull gear and the pinion
were fabricated from 4140. These extensions were threaded into the respective
gear elements, and locked into position with a pair of countersunk cap screws.
The pinion drawings describe the location of the two thrust probes, and the
surface observed by these axial transducers. Typically, the probes should be
spaced a minimum of two probe diameters from any potential source of interference. In this case, the axial probes had to be located to avoid the Keyphasor
notch, the center punch on the shaft, as well as the axial cap screws. The radial
vibration probes were positioned to have a clear view of the pinion extension,
plus they must not be influenced by the timing notch milled into the outboard
end of the extension. Thus, there is no chance for cross talk or interference with
the thrust probes or the Keyphasor. Although the side view places the X-Y
probes fairly close to the timing probe, the end view in Fig. 6-23 reveals the 45
angular offset between each of the radial probes and the Keyphasor. Hence, the
radial probes have minimal potential for signal interference.
In virtually all cases, the Keyphasor probe should be oriented in a radial
direction to maintain a uniform pulse shape and size during all operating conditions and speeds. Placement of the Keyphasor transducer in an axial direction
will often jeopardize the consistency of the timing pulse. This is due to variations
in thrust position that manifest as substantial changes in the pulse signal. The
observed notch (or projection) should normally be 40 to 60 Mils deep (40 to 60
Mils high for projection), and the width should be one and a half (1.5) times the
probe diameter. For instance, a 1/4 diameter Keyphasor probe would produce
an acceptable pulse signal with a 3/8 wide slot that is 50 Mils deep (or 50 Mils
high for a projection). In the final output, the Keyphasor signal should have a
pulse height between 5 and 15 volts.
The necessity for a strong and consistent Keyphasor signal cannot be
overstated. Many mechanical malfunctions and the entire array of normal balancing activities are totally dependent upon a good once-per-revolution trigger
signal. Furthermore, most dynamic data acquisition systems and analytical
instruments require a rotational speed trigger signal to allow full utilization of
the capabilities of the instrumentation. Since the availability of this timing signal is so important to the business of machinery diagnostics, the manipulation
and proper utilization of this signal will be discussed in further detail in subsequent chapters 7, 8, and 11.
In the overview, the proximity probe transducer suite is applicable to a wide
range of process machinery. Due to the ability to measure relative position
changes as well as relative vibration measurements, the potential industrial
applications are substantial. As with any signal transducer, proximity probes
271
exhibit a variety of advantages, but they also have a series of disadvantages. For
purposes of comparison with other vibration transducers, the following two summaries of proximity probe features are presented for consideration:
Proximity Probe Advantages
Measures Shaft Dynamic Motion

Measures Shaft Static Position
Excellent Signal Response Between DC and 90,000 CPM (1.5 kHz)
Flat Phase Response Throughout Transducer Operating Range
Simple Calibration
Solid State Electronics
Rugged and Reliable Construction
Available in Many Physical Configurations
Suitable for Installation in Harsh Environments
Multiple Machinery Applications for the Same Transducers
Proximity Probe Disadvantages
Sensitive to Surface Imperfections and Magnetism

Sensitive to Material Properties
Shaft Surface must be Conductive
Low Dynamic Signal Response Above 90,000 CPM (1.5 kHz)
External Power Source Required
Correct Probe to Proximitor Cable Impedance Must Be Maintained
Minor Temperature Sensitivity in pre-1990 Probes
Sensitive to Interference from Adjacent Proximity Probes
Sensitive to Probe Mounting Bracket Resonance(s)
Potentially Difficult to Install
In the majority of applications, proximity probes are used for permanent

monitoring and machinery protection measurements on units with fluid film
bearings. In many process plants this would be the critical and essential machinery trains. The probes would typically be connected to dedicated monitors for
functions such as radial position and vibration, axial position and vibration,
rotational speed, differential expansion, and eccentricity. The monitors provide
power to the Proximitors. These monitors also provide additional signal conditioning, various display functions, plus dedicated alarm capabilities. Furthermore, many monitoring systems are commonly interfaced with Distributed
Control Systems (DCS), and dedicated dynamic data acquisition and processing
systems such as a Transient Data Manager system.
272
Chapter-6
VELOCITY COILS
Velocity transducers represent one of the earliest forms of vibration probes.
References from the 1950s often refer to these transducers as vibrometers. They
are used to obtain absolute velocity measurements of stationary machinery elements. These are fully contacting probes that are mounted directly on a mechanical structure (e.g., bearing housing); and they measure the dynamic motion of
that structure. Velocity coils are supplied in a several basic configurations. For
example, the photograph in Fig. 6-24 depicts a typical industrial velocity pickup.
Output Cable to
Instrumentation
Spring
Magnet
Coil
Case
Mounting Stud
Fig. 624 Typical Industrial Velocity Coil
Fig. 625 Velocity Coil Cross Section
This type of vibration transducer generally has an output sensitivity of 500

millivolts/IPS. The usable frequency range extends from a bottom end of 600 to
900 CPM (10 to 15 Hz), to an upper limit of approximately 90,000 CPM (1,500
Hz). Special velocity coils are built specifically for low frequency measurements;
and although the lower end of the frequency response curve is significantly
improved, the physical size and weight of the transducer grows substantially. For
instance, some low frequency applications use velocity coils that weigh in excess
of 200 pounds.
Regardless of the physical configuration, velocity coils consist of the same
fundamental components. For example, consider the cross section of a typical
velocity probe shown in Fig. 6-25. The transducer casing is directly attached to
the vibrating surface. The method of attachment could be a screwed connection
between the transducer mounting stud and a drilled and tapped hole at the measurement point. Another common approach would be to screw the velocity pickup
into a high strength double bar magnet, and use the magnet for attachment of
the pickup to the vibrating surface. This approach provides good mobility to the
Velocity Coils
273
probe. However, the rigidity of the attachment should always be checked before
believing the resultant vibration data. In all cases, rigid mounting of the probe to
the vibrating surface insures that the transducer will move in unison with the
measurement point. The transducer senses the vibrating surface via a circumferential electrical coil that is attached to the inside of the housing. This coil must
move together with the casing due to the physical attachment between elements.
Hence, the coil motion is presumed to be virtually identical to the vibratory
motion of the attached surface.
Located within the center of the electrical coil is a permanent magnet
mounted on very soft springs. This spring supported magnet is confined to oscillate in the principal axis of the transducer (e.g., up and down for a vertical
probe). The combination of a heavy magnet (mass), plus a soft spring yields a low
natural resonant frequency for the assembly. In actual operation, the transducer
case and coil vibrate in sympathy with the attached surface, and the spring
mounted magnet tends to remain stationary. This relative motion between the
essentially stationary magnet, and the vibrating coil results in the generation of
a coil voltage that is proportional to the velocity of the transducer outer casing.
Since this type of vibration probe produces a signal without the necessity of
an external power source, the transducer is considered to be self-generating. This
feature simplifies the field installation of a velocity transducer system. For
instance, a velocity probe is commonly referred to as a Geophone within the field
of seismic testing. This category of testing is applied to seismic measurements for
earthquake detection, and petroleum exploration. In these applications a selfgenerating probe is highly desirable.
The natural frequency of the spring supported magnet typically falls in the
range of 300 to 600 CPM (5 to 10 Hz). The severity of this fundamental system
resonance might easily dominate the resultant output signal, and render the
data unusable. In order to restrain this resonant spring mass response, the outFrequency (Cycles/Minute)
300
600
6,000
60,000
Fig. 626 Typical Velocity

Coil Frequency Response
Characteristics
30,0000
180
500
160
Transducer Sensitivity
450
140
400
120
350
100
300
80
250
60
200
40
150
20
Phase Error
100
50
5
10
100
Frequency (Hertz)
1,000
-20
5,000
Phase Error (Degrees)
Velocity Coil Sensitivity (millivolts/IPS)
550
274
Chapter-6
put is either damped electronically, or the mechanical system is over damped

with a viscous fluid surrounding the magnet. The damped response of the spring
mass system is subject to the same physical laws that were previously discussed
in chapter 2. Since the transducer system is over damped, the natural resonance
is fully suppressed as shown in the calibration plot of Fig. 6-26.
The characteristics displayed in this velocity coil calibration plot are for a
600 CPM transducer. By definition, this transducer should display an attenuation in output sensitivity of -3 dB at 600 CPM. In fact, this probe does exhibit 354
mv/IPS (= 500 x 0.707) at the resonant frequency of 600 CPM (10 Hz). In addition, it should be noted that damping not only suppresses amplitude sensitivity,
it also influences the output signal phase. This results in a 90 phase error at 600
CPM. Even at 6,000 CPM, the phase error is still 10. Thus, the diagnostician
should be cautious during any low frequency applications of velocity pickups.
When accurate amplitude and phase characteristics are required for these transducers, each pickup must be individually calibrated to determine the specific
characteristics for each velocity coil.
Full calibration of a velocity transducer requires the integration of several
instruments as depicted in Fig. 6-27. The dual channel Dynamic Signal Analyzer
(DSA) is used to generate a white noise output that attempts to simultaneously
produce all frequencies within the analysis bandwidth. Another way to visualize
a white noise source is to consider a signal that contains a uniform distribution
of energy across the analysis bandwidth. In most instances, this type of excitation allows a faster calibration when compared against a single frequency oscillator that is used to sweep the transducer operating range.
Test Probe Output
Dual Channel
Dynamic
Signal
Analyzer
Fig. 627 Typical Instrumentation Required For

Proper Calibration of Seismic Vibration Transducers
White Noise Excitation
Reference Probe
Output
Shaker
Table
Power Amplifier
The white noise signal is used as an input to the Power Amplifier. This
amplifier actually drives the vertically oscillating Shaker Table in accordance
with the frequency content of the excitation signal. Since the shaker table is a
mechanical device, the upper frequency will be governed by the design of the particular table. However, for the calibration of velocity coils, the testing devices will
generally have the capability to exceed the frequency response characteristics of
the test object (i.e., the velocity coil).
A calibrated reference transducer is mounted on the shaker table next to
the test probe. This reference transducer is usually an accelerometer that has a
flat amplitude and frequency response across the bandwidth of the velocity
Velocity Coils
275
pickup. Output signals from the reference transducer, and the probe under test
are directed back to the two channel DSA. At this time, a frequency response
function (also known as a transfer function) is performed between the two signals (test probe to reference probe). The resultant amplitude and phase characteristics versus frequency are documented in a plot similar to Fig. 6-26.
General characteristics for any particular type of velocity transducers will
govern the average results of this type of transducer testing. In most cases, deviations between transducers will be quite apparent. Hence, for any critical work
with velocity coils (especially low frequency), this type of detailed response testing is considered to be mandatory.
Since velocity coils contain internal springs and moving parts, they are subject to fatigue failures as a function of time. Hot service in a high vibration environment will shorten the life span of these elements. Conversely, moderate
operating temperatures combined with low vibratory surfaces will allow the
longest transducer life. In many cases, the condition of the velocity coils and
their suitability for service may be determined with a routine calibration check.
Generally, significant deviations from the initial calibration curves would be
grounds for refurbishing (i.e., rebuilding) or replacing the velocity pickup.
The velocity coil calibration curve and the discussed calibration check both
presume that the transducer is securely bolted to the vibrating surface. In many
instances, magnets or extension stingers are used between the velocity pickup
and the measurement surface. These additional interface devices degrade the
performance of the probe, and they may result in mounting resonances between
500 and 3,000 Hz (30,000 and 1800,000 CPM). Thus, the type of mounting fixture
should always be considered. If possible, the velocity coil mounting device (e.g.,
magnet) should be incorporated into the calibration setup to determine the frequency response characteristics of the entire transducer system.
Based upon the calibration curve and the fundamental operation of the
velocity pickup, it is clear that position measurements (i.e., DC or zero frequency) cannot be made with velocity coils. In all applications the orientation
rules, and the interaction with the Keyphasor timing probes are identical to
the behavior previously discussed with the proximity probes.
It should also be restated that velocity amplitudes are zero to peak values,
whereas displacement measurements are expressed as peak to peak amplitudes.
Velocity signals may be integrated with respect to time to obtain displacement,
and they may also be differentiated with time to obtain acceleration. These
manipulations of the raw velocity signals may be performed with electronic circuits dedicated to these integration or differentiation functions; or they may be
performed more effectively in a digital format within a DSA.
For velocity amplitudes at a given frequency, the equivalent displacement
or acceleration amplitudes may be computed with equations 2-17 through 2-22
from chapter 2. In many cases, the velocity amplitudes at a single frequency are
defined as vector quantities. The conversion equations allow the computation of
proper amplitudes, but the diagnostician must convert the vector phase angles
by 90. The angle conversions are performed in accordance with equations 2-14
276
Chapter-6
through 2-16. For velocity measurements, the angle conversions are as follows:
Phase displacement = Phase velocity + 90
Phase acceleration = Phase velocity 90
Conversion of vibration units can be quite useful. For instance, if a shaft
sensing proximity probe is attached to a bearing housing, and a velocity coil is
mounted directly in line with the proximity probe, a variety of measurements are
possible. The proximity probe can measure position and vibration of the shaft
with respect to the housing. The casing mounted velocity coil can measure absolute vibration of the bearing housing. A summation of shaft plus casing vibration
will result in shaft absolute motion. Fig. 6-28 depicts this type of dual transducer
installation on a machine with a fluid film bearing.
Casing
Signal
Velocity
Probe
Shaft
Signal
Proximity
Probe
Shaft
Bearing
Fig. 628 Simultaneous

Measurement Of Absolute
Casing Velocity And Relative Shaft Displacement
Bearing Housing
In this type of arrangement, the casing velocity data may be integrated to

casing displacement, and the resultant signal electronically summed with the
proximity probe output. This type of signal manipulation is performed with a
voltage summing amplifier. A differential voltage amplifier may also be used, but
one of the signals must be inverted to obtain the proper result. The complex
summed signal from the amplifier may then be processed in a variety of manners
with additional diagnostic instrumentation.
Alternately, the velocity signal may be digitally integrated to displacement
in one channel of a DSA. If the shaft displacement signal is connected to another
channel of the DSA, the integrated velocity may be added to the shaft signal
Velocity Coils
277
within the DSA. This summed signal may then be viewed as time or FFT data.
Another approach that is quite convenient for signals that are dominated
by a single frequency component is a direct vector summation of the output from
each transducer. For example, consider the following calculations for 1X vectors
emitted by a 3,600 RPM synchronous motor:
Shaft Relative Displacement = 2.85 Mils, p-p 165
Casing Absolute Velocity = 0.24 IPS, o-p 31
Converting the casing velocity amplitude into the equivalent casing displacement amplitude may be accomplished with equation (2-20) as follows:
19, 099 0.24
19, 099 V
Casing Absolute Displacement = ---------------------------- = ---------------------------------- = 1.27 Mils, p-p
Rpm
3, 600
The phase of the casing displacement vector is determined from the velocity
phase angle using equation (2-14) as follows:
Case Phase displ = Case Phase velocity + 90 = 31 + 90 = 121
These two conversions define the casing displacement vector, which may
now be added to the relative shaft displacement vector to determine the shaft
absolute vector. Using the equations (2-31) to (2-34) for vector addition, the summation of these two running speed vectors yields the following:
Shaft Absolute Vector = 2.85 165 + 1.27 121 = 3.86 Mils, p-p 152
This is an important solution for machines with flexible supports. In many
cases, neither the shaft relative nor the casing absolute vibration vectors are
fully descriptive of the overall shaft motion. For these types of machines, the
shaft absolute measurement may be mandatory. Furthermore, during vibration
analysis of machinery with flexible supports, the relationship of shaft to casing
to absolute motion is often a key ingredient in the determination of the mechanical element(s) responsible for the vibration problem. It should also be mentioned
that support flexibility might occur in the bearing housing assembly, or the soft
members might be the machine foundation or even the subsurface soil.
As shown in the previous example, the absolute motion was computed from
the transducer outputs for simple single frequency signals. Alternatively, the
information may be obtained by electronic summation of the signals, followed by
synchronous filtration. In either method, the shaft displacement signal should be
corrected for shaft runout to obtain the best possible representation of the absolute shaft motion. For complex transducer signals containing multiple frequencies, the shaft plus casing data must be handled by electronic summation, plus
further data processing such as FFT analysis.
Due to the self-generating signal characteristics, and the general ease of
installation of these probes on different mechanical elements, the velocity coil
has been a popular transducer. As with any signal transducer, velocity coils
exhibit several advantages, but they also have a series of disadvantages. For pur-
278
Chapter-6
poses of comparison with other vibration transducers, the following two summaries of velocity coil features are presented for consideration:
Velocity Coil Advantages
Measures Casing Absolute Motion

Easily Attached to Machinery Externals, Piping, Baseplates or Structures
Good Signal Response Between 900 and 90,000 CPM (15 and 1,500 Hz)
Self-Generation Signal Electronics
No Special Wiring Required
Available in Several Configurations
Velocity Coil Disadvantages
Sensitive to Mounting Fixture and Transducer Orientation

Unable to Measure Shaft Vibration or Position
Difficult Calibration Check
Poor Signal Response Below 900 CPM (15 Hz) Above 90,000 CPM (1.5 KHz)
Amplitude and Phase Errors at Frequencies below 1,800 CPM (30 Hz)
Operates Above Transducer Natural Frequency of 600 CPM (10 Hz)
Potential for Failure due to Fatigue of Moving Internal Parts
Temperature Sensitive, Typical Upper Limit of 250F, Lower Limit of 30F
Difficult to Install in Cramped Areas
During recent years the cost of velocity coils has increased, and the price of
accelerometers has substantially decreased. This cost change, combined with the
previously listed disadvantages, has eliminated the use of velocity coils in many
locations. For applications where velocity remains the preferable measurement,
the traditional mechanical velocity coil has been replaced by an accelerometer
with integral electronics to provide a velocity output from the accelerometer.
PIEZOELECTRIC ACCELEROMETERS
Accelerometers are versatile vibration transducers for absolute measurement of stationary machinery elements or structures. These devices are fully
contacting probes that are mounted directly on a mechanical element (e.g., bearing housing). They are available in a variety of configurations, and they may be
designed to cover a wide range of operating and environmental conditions. For
example, the photograph in Fig. 6-29 depicts four different sizes and configurations of standard PCB Piezotronics high frequency ICP accelerometers.
The ICP designation is a registered trademark of PCB Piezotronics, Inc.,
and it stands for Integrated Circuit Piezoelectric. This type of probe contains
much of the necessary signal conditioning electronics within the body of the
transducer. The fundamental flexibility of the basic transduction scheme allows
Piezoelectric Accelerometers
279
Fig. 629 Typical Configurations Of Accelerometers

Manufactured By PCB
Piezotronics, Inc.
significant interchangeability of transducers with common external power supplies. In addition, many instruments (e.g., HP-35670A) contain a constant current power supply that allows the DSA to power the accelerometer directly.
Although the four accelerometers shown in the Fig. 6-29 may be driven by
the same power supply, the characteristics of the accelerometers are significantly
different. For comparative purposes, the accelerometer shown in the upper left
hand portion of Fig. 6-29 weighs 60 grams. This pickup displays a 5% operating
range from 60 to 180,000 CPM (1 to 3,000 Hz). It has a mounted natural frequency of nominally 1,680,000 CPM (28.0 kHz), and an output sensitivity of 100
millivolts per G. Pickups of this type are generally used for measurement of low
to medium frequency behavior.
In the upper right hand corner of Fig. 6-29, the displayed accelerometer is
smaller with a total weight of 25 grams. It has an extended operating range
(5%) of 60 to 600,000 CPM (1 to 10,000 Hz). This transducer has a mounted resonant frequency in the vicinity of 2,100,000 CPM (35.0 kHz), and an output sensitivity of 10 millivolts per G. At the bottom of Fig. 6-29, miniature
accelerometers are shown. These devices weight 2 grams and 1 gram respectively. Due to the small size and weight of these accelerometers the cables are
integral with the transducers. The probe in the lower left hand corner has a 5%
operating range between 60 and 1,200,000 CPM (1 and 20,000 Hz). The mounted
resonant frequency resides in the vicinity of 4,620,000 CPM (77.0 kHz), and the
output scale factor is 10 millivolts per G. The smallest transducer shown in the
lower right hand corner has a 10% operating range of 180 to 1,800,000 CPM (3
to 30,000 Hz). The mounted resonant frequency for this accelerometer is
7,200,000 CPM (120.0 kHz), and the output scale factor is 5 millivolts per G.
Thus, it is apparent that as accelerometers decrease in size, the transducers exhibit lower scale factors combined with extended frequency response characteristics. These smaller accelerometers are used for measurements on small
mechanical elements such as turbine blades and circuit boards. Conversely,
larger accelerometers have higher output sensitivities plus a reduced frequency
response range. For instance, one particular ICP transducer that is suitable for
low frequency seismic testing has a scale factor of 10,000 millivolts per G. The
280
Chapter-6
5% operating range on this pickup runs from 9 to 6,000 CPM (0.15 to 1,000 Hz),
and the rated output is only 0.5 Gs. If the examination of lower frequencies are
required, then piezoresistive accelerometers may be used, and they allow measurement of frequencies down to DC (zero frequency). These specialized piezoresistive transducers are more commonly used for static plus dynamic pressure
measurements, and additional discussion of this sensing element will be provided in the following section of this chapter.
Although piezoelectric accelerometers are manufactured in a wide variety
of physical configurations to address a large range of applications, the fundamental internal elements for this class of transducer remains fairly consistent.
For instance, consider the cross section of a typical ICP industrial accelerometer as presented in Fig. 6-30. This type of vibration transducer depends on the
Integrated
Circuit
Output Cable
Preload Screw
Seismic Mass
Insulator
Piezoelectric
Crystal
Outer Case
Case Insulator
Fig. 630 Cross Section

Of A Typical Industrial
ICP Accelerometer
Mounting Stud
electromechanical properties of the piezoelectric crystal. Specifically, the crystal

will emit an electrical charge when a mechanical load or stress is applied. Conversely, when an electrical charge is applied to the crystal, it will physically
deform in a direct relationship to the magnitude of charge. This concept is
applied in many electronic devices varying from computers, to communication
equipment, to accelerometers. Although various materials exhibit piezoelectric
properties, the majority of the industrial accelerometers employ either natural
quartz crystals, or man-made polycrystalline ceramics. Each type of material displays specific features, and the final selection of the piezoelectric crystal depends
on the eventual application of the transducer.
The diagram in Fig. 6-30 depicts a compression accelerometer that includes
a stainless steel base that supports the crystal, plus a seismic mass. Full contact
between the mass and crystal is insured by the preload screw that joins the
mass, crystal, and base into an essentially solid structure. When this assembly is
bolted onto a vibrating surface, the seismic mass imposes a definable force upon
281
the crystal. In accordance with the second law of motion by Sir Isaac Newton
(1643 to 1727), force is equal to mass times acceleration. Within an accelerometer, the crystal is subjected to a force from the mass, and the output charge is
proportional to the acceleration. Obviously, small forces will produce low acceleration levels, and large forces will manifest as high acceleration.
The charge output from the crystal in Fig. 6-30 is wired directly to an internal Integrated Circuit (IC). This electronics package provides the necessary signal conditioning to convert the crystal charge signal (picocoulomb/G) to a voltage
sensitive signal (millivolts/G). The presence of this IC within the transducer normally limits the maximum operating temperature to approximately 250F. For
applications at higher temperatures, the signal conditioning electronics may be
located in a separate Charge Converter that is placed in a cooler environment.
Removing this IC from the accelerometer allows the transducer to be designed
for effective operation at elevated temperatures. Standard transducers may be
purchased for operation up to 600F, and custom accelerometers have been built
to withstand temperatures in excess of 1,200F. This type of transducer is commonly referred to as a Charge Mode accelerometer. It is sensitive to cable whip,
and electrical interference of the wiring between accelerometer and Charge Converter. Once the signal is converted to a voltage sensitive signal, the wiring
downstream of the Charge Converter may be safely directed to the measurement
or recording instrumentation with normal coaxial cable.
For most industrial applications, the internal IC plus a piezoelectric crystal
are combined into an ICP transducer that is suitable for use up to 250F. The
accelerometer in Fig. 6-30 represents an upright compression configuration.
Accelerometers are also built in an inverted compression, an isolated compression, a shear mode, and a flexural mode configuration. Each particular model
exhibits different technical characteristics for various measurement applications, combined with a range of transducer prices. As with any technical selection, the individual must arrive at an equitable balance between performance
and cost. From an operational standpoint, the previous explanation remains
generically applicable throughout this suite of transducers.
As previously noted during the accelerometer transducer review, the natural frequency of an accelerometer resides above the operating range. Hence,
accelerometers function as a rigid system, and phase excursions do not appear
within the transducer operating frequency range. In virtually all cases, acceleration phase angles may be converted to displacement angles by adding or subtracting 180. The amplitude response of an accelerometer is somewhat more
complicated due to the low frequency roll off of the transducer, plus the mounting
resonance located well above operating frequency range.
Typical frequency response characteristics of an accelerometer are presented in the calibration plot of Fig. 6-31. In this diagram, the accelerometer output sensitivity in millivolts per G is plotted against frequency. This particular
transducer exhibits the anticipated low frequency attenuation below 900 CPM
(15 Hz), and the mounted resonance at a frequency of about 1,800,000 CPM
(30,000 Hz). Clearly, measurements made in the vicinity of the resonant frequency would be influenced by the amplification associated with this resonance.
282
Chapter-6
The output sensitivity between 1,200 and 600,000 CPM (20 and 10,000 Hz)
remains reasonably flat, and an average scale factor of 100 millivolts per G is
appropriate to apply within this region.
300 600
6,000
60,000
600,000
3,000,000
100
1,000
10,000
50,000
Fig. 631 Typical Frequency

Response Characteristics For
An Industrial Accelerometer
Accelerometer Sensitivity (millivolts/G)
250
200
150
100
50
0
5
10
Frequency (Hertz)
It must be mentioned that accelerometer calibration plots are generally

performed with a stud mounted transducer. If the accelerometer is used in the
field with an alternate mounting technique, the mounted resonant frequency
may drift into the normal operating range due to a reduction in the attachment
stiffness. This reduction in the mounted natural frequency serves to reduce the
effective operating range of the accelerometer. For example, Table 6-2 summarizes the reduction in the maximum operating frequency with various types of
accelerometer mounting techniques:
The direct stud mount presumes a smooth and flat mounting surface on the
structure to be measured that will interface with the machined bottom of the
accelerometer. The direct adhesive mount is typically a super glue type of material that is used to attach small one and two gram accelerometers. This type of
Table 62 Influence Of Mounting Technique On Frequency Response Characteristics
Type of Accelerometer Mount
Maximum Frequency Response
Direct Stud Mount
15,000 to 20,000 Hz
Direct Adhesive Mount
8,000 to 16,000 Hz
Epoxy Glue Base
8,000 to 12,000 Hz
Strong Double Bar Magnet
3,000 to 5,000 Hz
Mechanical Clamp
500 to 1,500 Hz
Hand Held
250 to 1,000 Hz
283
adhesive provides a good solid mount for small accelerometers. At the conclusion
of the test, the accelerometer may be pried or twisted off of the mounting surface.
The epoxy glue bases provide one of the most convenient methods for field
mounting accelerometers. These glue bases are made of anodized aluminum,
with a drilled and tapped center hole to accept the accelerometer mounting
screw. These bases are attached to the measurement surface with a quick setting
(5 to 10 minute cure time) two-part epoxy cement. One surface of the glue base is
finished flat for contacting the accelerometer, and the other side is normally
grooved to accept the epoxy.
Strong double bar magnets may be used to attach accelerometers, but the
overall frequency response is significantly reduced. The use of weak magnets is
discouraged since the frequency response characteristics of these mounts is generally unacceptable. A variety of mechanical clamps or attachments may be used
to mount accelerometers. Again, a further reduction in overall frequency
response should be anticipated. The final category in Table 6-2 of hand held
accelerometers may be used if no other form of attachment is available.
From this summary, it is clear that an accelerometer may be rendered ineffective for high frequency measurements simply due to the method of transducer
attachment. This characteristic may also be used to control the frequency
response of the final measurement. For instance, assume that low frequency
information is required on a machine that emits significant high frequency excitations. In this situation a mechanical isolation pad between the measurement
surface and the accelerometer may be used to eliminate or suppress the high frequency components. Whenever possible, the calibration procedure should employ
the same accelerometer mounting technique that will be used in the field. Within
the low frequency domain, variations in accelerometer mounting techniques will
not cause appreciable differences in the data. However, the diagnostician is
always encouraged to utilize a substantial and rigid mount to eliminate any concerns or potential data corruption associated with a poor transducer attachment.
The ideal accelerometer mounting surface should be both smooth and flat.
Historically, transducer manufacturers have specified surface finishes that are
often unattainable on cast machinery structures. Spot facing of accelerometer
mounting locations in the field, or even machine shop finishing generally results
in surfaces that do not meet the ideal vendor requirements. In recent years,
these stipulations have been somewhat relaxed, and an emphasis has been
placed on using a thin coating of silicone grease or acoustic couplant between the
accelerometer base and the mounting device. This approach allows an improved
contact between the transducer and the mounting surface by allowing the grease
to fill in the voids in both metallic surfaces. The appropriateness of this technique has been verified with shaker table tests, and extensions of the upper frequency limit of 5% to 20% have been documented on several occasions.
In addition to accelerometer sensitivity to the method of attachment, the
machinery diagnostician must also secure the coaxial cables leaving the accelerometer. This is particularly important in high vibratory environments were the
cable motion could provide addition strain to the accelerometer, and result in
false and/or erratic electronic signals. To prevent the occurrence of this potential
284
Chapter-6
signal interference problem, it is generally desirable to tape, clamp, or otherwise

secure the signal cable within 2 to 4 inches from the accelerometer. In addition, a
slight bend or relief of the cable should be maintained.
The same type of shaker table calibration techniques used for velocity coils
may generally be applied for accelerometers (e.g., Fig. 6-27). Since top end frequencies are higher with accelerometers, particular attention must be paid to
the mounting techniques used for the test as well as for the reference accelerometer. Whenever possible, the accelerometer should be attached to the test surface
(e.g., shaker table) in the same manner that it will be mounted in the field. Single point, swept frequency, as well as multi-frequency test systems are used for
accelerometer calibration. Other common techniques such as gravitational drop
tests may also be used for certain types of transducers. In all cases, the downstream electronics should be included in the calibration test to allow documentation of frequency response characteristics for the entire system.
Due to the solid state reliability, and the extended frequency response
range of most accelerometers, the applications are substantial. As with any
transducer, accelerometers exhibit a variety of advantages and disadvantages.
For purposes of comparison with other vibration transducers, the following two
summaries of accelerometer features are presented for consideration:
Accelerometer Advantages
Measures Casing or Structural Absolute Motion

Easily Attached to Machinery, Piping, Baseplates or Structures
Good Signal Response Between 900 and 600,000 CPM (15 >10,000 Hz)
Flat Phase Response Throughout Transducer Operating Range
Solid State Electronics with Rugged and Reliable Construction
Operates Below Mounted Natural Resonant Frequency
Same ICP Signal Conditioning Usable with Various Transducers
Special Units Available for High Temperature Applications (>1,200F)
Available in Many Configurations
Small Transducer Size, Easiest to Install in Cramped Areas
Accelerometer Disadvantages
Sensitive to Mounting Technique and Surface Condition

Unable to Measure Shaft Vibration or Position
Difficult Calibration Check
External Power Source Required
Low Dynamic Signal Response Below 600 CPM (10 Hz)
Transducer Cable Sensitive to Noise, Motion, and Electrical Interference
Temperature Limitation of 250F for ICP Transducers
Extended Frequency Range Often Requires Signal Filtration
Double Integration Often Suffers from Low Frequency Noise
Pressure Pulsation Transducers
285
The characteristics of piezoelectric transducers revolve around the relationship between force, mass, and acceleration. It is reasonable to deduce that any
other physical phenomena that can apply a force or load upon a piezoelectric
crystal may also be measured with this technology. In fact, the following sections
addresses some of these similar dynamic transducers.
PRESSURE PULSATION TRANSDUCERS

One of the most useful devices for analyzing fluid behavior is the dynamic
pressure pulsation transducer, as depicted in the photograph of Fig. 6-32. The
unit at the bottom of the photo is the actual pressure pulsation transducer, and
the assembly shown at the top of the photograph portrays an identical pressure
transducer mounted in a 1/2 NPT bushing.
Fig. 632 Typical Configuration of Pressure Pulsation

Probes Manufactured By
PCB Piezotronics, Inc.,
Plus Installation Pipe Plug
These small transducers contain an Invar diaphragm that is exposed to the

process fluid, and it is connected to the piezoelectric crystal. Application of pressure to the diaphragm causes the crystal to produce a charge that is converted to
a voltage sensitive signal by the internal integrated circuit. For the pressure
transducers depicted in Fig. 6-32, the output sensitivity is equal to 1.0 millivolt
per Psi. These particular transducers are rated for a maximum pressure of
15,000 Psi, and the transducer operating temperature range generally extends
between -100F and +275F. These pressure probes exhibit a large frequency
range, with a high resonant frequency of 28,500,000 CPM (475,000 Hz).
Other configurations of static and dynamic pressure pulsation transducers
are available. There are entire families of transducers dedicated specifically for
high frequency measurements of shock waves, ballistic phenomena, engine combustion, acoustics, explosive blasts, and machinery related pressure pulsations.
Although the physical transducer configurations and performance do vary, the
fundamental transducer concepts remain constant.
Although the pressure pulsation transducer technology is fairly straightforward, the application and field utilization of these probes can become quite complicated. In most instances, the complexity is associated with the mounting
location of the transducer versus the fluid stream to be investigated. Ideally, the
286
Chapter-6
pressure probe would be installed within the fluid to obtain the best possible
indication of pressure fluctuations. However, this mounting location is often
physically impossible to achieve. The next thought might be to mount the pressure probe flush with the pipe inner diameter. This sounds plausible, but in
many instances the desired dynamic data is masked by the fluid boundary layer
effects at the pipe wall.
In many applications, the pressure transducer is simply attached to the
atmospheric side of an available vent or drain valve. This type of location is physically accessible, but the fluid pressure variation is now attenuated by the intermediate nipples, fittings, and valve(s). In addition, the dynamic signal may also
be adversely influenced by a standing acoustic wave within the measurement
piping (blowing across an empty bottle effect). In some cases, it may be desirable
to fool the fluid mechanics of the system by externally increasing the length of
the measurement cavity. Specifically, a fixture may be constructed with several
feet of stainless steel tubing connected to the transducer measurement chamber.
This tubing is generally rolled into a coil to be physically manageable, and the
end of the tubing is plugged. Since the acoustic resonant frequency is inversely
proportional to the passage length, an increased physical length will result in a
significant reduction in the value of the acoustic frequency. A typical field application of this concept is shown in Fig. 6-33.
Pressure Pulsation Probe
Signal Output
To Process
Connection
Probe Reference Line

End
Plug
Stainless Steel Coil

of 1/4 Diameter
Stainless Steel Coil

of 1/16 Diameter
Fig. 633 Measurement Fixture For Extending Pressure Pulsation Probe Passage Length
This assembly is connected to the process stream through whatever nipples

or valves are required. The 1/4 diameter coil is connected directly to the measurement chamber, and a smaller coil with a tube diameter of 1/16 is connected
to the end of the large coil. This functions as a long tube that is open at one end
(process end), and closed at the opposite end (plugged end). One quarter of a
standing acoustic wave is contained within this device, and the fundamental
wavelength is equal to four times the total passage length. The acoustic natural
frequency is determined by dividing sonic velocity of the process fluid by the fun-
Pressure Pulsation Transducers
287
damental wavelength. A more detailed explanation of this phenomena is

described in chapter 10. However, for this discussion it should be recognized that
the acoustic resonant frequency can be adjusted, within physical limits, to a frequency that is outside of the frequency domain of interest.
If the pressure sensor installed in the measurement fixture consists of a
sealed assembly containing a diaphragm plus attached piezoelectric crystal, the
output will be dynamic pressure pulsation via a coaxial cable (signal plus shield).
However, other types of pressure probes are in common use. Manufacturers such
as Endevco Corporation and Kulite Semiconductor Products, Inc. offer miniaturized solid state devices that consist of a four arm Wheatstone bridge mounted on
the surface of a silicon diaphragm. This type of pressure transducer is known as
a piezoresistive transducer. In some probes, a pair of silicone strain gages are
combined with a pair of fixed resistors in the transducer bridge. A higher sensitivity is obtained by using four active gages in the Wheatstone bridge circuit. In
either case, the primary advantage of a piezoresistive transducer is the ability to
measure down to zero frequency (DC). With respect to pressure measurements,
this provides the unique and desirable capability of simultaneously measuring
the static pressure as well as the dynamic pressure pulsation. The ability of
piezoresistive pressure pickups to measure static pressures also allows an easy
calibration with a simple hydraulic dead weight tester. These types of pressure
transducers are easily recognizable by the four wire output cable configuration
that is often combined within an external shield for noise suppression.
Another distinguishing characteristic of some piezoresistive pressure transducers is the presence of a reference line or vent tube protruding from the probe.
An example of this thin wall tubing is shown in Fig. 6-33 as the Probe Reference
Line. This tubing is connected to the back side of the diaphragm, and allows several different operating modes for the pressure pickup. If the tubing is open to
the atmosphere as shown in Fig. 6-33, the transducer will be referenced to atmospheric pressure, and the resultant output will be a relative pressure (Pounds/
Inch2 Gage). If the reference tube is evacuated in a vacuum chamber, and the
tube crimped shut, the transducer output will be an absolute pressure (Pounds/
Inch2 Absolute).
Alternatively, the probe reference line may be connected to other pressure
sources to measure a differential pressure. This same capability may be
employed to eliminate the static pressure from the final output signal. For
instance, if the reference line shown in Fig. 6-33 is connected to the end of the
small coil, the static pressure will be equalized across the diaphragm. This would
effectively AC couple the output signal, and allow enhanced observation or
amplification of the dynamic pressure pulsation portion of the electronic signal.
Clearly, there are many items to be considered in the proper selection and
application of pressure pulsation transducers. A properly engineered installation
produces meaningful information, and a poorly conceived installation generates
more questions than answers. As with most dynamic transducers, the machinery
diagnostician must carefully consider all aspects associated with the installation
and utilization of pressure pulsation probes.
288
Chapter-6
SPECIALIZED TRANSDUCERS
The piezoelectric transducer concept is not only applicable to accelerometers and pressure probes, it is also used for other types of dynamic transducers
such as force and shock probes. For ICP devices, the end user enjoys the convenience of interfacing with identical power supplies. Hence, a variety of ICP
transducers may be driven by the same power supplies, or a Dynamic Signal
Analyzer (DSA) such as an HP-35670A.
Another common application of piezoelectric transducers resides in the
domain of force measurement. Force sensors are used to measure compression,
tension, and impact forces involved with a wide variety of manufacturing processes. There are also many applications for load or force transducers for measurement of radial or axial bearing loads on machines. Some of these
installations are for temporary test measurements, and other installations are
designed for long-term continuous monitoring of machinery forces.
In the realm of structural testing, piezoelectric force transducers are used
in devices such as impact or impulse hammers to deliver a short duration force
pulse to a structure. The force transducer measures the output characteristics of
the force pulse, and a separate accelerometer is used to measure the resultant
structural response. The clarity of these signals are verified in the time domain,
and a frequency response function (FRF) is performed with a Dynamic Signal
Analyzer between acceleration and force. The resultant frequency response curve
is used to identify structural resonances plus damping characteristics.
In critical applications, it may be desirable to perform a system calibration
that involves the accelerometer, the impact hammer, plus the data processing
instrumentation. This apparently difficult task may be accomplished by the simple test fixture depicted in Fig. 6-34. In this diagram, a steel cylinder is suspended from a single stationary point using a two point hitch. This type of
Rigid Support
Support String
Diameter (D)
Accelerometer
Fig. 634 Typical Test

Setup Used For Mass
Calibration of Piezoelectric Force Hammer And
Accelerometer
Impact Hammer
Length (L)
Accelerometer
Output to DSA
Impact Hammer
Output to DSA
Specialized Transducers
289
suspension is critical for obtaining proper accuracy. This cylinder represents a

mass at rest. If the mass is excited with an input force, the mass should accelerate in proportion to the applied force. More succinctly, this is known as Newtons
second law of motion, and it is customarily stated as follows:
Force = Mass Acceleration
(6-7)
In the test setup shown in Fig. 6-34, the mass is constant. Assume that the
input force is measured with a force transducer that is integral with the impact
hammer, and acceleration of the mass is measured with the axial accelerometer.
Intuitively, mass should be determinable with a FRF of force over acceleration.
As an example, consider a solid steel cylinder that is 1.50 inches in diameter, and
3.5 inches long. The weight is computed with equation (3-7) as follows:
2
LD
3.50 Inches 0.283 Pounds/Inch ( 1.50 Inch )
W = ------------------------------------- = ------------------------------------------------------------------------------------------------------------------------- = 1.75 Pounds
4
4
To check these calculations, the steel cylinder was weighed on a digital laboratory scale at 790 grams. This metric mass unit may be converted to equivalent English units of pounds in the following manner:
1 Pound
W = 790 Grams ----------------------------- = 1.74 Pounds
453.6 Grams
Hence, the calculated and measured weights are equal. The next step consists of connecting the impact hammer force transducer to channel 2, and the
accelerometer to channel 1 of an HP-35670A analyzer, and performing an FRF
during several sample hits with the hammer. The results of this test are shown
in Fig. 6-35 for a total of eight samples. At the midscale of 48,000 CPM (800 Hz),
the amplitude is equal to 1.744 pounds. Obviously this is excellent agreement
with the calculated and the measured weight of this steel cylinder. If significant
differences appear between the actual weight and the FRF weight, the calibration of the force transducer or the accelerometer might be questionable. This
could be resolved by individual calibration of both piezoelectric transducers.
Alternatively, a percentage deviation could be established between the actual
weight and the FRF weight. This constant offset could then be applied to all
future FRFs with the DSA math functions.
For reference purposes, the test setup shown in Fig. 6-34 and the resultant
FRF presented in Fig. 6-35 were performed prior to a turbine blade test. The
overall test weights were fairly small, and a 1 gram accelerometer was combined
with a miniature force hammer for this test. The same results may be achieved
with physically larger systems, but the diagnostician should be very careful with
the physical setup as well as the DSA operation. The transducers should be connected directly to the DSA, and powered by the internal ICP power source. A
suitable force window should be used for the hammer input on channel 2, and a
proper exponential window selected for the accelerometer on channel 1. The
hammer should be fixed to a constant pivot point so that it strikes the mass at
essentially the same point for every sample. Also, the mass must be stationary
290
Chapter-6
Fig. 635 Frequency

Response Function (FRF)
Used For Mass Calibration
Of ICP Force Hammer
And Accelerometer
before every hit. This is the only way to guarantee comparable data for each of
the FRF samples. Finally, if the mass line is not straight, or if the phase differential is higher than 2, or if the coherence falls below 0.99, then there is something wrong with the test setup or the DSA.
This same data processing scheme may be used for examining force divided
by an acceleration signal that is double integrated to displacement. The resultant FRF of force over displacement provides a measurement of dynamic stiffness
versus frequency. Certainly this is important parameter that may be necessary
to properly describe or model the behavior of a particular machine and/or associated structure. This technique was discussed in chapter 4, and case histories 8
and 9 describe the support stiffness measurements on a steam turbine and a gas
turbine bearing housing support.
Further explanations and discussions of these various piezoelectric transducers are beyond the current scope of this text. The machinery diagnostician
should recognize that these devices do exist, and they may provide essential
information on specific mechanical problems. In addition, there are other specialized transducers that are commonly available within the industrial community.
For instance, lasers are used for static machinery alignment measurements as
discussed in chapter 12. Lasers are also used for non-contacting lateral and torsional vibration measurements on various machines and structures. Although
the laser based instruments are both expensive and complex, the capabilities are
significant. In some cases, laser based transducers provide the only method for
direct measurement of a vibrating surface.
Another type of optical transducer is the timing probe shown in the photo-
291
Fig. 636 Typical Optical

Transducer Used For
Machinery Timing Measurements
graph in Fig. 6-36. This device is used to observe a piece of reflective tape, or
other variable contrast media attached to a rotating shaft or a reciprocating surface (e.g., compressor drive rod). The coincidence of the reflective tape with the
optical probe produces a pulse signal that may be used as a Keyphasor for field
balancing or malfunction diagnosis. This type of transducer is ideally suited to
machines that cannot tolerate drilled holes or milled slots in exposed shaft surfaces. High speed machines are often candidates for this type of timing pickup.
Unfortunately, the use of reflective tape for the shaft timing mark is usually limited by the surface speed of the shaft. For cases of high surface velocities, the use
of black spray paint or layout bluing may be used to uniformly darken the entire
shaft. A strip of reflective paint may then be applied to the shaft surface to act as
the trigger mark. The signal output from the optical driver may then be observed
on an oscilloscope, and the suitability of the resultant pulse signal evaluated.
Obviously, corrections may be performed to the reflective mark to improve the
clarity and consistency of the final optical Keyphasor signal.
Other dynamic transducers such as strain gages may be successfully used
on mechanical structures as well as on rotating shafts. There are many available
techniques for extracting this type of information, including direct wiring, telemetry, brush contacts, and some newer optical transmission techniques. It should
be recognized that strain gage application, installation, and proper operation are
almost an independent branch of measurement technology. Strain gages may be
installed individually to measure strain in one direction, or three gages may be
mounted in rectangular or various rosette configurations to determine the principal strains. From the strain data, the principal stresses may be computed by
incorporating the modulus of elasticity E for the material. The resultant stress
levels may then be compared with the elastic strength of the material under test.
Typically, various safety factors are also included to form the final assessment of
measured versus allowable stress limits.
The proper application of strain gage technology is a complex endeavor that
should categorically be classified as an experimental stress analysis technique.
Certainly there are occasions when it is necessary to examine the stress charac-
292
Chapter-6
teristics of a mechanical element. In this situation, the machinery diagnostician

is encouraged to examine additional technical references, such as chapter 17 of
the Shock and Vibration Handbook2, or Practical Strain Gage Measurements by
Hewlett Packard3. More detailed discussion of strain gage characteristics and
specific application notes may be obtained from the vendors of these devices.
Companies such as Measurements Group, Inc.4 provide excellent technical literature as well as extensive training materials.
Strain gages are used in devices such as piezoresistive accelerometers and
pressure transducers. These probes have the ability to measure constant acceleration or pressure levels (via DC output) combined with high frequency behavior
or long duration shock pulses. Strain gages may also be applied for the measurement of tension and torsion in a machine shaft. This is a simple measurement on
a non-rotating shaft. However, when the shaft is turning, the complexity of the
strain measurement increases significantly. For temporary tests on rotating
shafts, the radio telemetry system offered by Binsfeld Engineering (Maple City,
Michigan) works quite well. In this system, the shaft mounted strain gages are
powered by a nine volt battery and a transmitter that are attached to the shaft
with strapping tape. A stationary circumferential antenna picks up the strain
output from the rotating transmitter, and directs the signal to a Receiver Demodulator. After this device demodulates the signals, traditional instruments may be
used to examine the static and dynamic strain signals.
For permanent or continuous monitoring of shaft torque, the non-contacting
TorXimitor by Bently Nevada Corporation provides an attractive offering.
Within this system, a balanced strain gage bridge is mounted on a machine coupling or spool piece. The strain gages are combined with inductive power receivers, and FM signal transmission electronics. Adjacent to the coupling, a
stationary module provides inductive energy to the rotating coupling electronics,
and receives the strain gage signals broadcasted by the coupling transmitter.
Finally, the FM signals are demodulated and conditioned in a separate module to
provide an electrical signal that is proportional to torque. If a Keyphasor signal
is included, the transmitted horsepower is determined from equation (2-98).
In addition to the diversity of specialized dynamic transducers that are
available, there are many variations of the standard displacement, velocity, and
acceleration probes. For instance, extended range proximity probes exist that
have linear operating ranges that easily exceed one inch. These types of probes
generally have low sensitivity values in the range of 5 to 10 millivolts per mil.
Other proximity probes are designed to be waterproof, resistive to a variety of
industrial chemicals, and capable of withstanding significant operating pressures and temperatures.
Magnetic probes may be used to measure repetitive events when they are
1996), chap. 17.
3 Practical Strain Gage Measurements, Application Note 290-1, (Hewlett Packard, printed in
USA, 1987).
4 Strain Gage Technology-Technical Reference Binder, (Raleigh, North Carolina: Measurements
Group, Inc., 1996 update).
293
positioned over conductive materials with discontinuities. These types of transducers are often used as primary sensors in turbine speed control systems. Typically, they are mounted in a radial direction over multi-tooth gears, and they
produce an electrical pulse as each gear tooth passes underneath the probe. In
control systems, three probes are used over the same gear to provide full redundancy and backup for the measurement. Dividing the pulse passing frequency by
the number of gear teeth will yield the shaft rotative speed. It should be mentioned that the electrical signal output from these transducers is dependent
upon the size and shape of the gear teeth, the peripheral speed, and the distance
between the magnetic probe tip and the top of the gear teeth. In many applications these pickups are gapped at only 10 to 15 Mils (0.010 to 0.015 inches) from
the gear teeth. This close distance makes the probes susceptible to damage.
Magnetic pickups or proximity probes are also used for the measurement of
torsional vibration. For this application, the transducer is positioned to observe a
set of gear teeth or a precision slotted wheel. The resultant tooth (or slot) passing
frequency is radially influenced by the torsional motion. In this application, the
transducer output is a frequency modulated (FM) signal. The high frequency carrier signal is the tooth (or slot) passing frequency, and the lower frequency modulating signal is proportional to the torsional vibration. Thus, demodulation of the
FM signal produces the torsional vibration. Typically, two non-contacting pickups, spaced 180 apart, are used to observe the slotted wheel. Both of the signals
are fed into a device such as the Bently Nevada TK17 Torsional Vibration Signal
Conditioner. Within this instrument the two opposed signals are used to cancel
out radial vibration, and thereby minimize measurement noise. The final output
of the TK17 is the conditioned and demodulated signal proportional to torsional
vibration. This signal carries engineering units of volts per degree, and it may be
processed or recorded with any traditional array of diagnostic instrumentation.
Historically, torsional velocity coil transducers have been installed on the
blind end of gear shafts for short-term unit testing. These devices operate similar
to casing velocity pickups, with the significant difference that the output signals
from the velocity pickup must be obtained through slip rings. This requirement
limits the total useful operating life of this type of transducer to a few hours.
Furthermore, the machinery train must be shutdown for removal of the torsional
velocity coil at the end of the test. This may not be a particular problem for the
small blower installation, but it can be a significant expense to shutdown and
restart a large turbine generator set.
In the overview, a wide variety of static and dynamic measurement transducers are available in the industrial marketplace. The machinery diagnostician
must define the specific test measurements required. Based on these requirements, the best possible transducer suite should be selected. There may be distinct differences in the transducers temporarily installed for a field test versus
the pickups normally used for permanent monitoring and machinery protection.
This applies to the type and quantity of the transducers as well as to the accuracy of the calibration of the various devices. In all cases, direct measurements
are preferable to implied measurements, and correlation of multiple measurements remains as a highly desirable objective.
294
Chapter-6
ASPECTS OF VIBRATION SEVERITY

Machinery vibration measurements are often viewed as quality control, or
machinery condition measurements. Although this may be an oversimplification,
most people consider low vibration levels to be indicative of proper machinery
behavior. When vibration amplitudes are high, the general tendency is to believe
that the machinery is experiencing some type of distress or mechanical malfunction. Ideally, this concept should be predicated upon the existence of a definitive
method, engineering specification, or set of guidelines that allow a clear cut definition of acceptable versus unacceptable vibration amplitudes. In reality, the
only statement that applies to all machines, under all conditions, is as follows:
There Are No Universal Vibration Severity Limits
Over the years many organizations and corporations have tried to establish
vibration severity standards. Although partial success has been attained with
some machines, and some specific applications, the establishment of universal
vibration tolerances has not been achieved. When the measurement complexities
are combined with the intricacies and variations between machines, it is self-evident that development of universal vibration severity limits may be beyond any
reasonable expectation.
Even though it is doubtful that a universal criteria can be established, that
still does not eliminate the need for a methodology to define workable vibration
limits. Within the context of this book, it is reasonable to examine the various
parameters that influence the measured vibration, and provide some guidance
for addressing this complex question. The first issue that should always be considered is the fundamental accuracy of the measurement. The vibration transducer, cables, interface devices, signal converters, and final readout must be
checked and calibrated as a system. If a problem exists in any part of the measurement chain, the validity of the entire measurement is compromised. Hence,
routine calibration checks are mandatory.
The next consideration is the proper operation and calibration of any diagnostic instrumentation that is used to analyze the vibratory behavior. This
equipment must also be rigorously and periodically checked for proper operation
and calibration. Fortunately, many modern digital instruments are both selfchecking and self-calibrating. This is a comforting feature that provides an
improvement in data consistency and accuracy. However, with any instrumentation system, the diagnostician must continually compare the final hard copy output data with the transducer output signals to verify that some anomaly has not
occurred within the data processing system.
For many measurements, a more fundamental question is the utilization of
the correct vibration transducer for a specific measurement task. In chapter 2 of
this text it was shown that displacement, velocity, and acceleration of an element
are integrally related. If one of the three vibration parameters and the frequency
was known, the other two vibration values could be easily calculated with equations (2-20) through (2-22). Phase difference between the three fundamental
Aspects of Vibration Severity
295
Proximity
Probe
10
Piezoelectric
Accelerometer
1
Velocity Coil
0.1
60,000
1000
100
6,000
600
10
160
Vibration Amplitude (Mils, IPS, Gs)
100
600,000
10000
B B
H
H
J H
H J
BH B B
1,200,000
20000
motion properties was also defined in equations (2-14) through (2-16). Hence, it
might appear that any vibration value or vector could be easily converted into
any other convenient set of engineering units. Indeed, this was the case presented in Fig. 2-4 where a constant velocity of 0.3 IPS,o-p was converted to equivalent displacement and acceleration amplitudes for a wide range of frequencies
(1 to 20,000 Hz). This earlier plot considered nothing more than the pure transformation of vibration parameters.
However, as demonstrated within this chapter, the individual vibration
transducers have specific areas of application, and definite limitations of frequency response characteristics. If the data from Fig. 2-4 is replotted with some
realistic limitations on the actual frequency operating ranges of the vibration
pickups, the following Fig. 6-37 may be drawn.
B BB
J
J
J
20,000
10,000
1,000
100
10
0.01
Frequency (Cycles/Second or Hertz)
Fig. 637 Typical Frequency Operating Ranges For Traditional Vibration Transducers
This log-log plot reinforces the fact that proximity probes are more suitable
for the lower frequency measurements around machine operating speed. Accelerometers are the correct transducer for high frequency measurements such as
blade passing or gear meshing characteristics. Furthermore, the relative position
and vibration capabilities of proximity probes makes them eminently appropriate for measurements on machines with fluid film bearings. On machines
equipped with rolling element bearings, or units that normally emit high frequency excitations, the external mounting of casing accelerometers makes the
most engineering sense.
296
Chapter-6
Direct velocity measurements have often been touted as the most informative type of vibration data. This reputation is based upon the fact that frequency
and displacement are combined into one value, for example, equation (2-20).
Although a velocity coil has a limited frequency range (as shown in Fig. 6-37),
the concept of using specific velocity amplitudes for severity evaluation has been
popular in many applications, and for many years.
For bearing housing vibration measurements, it is common to assign specific velocity amplitudes to varying levels of mechanical condition. For instance,
Fig. 6-38 identifies nine different levels of probable machinery condition versus
the associated casing velocity amplitudes. This varies from a category of excellent
condition at 0.002 IPS,o-p to danger at 1.0 IPS,o-p. This plot converts the velocity
amplitudes into bearing housing displacement amplitudes with associated engineering units of Mils,p-p.
Bearing Housing Vibration Amplitude (Mils, p-p )
1
60
10
600
100
6,000
1000
60,000
5000
300,000
200
100
Hig
he
st
Ob
se
rve
dV
elo
cit
y2
.5
IPS
,o-p
Go
Ro
Ac
Da
Ex
Ve
Sm
Ma
Ve
c
ug
od
ce
ry
r
r
h 0 nger
lle
Sm ooth y Go
0.0 eptab gina
nt
1
.5
l0
0.0
od
oo
5I
le
0.0
IPS .0 IP
.
t
2
P
0
1
0
h
.1
S,o
02
IPS
IPS .02
0.0
S,o
,o-p
I
P
IPS
p
I
,
-p
05
,
PS
o-p
S,o
,o-p
IPS o-p
,o-p
-p
,o-p
10
0.1
0.01
1
10
100
1,000
5,000
Fig. 638 Typical Vibration Severity Chart For Properly Supported Bearing Housings
Understandably, the highest amplitudes are tolerated at low frequencies,

and the acceptable vibration level decreases with increasing frequency. This
behavior of allowable vibration amplitude versus frequency is reasonable, and it
is incorporated on virtually all severity charts. Many organizations use vibration
severity charts that are similar to Fig. 6-38. Although the names of the categories will change, and the assigned velocity amplitudes will vary, the general
chart configuration will resemble Fig. 6-38.
Although the concept of setting vibration limits based upon constant velocity sounds like a reasonable approach, it does not work in many cases. For
instance, Fig. 6-38 shows a line of constant velocity at 2.5 IPS,o-p. This line is
297
labeled as highest observed velocity. Casing vibration amplitudes at this level

have been observed by the senior author on several occasions. Although failures
did eventually occur, the machinery operated for a considerable length of time. In
two other cases, speed increasing gear boxes displayed bearing cap vibration levels in excess of 2.0 IPS,o-p. These gear boxes sustained large casing vibration
amplitudes for many years with no adverse influence upon the gear teeth or the
bearings. At the opposite end of the severity chart, mechanical failures have been
documented with casing vibration levels of less than 0.05 IPS,o-p.
Clearly, the correlation between casing vibration and the excitation associated with various types of failure mechanisms is neither a definable, nor a constant quantity. For these types of measurements, the machinery diagnostician
must recognize that different machines, operating in different services, with different mechanical components are always installed with a variety of foundation
and piping systems. It is only logical to conclude that different machines must
exhibit different casing vibration characteristics. In practice, casing vibration
measurements are normally applied on small general purpose equipment trains
equipped with rolling element bearings. Since universal casing vibration limits
cannot be accurately quantified, the typical approach for this type of machinery
is to perform trend analysis as a function of time. Although routine variations in
casing amplitudes and frequencies will appear, the envelope of normal behavior
should eventually become visible, and reasonable alert and danger levels may be
established for each machine.
For larger machinery trains, the main bearings are normally fluid film
bearings as discussed in chapter 4. For these critical units, the additional flexibility and damping at the journal bearings makes a severity chart such as Fig. 638 even less appropriate. On these types of machines, the primary vibration
measurements are made with non-contacting proximity probes mounted at each
bearing housing. As discussed earlier in this chapter, the proximity probes have
the ability to measure journal position within the bearing, plus relative shaft
vibration. If the peak to peak shaft vibration is compared against the available
bearing diametrical clearance, a measure of severity may be deduced. For
instance, the values listed in Table 6-3 are considered to be appropriate for a
variety of machine trains equipped with fluid film bearings.
The proper utilization of Table 6-3 requires knowledge of the radial bearing
clearance. If this diametrical clearance is not available, the machinery diagnostiTable 63 Machinery Condition Based Upon Runout Compensated Shaft Vibration As A
Percentage Of The Total Available Diametrical Journal Bearing Clearance
Machine
Condition
Peak to Peak Shaft Vibration As A Percentage of Diametrical Bearing Clearance
Appropriate Action
Normal
Less Than 20% of Available Clearance
Continue Monitoring
Alert
40% to 60% of Available Clearance
Initiate Corrective Action
Danger
More than 70% of Available Clearance
Shutdown Machine
298
Chapter-6
cian might estimate the bearing clearance based upon the shaft diameter. As discussed in chapter 4, a bearing clearance ratio (BCR) of 1.5 Mils per inch of
diameter is quite common on large machines. Thus, if a 6.0 inch diameter shaft
journal is encountered, a normal diametrical clearance would be in the vicinity of
9.0 Mils. Applying the percentages shown in Table 6-3 would indicate that normal runout compensated shaft vibration should be less than 1.8 Mils,p-p. By the
same token, a danger or shutdown level would be at 70% of the available bearing
clearance, or a trip setpoint of 6.3 Mils,p-p.
These are reasonable amplitudes for shaft vibration of a 6 inch journal.
However, they do not directly address the speed or load characteristics of the
bearing. It is implied that large shafts run at slower speeds with larger absolute
clearances, and smaller diameter shafts run at higher speeds with proportionally
smaller bearing clearances. The work by Jim McHugh5 incorporated the static
rotor load on the bearings. For a simplified case of synchronous 1X shaft vibration in fluid film bearings, McHugh offered the following empirical equations:
where:
25
Acceptable Level -------------- D c + 1
BUL
(6-8)
50
Alarm Level = -------------- D c
BUL
(6-9)
100
Danger Level = -------------- D c
BUL
(6-10)
Acceptable Level
Alarm Level
Danger Level
Dc
BUL
= Acceptable Shaft Vibration Amplitude (Mils,p-p)

= Setpoint for First Vibration Alarm (Mils,p-p)
= Setpoint for Second Vibration Alarm (Mils,p-p)
= Bearing Diametrical Clearance (Mils)
= Bearing Unit Load (Pounds / Inch2)
The bearing unit load (BUL) in the above three expressions is simply the
shaft weight upon the bearing divided by the plane area of the bearing. This concept was previously discussed in chapter 4, and the BUL is easily calculated with
equation (4-1). McHugh includes a +1.0 Mil,p-p additional amplitude in equation
(6-8) to allow for runout, or for measurements at other than the mid-plane of the
bearing. Of course, a runout vector could be directly opposed to the synchronous
vibration vector, and this would result in a decrease in uncompensated shaft
vibration. It could also be argued that equations (6-9) and (6-10) should also
include some type of runout correction. Nevertheless, these three equations do
reinforce the concept that bearing load is an important consideration in establishing vibration severity limits. In essence, machines with high static loads are
less tolerant to excessive vibration than machines with low bearing loads.
5 James D. McHugh, Setting Vibration Criteria for Turbomachinery, Proceedings of the Eighteenth Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University, College
Station, Texas (October 1989), pp. 127-135.
299
For a typical case of a bearing unit load of 150 Pounds/Inch2 and a diametrical clearance of 9.0 Mils, the acceptable vibration amplitude from equation (6-8)
would be 2.5 Mils,p-p. Similarly, the danger or trip level may be determined from
equation (6-10) as 6.0 Mils,p-p. These values are in general agreement with the
bearing clearance percentages presented in Table 6-3.
Another approach incorporates machine operating speed into the severity
evaluation. For example, consider Fig. 6-39 that plots relative shaft vibration
amplitude in Mils,p-p versus machine running speed. Clearly, this diagram is
similar in construction to other vibration severity charts. In this case, the magnitudes and severity categories are based upon a myriad of measurements by the
senior author on a wide variety of industrial machines. These vibration levels are
from proximity probes mounted adjacent to the bearings, and the amplitude
readings are corrected for shaft runout
Speed (Revolutions/Minute)
600
1,200
3,000
6,000
12,000
30,000
60,000
500
1,000
Shaft Vibration Amplitude (Mils, p-p )
10.0
Dange
r - Sh
utdow
n
Alert Alarm
5.0
Rough
2.0
Accep
table
Very G
ood
1.0
Excell
ent
0.5
0.2
10
20
50
100
200
Speed (Revolutions/Second or Hertz)
Fig. 639 Runout Compensated, Relative Shaft Vibration Severity Chart
Once again, it is noted that vibration tolerance decreases with operating

speed. This is both reasonable and expected behavior. At a fixed speed, it is presumed that deterioration on a machine will propagate as increased vibration
amplitudes, and a shutdown will eventually be required for problem correction.
It can be argued that this graph is conservative in some areas, and it is liberal in
other regions. Some individuals might disagree with the categories or the specific
amplitudes displayed on Fig. 6-39, but the overall concept is generally sound
from a machinery management standpoint. Please note that any of the vibration
severity charts presented in this text (or from any other common source), should
300
Chapter-6
only be considered as general information. These vibration severity charts must

not be construed as absolute or specific guidelines.
It is often difficult to draw a line of demarcation between acceptable and
dangerous vibration amplitudes. The fundamental issue always comes down to
the situation of shutting down the machine before a catastrophic failure occurs,
versus the potential for unnecessary shutdowns due to normal changes or variations of the machinery. On large process machines many factors influence the
shaft vibratory behavior, and the following list summarizes typical considerations for severity evaluation, and the establishment of suitable vibration limits.
Bearing Diametrical Clearance

Machinery Maximum Continuous Operating Speed
Bearing Static Load (that is, shaft weight upon journal)
Bearing Dynamic Load (loads due to gear contact forces, unbalance, etc.)
Actual Stress/Strain Levels Imposed by the Static plus Dynamic Loads
Shaft Centerline Position
Bearing Temperature at Minimum Oil Film
Modal Location of Measurement versus Physical Bearing Location
Shaft Electrical Runout
Shaft Mechanical Runout or Eccentricity
Frequency Distribution of Shaft and/or Casing Vibration
Journal Bearing Configuration
Bearing Housing Support Flexibility
Foundation Support Flexibility
Casing to Rotor Weight Ratio
Potential Influence from Attached Large Bore Piping
Potential Influence from Adjacent Machinery or Other External Excitations
Sensitivity to Process and/or Load Variations
Sensitivity to Variations between Day and Night
Sensitivity to Variations between Summer and Winter
New machines are often started with somewhat loose vibration limits due
to the unknown behavior in many of the above categories. Although computer
simulations can provide meaningful information of some mechanical conditions
(e.g., unbalance), the final installed field vibratory behavior must always be measured and evaluated. Over a period of time, the vibration limits may be
decreased as the operating experience increases. Again, the objective is always to
protect the personnel, protect the machinery, and minimize false alarms.
With respect to new plants and machinery installations, one of the issues
that continually reappears is the allowable vibration limits on piping. Acceptable
piping vibration amplitudes are not well defined within the process industries.
Technical organizations and specifications often bypass this topic with statements such as in cases of excessive piping vibration, the problem shall be corrected by adjusting supports, dampers, and snubbers accordingly Although
301
6,000
3,000
10
1,200
600
300
120
60
18,000
that may be acceptable in a design specification, it may be inordinately difficult

to evaluate and correct in the field.
Generally, there are three acceptable methods for evaluating piping vibration severity. The first approach is the direct technique of measuring strain, and
multiplying by the modulus of elasticity to determine stress. Comparison of the
stress against established endurance limits results in an evaluation of the potential for high cycle fatigue. In many cases of carbon steel piping, a measured level
of 100 microstrain is considered to be acceptable. This value of 100 microstrain
includes stress concentration plus safety factors. Hence, experimental stress
analysis may be used to determine the suitability of vibrating pipe.
The second approach for evaluating the suitability of a piping system is to
measure the peak velocity of the piping, and apply a constant velocity criteria to
all frequencies. Historically, acceptable levels varying from 0.5 to 2.0 IPS,o-p have
been suggested as maximum piping vibration limits. Again, this depends upon
the specific technical application, and method of pipe support.
The third method for evaluating piping systems is also based upon piping
vibration measurements. In this technique, vibration amplitudes are measured
in displacement, and the amplitudes are compared against frequency for various
severities. The empirical results by J.C. Wachel6 have been widely published, and
a rendition of this piping severity chart is shown in Fig. 6-40. In addition to the
100
300
100
300
200
Constant Velocity = 0.5 IPS,o-p
Piping Vibration Amplitude (Mils, p-p )
100
50
Dan
ger L
evel
20
Corr
ectio
n Re
quire
d
10
Marg
inal
Desig
n
Perc
eptio
n Th
resh
old
0.3
1
10
20
50
Fig. 640 Typical Piping Vibration Severity Chart By J. C. Wachel and J.D. Tison
6 J.C. Wachel and J.D. Tison, Vibrations In Reciprocating Machinery and Piping Systems,
Proceedings of the Twenty-Third Turbomachinery Symposium, Turbomachinery Laboratory, Texas
A&M University, College Station, Texas (September 1994), pp. 243-272.
302
Chapter-6
five severity levels, a line of constant velocity corresponding to 0.5 IPS,o-p has
been included. Again it must be recognized that this type of information may be
directly applied on some installations, and in other cases it provides a reference
point to begin a more detailed evaluation of the installed mechanical system.
In order to be perfectly clear, it should be restated that There Are No Universal Vibration Severity LimitsTolerances will vary by machinery type, configuration, application, installation, vibration transducer, and industry. The only
real answer to the establishment of vibration severity criteria lies in careful
measurement and proper engineering evaluation of each specific machinery
installation. Finally, the machinery diagnostician should not be surprised when
some malfunctions result in a decrease in measured vibration. Although this
only occurs under unique conditions, it is a true physical reality.
BIBLIOGRAPHY
1. Harris, Cyril M., Shock and Vibration Handbook, Fourth edition, chap. 17, New
York: McGraw-Hill, 1996.
2. McHugh, James D., Setting Vibration Criteria for Turbomachinery, Proceedings of
the Eighteenth Turbomachinery Symposium, Turbomachinery Laboratory, Texas
A&M University, College Station, Texas (October 1989), pp. 127-135.
3. Practical Strain Gage Measurements, Application Note 290-1, (Hewlett Packard,
printed in USA, 1987).
4. Strain Gage Technology - Technical Reference Binder, (Raleigh, North Carolina:
Measurements Group, Inc., 1996 update).
5. Vibration, Axial Position, and Bearing Temperature Monitoring Systems - API
Standard 670, Third Edition, American Petroleum Institute, (Washington, D.C.:
American Petroleum Institute, November 1993).
6. Wachel, J.C. and J.D. Tison, Vibrations In Reciprocating Machinery and Piping Systems, Proceedings of the Twenty-Third Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University, College Station, Texas (September 1994),
pp. 243-272.
Dynamic Signal Characteristics 7
ynamic signals may contain a myriad of

discrete frequencies, associated harmonics, interactions between components,
plus various types of noise and signal interference. As the transducer frequency
range is expanded, the signal complexity increases. For instance, a proximity
probe observing a clean radial surface on bull gear shaft might exhibit the bull
gear rotational frequency, plus an excitation at the pinion speed. By comparison,
a casing mounted accelerometer might pick up the running speed components,
the gear mesh frequency, the frequencies associated with box structural resonances, blade passing excitation from the lube oil pump, plus excitations from an
adjacent machinery train. Obviously, the acceleration signal will be considerably
more complex than the displacement signal.
In order to attempt any type of mechanical assessment of machinery behavior, it is necessary to dissect the dynamic signals into discrete and understandable portions. To accomplish this goal, a wide variety of electronic instruments
are commercially available. In virtually all cases, these devices incorporate some
fundamental types of electronic filters to assist in examination of the dynamic
signals. These electronic filters may be analog, digital, or a combination of both.
ELECTRONIC FILTERS
The array of electronic filters may initially appear to be staggering, but in
essence there are only two fundamental types of filters. These basic filters are
the low-pass, and the high-pass filters. All other filter types and configurations
are merely combinations of the two basic types. Hence, if the characteristics of
the basic filters are understood, then the more sophisticated combinations of filters may be comprehended, and these electronic devices properly applied for
dynamic signal enhancement and examination.
In order to explain filter characteristics in a consistent manner, a series of
examples have been prepared. In the following cases, a Hewlett Packard 35665A
Dynamic Signal Analyzer (DSA) was used to generate a broad band 1.0 volt signal that maintained the same amplitude over a wide frequency range. This white
noise signal was used as the input into the various filters as shown in Fig. 7-1.
303
304
Chapter-7
Dual Channel
Dynamic
Signal
Analyzer
Source
Ch.A
Input
Ch.B
Input
White Noise
Fig. 71 Instrumentation
Arrangement For Frequency
Response Testing Of An
Electronic Filter
Input
Output
Electronic Filter
The broad band noise signal was also used as a direct input into channel A of the
DSA. The noise signal was filtered by the external electronic filter, and the filtered output signal was used as an input into channel B of the DSA. The actual
influence of the electronic filter was determined by performing a frequency
response function (FRF) between the filter output and the noise source input.
In this type of testing, it is mandatory to consider the filter effect upon the
signal amplitude as well as the influence upon the phase angle. Although this
information could be generated with an oscilloscope and a function generator, the
DSA performs this output/input comparison much faster, and with greater accuracy. As an example, consider the frequency response function (FRF) shown in
Fig. 7-2 of a low-pass filter set at 50 Hz (3,000 CPM).
For this data, a Krohn-Hite, model 3323 dual channel analog filter was used
(-24 dB/Octave per channel). This class of filter passes the low frequencies (i.e.,
low-pass), and it rejects the high frequencies. By definition, this filter has a single transmission band extending from the lower frequency limit of the device to
some finite upper cutoff frequency. The filter frequency is coincident with an
amplitude attenuation of -3 dB. In this case, the filter was manually set at 50 Hz
(3,000 CPM). From the test data in Fig. 7-2 it is noted that the 1.0 volt input signal was reduced to 0.702 volts at 50 Hz (3,000 CPM). This is very close to the filter design value of 0.707 volts (i.e., -3 dB = 0.707).
Note that the signal amplitude varies with frequency. Even though the filter was set at 50 Hz (3,000 CPM), amplitude attenuation occurs down to a frequency of 30 Hz (1,800 CPM). At the upper end of the frequency scale, it is clear
that 10% of the voltage signal is still passed through the filter at 90 Hz (5,400
CPM). Hence, it must be recognized that the crossover between acceptance and
rejection of a signal is not an instantaneous event. It does occur over a finite frequency range. By increasing the sharpness to -48 dB/octave, the filter characteristics are imposed over a much narrower frequency range.
The presented phase data in Fig. 7-2 for the low-pass filter also reveals significant changes across the examined 100 Hz (6,000 CPM) frequency range. This
is normal behavior for this type of device; and failure to consider this characteristic can easily corrupt the final data interpretation. In all cases, it must be
acknowledged that filters may be successfully applied to reduce the complexity,
Electronic Filters
Fig. 72 Low-Pass Filter Set At 50 Hz
305
Fig. 73 High-Pass Filter Set At 50 Hz
or clean up a dynamic signal. However, the resultant amplitude and phase data
may be distorted. It is mandatory that the diagnostician be fully aware of the
characteristics of any applied electronic filters.
A low-pass filter is quite useful for eliminating high frequency interference.
For instance, a proximity probe signal may be passed through a low-pass filter to
minimize the influence of shaft scratches or other surface imperfections. This is
particularly significant when X-Y signals are displayed as a shaft orbit. A clean
pair of vibration signals will result in a shaft orbit that is definitive of the journal
centerline motion. Conversely, an orbit that includes multiple scratches might be
totally illegible (e.g., ball of string display).
The low-pass represents one of the basic types of electronic filters. The companion filter that exhibits the opposite frequency characteristics is a high-pass
filter. This type of filter only passes the higher frequencies (i.e., high-pass), and it
rejects the lower frequencies. By definition, this filter has a single transmission
band extending from a defined finite lower cutoff frequency (amplitude at -3 dB)
to the upper frequency of the device. The diagram in Fig. 7-3 depicts the behavior
of a 50 Hz (3,000 CPM) high-pass filter. Again, the filter was manually set at 50
Hz, and the 1.0 volt input signal was reduced to 0.711 volts at the set frequency.
This is very close to the predicted value of 0.707 volts (i.e., -3 dB = 0.707).
Once more it is noted that the filtered amplitude varies with frequency.
Even though the filter was set at 50 Hz (3,000 CPM), amplitude attenuation
occurs up to a frequency of 80 Hz (4,800 CPM). At the bottom end of the frequency scale, it is noted that 10% of the signal is still passed through the filter at
28 Hz (1,680 CPM). The phase data for the high-pass filter also reveals significant variations across the 100 Hz (6,000 CPM) frequency range. This is normal
behavior for this type of electronic filter. Once more, the machinery diagnostician
must consider this characteristic during data analysis.
A high-pass filter is quite useful for eliminating low frequency interference.
For example, double integration of an acceleration signal often produces substantial levels of low frequency noise. This condition may drive the significant
vibration data down into the noise floor of the analytical instrumentation due to
306
Chapter-7
the overwhelming level of the low frequency components. One way to avoid this
problem is to use a high-pass filter to eliminate the low frequencies from the
acceleration signal before integration. In some instances, a double high-pass filter might be necessary. One stage of the high-pass filter would be used on the
acceleration signal prior to integration, and the second stage would be used on
the velocity signal before integration to displacement.

Plus Low-Pass Filter Also Set At 50 Hz

Plus Low-Pass Filter Set At 70 Hz
Low and high-pass filters may be considered as the building blocks for other
filter types. For instance, the 50 Hz (3,000 CPM) low-pass and the 50 Hz (3,000
CPM) high-pass filter may be consecutively applied to the same signal as shown
in Fig. 7-4. In this combination, the manually tuned frequency of 50 Hz will be
subjected to a -3 dB attenuation from the low-pass, and another -3 dB reduction
from the high-pass filter. This combined attenuation of -6 dB is equivalent to a
50% voltage ratio, and that is exactly equal to the measured value of 0.500 volts
at 50 Hz (3,000 CPM). The phase change of the combined filter at 50 Hz is only 2. However, increased or decreased frequency deviation will result is significantly greater phase errors. Obviously, this 50 Hz low-pass and high-pass combination is an unrealistic filter configuration due to the overall signal attenuation
at all frequencies.
A separation of the tuned frequencies to a 30 Hz (1,800 CPM) high-pass
plus a 70 Hz (4,200 CPM) low-pass yields a much more useful filter combination.
This combination is displayed in the FRF in Fig. 7-5. This dual filter provides a
comfortable range for passing a band of frequencies with minimal signal reduction. Of course phase errors are encountered, and the magnitude of the phase
errors increase as the frequency varies from the center value of 50 Hz (3,000
CPM). This type of filter characteristic is commonly referred to as a band-pass
filter. In essence, a band of frequencies are passed by the filter, and the remainder of the signal frequencies are rejected. A filter of this configuration is very
useful for examination of a specific frequency component.
With respect to rotating machinery, the frequency of primary interest is
usually the machine running speed. Although many frequency components can
Electronic Filters
307
and do appear on process machines, the rotational speed motion should always
be scrutinized. Due to the potential influence of other vibratory sources, the
bandwidth of accepted frequencies should be reduced to allow examination of
only the running speed vibration. For instance, Fig. 7-6 displays a 2 Hz (120
CPM) band-pass filter from a Bently Nevada, Digital Vector Filter (DVF). This
data was produced by directing the DSA white noise signal into the DVF, and
connecting the DVF filtered output back into the second channel of the DSA as
shown in Fig. 7-1. The DVF was manually tuned to 3,000 RPM (50 Hz), and a frequency response function (FRF) performed between the filtered output signal
and the broad band noise input.
Fig. 76 Band-Pass Filter Set At 2 Hz
Fig. 77 Expanded 2 Hz Band-Pass Filter
This filter provides a close fit around a specific frequency component, plus a
minimal phase error at the center frequency. In addition, the DVF is designed to
allow the center frequency of the filter to be automatically tuned to coincide with
a Keyphasor pulse. This speed tracking characteristic of the DVF band-pass filter is extraordinarily valuable for measuring synchronous behavior during transient speed conditions. Some of the fundamental applications for this type of
filter include the generation of variable speed Bode and polar plots, plus the
examination of synchronous 1X response at a constant speed.
Amplitude accuracy is important for this type class of electronic filter. However, the 2 Hz (120 CPM) DVF band-pass filter presented in Fig. 7-6 exhibits a
maximum amplitude of 0.912 volts at the center frequency (i.e., 50 Hz). Since the
input signal is 1.000 volt, the apparent error of this filter approaches 9%. This
error would generally be considered as unacceptable, and the source of the deviation should be examined in greater detail. Fig. 7-7 represents another view of
this same DVF band-pass filter. In this diagram, the 0 to 100 Hz (0 to 6,000
CPM) frequency scale has been narrowed to a 25 Hz (1,500 CPM) bandwidth
extending from 40 to 65 Hz (2,400 to 3,900 CPM). This simple translation in frequency was performed within the DSA, and it has produced a fourfold improvement in frequency resolution (100 Hz versus 25 Hz bandwidth). Please note that
the four times improvement in resolution is directly coupled with a fourfold
increase in sample time.
308
Chapter-7
Examination of the 2 Hz (120 CPM) band-pass filter characteristics by

zooming in (Fig. 7-7) reveals an amplitude of 0.981 volts at the center of the filter. This value suggests an amplitude error of less than 2%. However, the filter
was not adjusted, only the frequency analysis range was changed. This behavior
suggests a difference in the DSA handling of the 100 Hz versus the 25 Hz analysis bandwidth. In actuality, the difference of a 2% versus a 9% amplitude error is
directly attributable to the coincidence of the 50 Hz DVF filter frequency with a
digital filter within the DSA. Also, a minimum of ten FFT frequency bins must
be contained within the half power bandwidth of the filter under test. If this
requirement is satisfied, then the amplitude accuracy will be 2% or better. If less
filter bins are contained within the half power bandwidth, the amplitude accuracy will suffer. Additional digital filter characteristics will be discussed in further detail later in this section. Prior to that discussion, it is desirable to
conclude the current review of band-pass filters.
As previously mentioned, the fundamental purpose of a band-pass filter is
to provide a narrow bandwidth filter around a specific frequency such as
machine running speed. The 2 Hz (120 CPM) bandwidth is quite adequate at
higher rotational speeds, but at low speeds, the fixed bandwidth might be excessive. For instance, a machine operating at 12,000 RPM (200 Hz) would be properly observed with a 120 CPM (2 Hz) band-pass filter. However, if the machine
was running at 200 RPM (3.33 Hz), the 120 CPM (2 Hz) filter width would not
provide the desired clean signal. In fact, the filter would include a good portion of
the frequency domain adjacent to the machine operating speed range.
To minimize this influence, the DVF and similar instruments are equipped
with a 12 CPM (0.2 Hz) bandwidth filter. The frequency response function (FRF)
shown in Fig. 7-8 depicts the characteristics of this filter between 0 and 100 Hz
Fig. 78 Band-Pass Filter Set At 0.2 Hz
Fig. 79 Band-Reject Filter Set at 2 Hz
(0 and 6,000 CPM). It is clear that the signal acceptance region is much tighter,
and the phase transition is much steeper than the 2 Hz (120 CPM) band-pass filter. Although it is not obvious from this data, the filter settling time for these two
filters varies by a factor of ten. That is, the 12 CPM (0.2 Hz) filter will require ten
times longer to settle on to a definitive amplitude and phase angle as a 120 CPM
Electronic Filters
309
(2 Hz) filter. Thus, vectors that change rapidly with respect to time should be processed with a 120 CPM (2 Hz) bandwidth filter.
Generally, the narrow 12 CPM (0.2 Hz) filter should be employed at speeds
below 1,000 RPM, and the wider 120 CPM (2 Hz) filter should be used for
machine operating speeds above 1,000 RPM. Obviously, there are exceptions to
this general rule, and the specific characteristics of the machinery should always
be considered when selecting filter bandwidths.
One final variation to the standard suite of electronic filters is presented in
Fig. 7-9. This 2 Hz (120 CPM) band-reject filter performs the opposite function of
a band-pass filter. That is, instead of passing only a small frequency range, this
type of filter rejects a small frequency range. For instance, the example reveals
that the tuned frequency of 50 Hz (3,000 CPM) has reduced the 1.000 volt input
level to 0.007 volts. In addition, the phase and amplitude effects upon the neighboring frequencies are minimal. This band-reject or notch filter is very useful for
eliminating a particular frequency from a dynamic signal. For instance, if a
machine is experiencing some minor subsynchronous motion, it may be advisable
to filter out (i.e., band-reject) the running speed vibration component. This filtration would allow better visibility of the sub-harmonic activity for detailed determination of the orbital precession plus other signal characteristics.
A band-reject filter is also quite useful for reducing the level of interference
on a dynamic signal. For example, if a measurement system displayed a 60 Hz
line frequency interference, a band-reject or notch filter might be used to attenuate or effectively remove the 60 Hz component. As with all filters, this type of
procedure does eliminate data. Thus, the application of a 60 Hz band-reject filter
might be totally appropriate for examining steady state data on a 10,000 RPM
turbine. However, a 60 Hz notch filter would be a poor selection for investigating
the full load behavior of a 3,600 RPM synchronous motor.
Electronic filters are an integral part of the tools used by the machinery
diagnostician. The previously discussed Krohn-Hite low and high-pass filters are
typical of the tunable analog filters available within the marketplace. This same
vendor also produces a line of digital filters that exhibit improved performance
over the analog variety. The automatic tracking band-pass and band-reject filters
available in the Bently Nevada DVF represent a digital filter configuration that
is mandatory for virtually any type of machinery analysis. However, when the
simultaneous examination of multiple frequencies is required, the diagnostician
must also employ an instrument with multiple filters.
An instrument equipped with multiple filters is commonly known as a
Dynamic Signal Analyzer (DSA), or a Fast Fourier Transform (FFT) analyzer. In
years past, this type of instrument was also referred to as a spectrum analyzer.
The device originally appeared in a single channel analog configuration that was
physically large, and weighed in excess of 150 pounds. Since 1970, it has evolved
into a smaller and lighter digital instrument. Currently, this device is available
as a portable, battery powered, 2 channel field instrument that weighs 5 to 10
pounds (e.g., HP-3560A). It is also common to find stand-alone 2 to 4 channel
instruments with substantial capabilities that weigh in the vicinity 20 to 30
pounds (e.g., HP-35670A). In addition, various multichannel computer interface
310
Chapter-7
configurations are also commercially available. These devices will be discussed in

greater detail in chapter 8 of this text.
Regardless of the specific configuration, or the user interface, all of the
DSAs share a common approach in signal processing. Basically, an analog input
signal is digitized, and subjected to a time domain windowing function. The frequency data could be considered as processed through a series of digital bandpass filters that are sequentially arranged in what is often referred to as a picket
fence pattern. A graphical representation of these sequential band-pass filters is
shown in Fig. 7-10.
Amplitude
Digital Filters
Fig. 710 Overlapping

Digital Filters In Dynamic
Signal Analyzer
Frequency
The analysis frequency bandwidth (span) of the DSA is divided by the number of digital filters to determine the frequency resolution. For instance, if a 0 to
100 Hz display contains 100 digital filters, the resolution is 1.0 Hz. Alternatively,
it may be stated that the filter spacing is 1.0 Hz (i.e., 100 Hz/100 filters = 1.0 Hz/
filter). The quantity of digital filters is generally referred to as the number of
lines or bins for the FFT. In this case, a 100 line display produces a low resolution
plot, whereas an 800 line spectrum would be considered as a high resolution display. Thus, a 0 to 100 Hz span spectrum with 800 digital filters has a resolution
of 100 Hz/800 filters = 0.125 Hz/filter.
The time necessary to acquire a block of data is determined by computing
the period of the particular frequency handled by the filter. For a resolution of
0.125 Hz/filter, the sample time is determined from equation (2-1) as:
1
1 Cycle
Sample Time = Period = ------------------------------- = --------------------------------------------- = 8.0 Seconds
Frequency
0.125 Cycle/Second
Obviously, if the sample time of the 100 line filter is subjected to this equation, the result would be a 1.0 second sample period. Thus, the data sampling
time is directly related to the number of filters. A 200 line display requires twice
as long to sample as a 100 line display. Similarly, an 800 line plot will take eight
times as long as a 100 line spectrum for the same frequency span.
Naturally, the frequency bandwidth selected will influence the sample time.
If a 6.25 Hz span is selected along with a 400 line display, the resolution is 6.25
Electronic Filters
311
Hz/400 lines = 0.015625 Hz/line. The sample time is the reciprocal of this value,
or 64.0 seconds to acquire a block of data. If high frequency data was to be examined, the sample time would be significantly reduced. For example, assume that
the number of lines remained at 400, and the span was increased to 25,600 Hz.
The resolution would be 25,600 Hz/400 lines = 64 Hz/line. Once again, the sample time is the reciprocal of this value, or 0.015625 seconds to acquire a block of
data. In all cases, low frequency bandwidths demand long sample periods, and
high frequency spans have short sample times.
A convenient way to directly associate frequency span with the length of
the time record and the FFT resolution is presented in Table 7-1. The frequency
span is listed in the first column. The associated time record length, and resolution are shown in the second and third columns respectively. This data is for a
400 line spectrum, and the presented values must be adjusted for any other frequency span range set on the DSA.
Table 71 Measurement Speed Versus Time Record Length And Resolutiona
Frequency Span
(Hertz)
Time Record Length

(Seconds)
Filter Resolution
(Hertz)
102,400
0.0039065
256
51,200
0.0078125
128
25,600
0.015625
64
12,800
0.03125
32
6,400
0.0625
16
3,200
0.125
1,600
0.25
800
0.5
400
200
0.5
100
0.25
50
0.125
25
16
0.0625
12.5
32
0.03215
6.25
64
0.015625
3.125
128
0.0078125
1.5625
256
0.00390525
0.78125
512
0.001953125
aValues
Listed Are For A 400 Line Display Only
312
Chapter-7
Viewed in another manner, for 100 lines at a 100 Hz bandwidth, the time
record would be 1 second long, and the resolution would be 1 Hz. That is:
100 Hz, at 100 lines, requires a 1 second time record, at 1 Hz resolution
This 100, 100, 1, 1 sequence is easy to remember. It may be quickly scaled
to any measurement configuration by multiplying or dividing the values by the
appropriate number of lines, and the span of set on the FFT. The values listed in
Table 7-1 are consistent with the previous discussion. In all cases, the lower frequency spans display high resolution combined with long sample times. Conversely, the higher frequency spans have very short time records, coupled with
increased filter resolution.
Additional time is required to process the data after sampling. For example,
processing time may be 12 to 14 milliseconds for a 1,024 block size. This
increases to 25 to 29 milliseconds for a 2,048 block size. The processing time for
each time record is normally added to the sample time. However, by using overlap processing the time records are overlapped and the FFT computation is
performed more frequently. Overlap processing reduces the measurement time,
and it uses the processor more efficiently. In addition, overlap processing will
recover some of the data lost in the windowing process. In all cases, the diagnostician must maintain a proper balance between resolution, data acquisition time,
duration of the event, and the machinery type under consideration.
The appearance and resolution bandwidth of the FFT filters is not only
dependent upon the selected number of filter lines, it is also controlled by the
type of window. This window is a time domain weighting function (filter) that is
applied to the input signal to remove spurious and non-periodic signals. These
digital filters do not influence the input time record, but they affect the displayed
information. Most DSA or FFT analyzers employ three basic filter shapes that
are referred to as the flat top, Hann, and uniform windows. A visual comparison
of these three window types is presented in Fig. 7-11.
The flat top window (also called a sinusoidal window) provides input signal
Flat Top
at -0.01 dB
Hann
at -1.5 dB
Fla
op
tT
n
Han
Uniform
Uniform
at -4.0 dB
Filter
Spacing
Frequency
Uniform
Fig. 711 Comparison Of

Uniform, Hann, And Flat
Top Filter Shapes And
Overlap Spacing Characteristics Between Filters
tT
op
nn
Ha
Amplitude
Fla
Electronic Filters
313
weighting plus a wide digital filter that results in exceptional amplitude accuracy. As shown in Fig. 7-11, the maximum amplitude error between digital filters
is only -0.01 dB. This translates to a voltage ratio of 0.9988, which may be
expressed as 99.88%. In other words, the component amplitude will be displayed
within -0.12% of the true value. Unfortunately, the frequency resolution suffers
with a flat top window, but the amplitude accuracy is unsurpassed. An additional
view of the flat top window is provided by the time domain and spectrum plots
shown in Fig. 7-12. In this example, a 50 Hz (3,000 CPM) sine wave with an
Fig. 712 Flat Top Window Characteristics In The Time And Frequency Domains
amplitude of 1.00 Volt,p-p is processed with a flat top window. It is noted that the
beginning and end of the time record are heavily attenuated, and the filter is
directed at the middle of the sample. This manifests as excellent amplitude accuracy, and the FFT plot in Fig. 7-12 reinforces the previous conclusion that the
minimum display amplitude will not be below 0.999 Volts,p-p of the 1.00 Volt,p-p
input signal. Again, this high amplitude accuracy is a function of the difference
of only -0.01 dB between adjacent filters.
The next filter displayed in Fig. 7-11 is a Hann window (also known as a
Hanning or Random window). This is the traditional filter shape found on the
original Real Time Analyzers. It provides a good compromise between the amplitude accuracy of the flat top window, and the frequency resolution inherent with
a uniform window. The Hann window is the most commonly used window for
vibration analysis, and random noise measurements.
As shown in Fig. 7-13, the Hann window attenuates the input signal at both
ends of the sampled time record, and it forces the signal to appear periodic. The
maximum error between Hann filters is typically -1.5 dB. This translates to a
voltage ratio of 0.8414, which may be expressed as 84.14%. Thus, a 1.0 Mil,p-p
signal would be displayed as 1.0 Mil,p-p if the component frequency was coincident with a filter bin. However, if the frequency was located precisely between
two filters, the displayed amplitude would be 0.84 Mils,p-p, which is equivalent to
an amplitude error of 16%.
314
Chapter-7
Fig. 713 Hann Window Characteristics In The Time And Frequency Domains
For additional explanation, the behavior of a Hann window is depicted in

Fig. 7-13, where a 50 Hz (3,000 CPM), 1.00 Volt,p-p sine wave was processed. In
this case, the input voltage was subjected to the potential of a 16% amplitude
reduction, and the minimum voltage of 0.841 Volts,p-p was noted in the spectrum
plot. Again, this potential for an amplitude error with a Hann window is a function of the difference of -1.5 dB between adjacent filters.
The third type of window is called a uniform (or a transient) window. This
particular window weighs all parts of the sampled time record equally. In
essence, the Uniform window does not influence or manipulate the input time
record. This is demonstrated in Fig. 7-14 where the 50 Hz (3,000 CPM) sine wave
displays equal amplitudes of 1.00 Volt,p-p throughout the entire time record.
Some of the technical literature refers to this type of window as a rectangular
shape, and other references consider this to be a non-window.
A Uniform window is very useful for examining short duration functions
that are basically self-windowing. The force hammer input, and the accelerometer output signals acquired during an impact test are good examples of these
Fig. 714 Uniform Window Characteristics In The Time And Frequency Domains
Electronic Filters
315
types of functions. For examination of periodic signals, the maximum amplitude

uncertainty between Uniform filters is approximately -4 dB as previously shown
in Fig. 7-11. This is equivalent to a voltage ratio of 0.6310, or this may be
expressed as 63.10%. Hence, if a component falls directly between Uniform filters, the amplitude error could approach 37% as shown in the spectrum plot of
Fig. 7-14. Obviously, this is unacceptable for most routine vibration measurements, and one of the other two windows would normally be selected.
For improved handling of transient events, two additional windows known
as the force and the exponential windows are in residence within most modern
DSAs. The force window passes the first portion of the time record, and sets the
remainder of the record to some specific value (e.g., zero). This is very useful during hammer impact tests, since this window provides the capability to delete all
electronic noise from the force signal past the initial impulse. Truncation in this
manner provides a cleaner signal from the impact hammer, and allows for
improved clarity of the final frequency response function (FRF).
The exponential window attenuates the input response signal at a decaying exponential rate. The actual rate of signal decay is determined by a user
defined time constant. Again, this type of window is very useful during hammer
impact tests to provide an even exponential decay to the vibration response signal. The validity of applying this type of filter shape was previously demonstrated by equation (2-71) and the companion Fig. 2-14. From this earlier
analysis of a simple mechanical system, it was shown that forced vibration will
decay in an exponential manner. Hence, this filter shape is quite appropriate. It
also should be mentioned that an exponential window does not attenuate the
response in the region that contains the best signal to noise ratio, but it does
gradually taper off the signal in the time domain where the signal to noise ratio
is the lowest. Overall, this window contributes to the elimination of spurious signals, and it improves the accuracy plus clarity of the FRF.
In the overview, it is clear that a variety of electronic filters are available to
the machinery diagnostician. Each instrument and each individual filter configuration possesses specific characteristics that may be used to enhance the measured dynamic data, and allow identification and resolution of the mechanical
behavior. Conversely, these same filters may totally obliterate the useful data if
they are improperly applied. Hence, the diagnostician must become intimately
familiar with the filter characteristics on all instruments used for machinery
analysis. The reader is also encouraged to study more detailed explanations of
FFT signal manipulation in documents such as The Fundamentals of Signal
Analysis1, and chapter 14 of the Shock and Vibration Handbook2. Many other
references are also available, but care should be taken to insure that the selected
educational materials are consistent with the machinery to be tested, and the
selected array of instrumentation.
1 The Fundamentals of Signal Analysis, Application Note 243, (Hewlett Packard, printed in
USA, 1995).
1996), chap. 14.
316
Chapter-7
TIME AND ORBITAL DOMAIN

Rotating machinery and associated perturbations usually follow shaft rotation, which always moves in a direction of increasing time. In virtually all cases,
careful examination of the time domain behavior of a machine is fundamental to
the analysis of that equipment. Since the other dynamic measurements are
extracted from the time domain motion, the full understanding of this topic is
essential. It is desirable to begin this discussion with an example of how vibratory motion of a machine element is translated into a time domain voltage signal. In this regard, consider Fig. 7-15 of a rotor turning in a clockwise direction,
with a vertical proximity probe observing the shaft. A series of eight angular
locations are identified with the letters A through H. These positions are stationary reference locations around the circumference of the rotor (not be confused
with the angular coordinate system on the rotating element). To state it another
way, locations A through H represent the actual shaft positions as the rotor
turns, and follows the circular dotted line path as a function of time.
In this example, presume that the rotor contains a single mass unbalance
as indicated by the black dot on the eight shaft positions. Also, assume that this
simple machine operates below the rotor natural frequency. It is self-evident that
centrifugal force due to the unbalance causes a radial deflection (bow) of the
shaft in the direction of the mass unbalance. In essence, the effective rotor heavy
spot is thrown to the outside of the orbit at all angular locations. This is the same
mechanism that was previously discussed in chapter 3 on the Jeffcott rotor. The
rotational behavior depicted in Fig. 7-15 is identical to moderate speed operation
of the Jeffcott rotor as shown in Fig. 3-30. As the shaft rotates, the heavy spot
will pull the rotor into a circular orbit if the stiffness is identical in both vertical
and horizontal directions. With assumed identical stiffness values, the resultant
circular orbit is depicted by the dotted line in Fig. 7-15.
Clearly, the heavy spot (unbalance mass) is identical and coincident with
the high spot (peak displacement). An actual physical example of this type of
Rotating Shaft
Orientation
at 45
Increments
A
H
1.0
Unbalance
Mass
R o t a ti o
1.5
Min. DC Voltage
B
W
0.5
0.5
0
C
Ti m e
F
Shaft Orbit
-0.5
Max.
DC
-1.0
F
Fig. 715 Output From A Vertical Proximity Probe Observing A Rotating Shaft
Vertical Displ. (Mils)
Stationary
Observation
Points at 45
Increments
Time and Orbital Domain
317
rotor would be a truck drive shaft. If the rear wheels of the truck are jacked up
off the ground, and the drive shaft runs at a moderate speed, the heavy spot may
be determined by carefully bringing a piece of chalk up to the rotating shaft.
When the chalk makes contact with the shaft, the high spot is marked. A balance
weight may be attached onto the shaft 180 away from this location. Fig. 7-15 is
representative of this physical behavior. However, in this diagram an accurate
distance sensing device in the form of a proximity probe is used instead of the
piece of chalk. As discussed in chapter 6, the proximity probe converts distance
between the probe and the observed surface into a DC plus an AC output voltage.
When the shaft rotation allows the high spot/heavy spot to be in a true vertical position, the output DC voltage from a vertical proximity probe will be minimal (smallest gap). This specific condition is defined by point A on Fig. 7-15. As
the shaft continues to turn in a clockwise direction, reference points B, C, D, etc.
expose a different shaft surface to the probe. In each case, the high spot continues to move further away from the probe until position E is attained. At this
point, the shaft has arrived at the farthest distance from the probe tip, and the
probe DC gap voltage will be at a maximum.
Continuing in 45 increments through stationary observation points F, G,
and H, the shaft returns back to the starting point A. The associated time
domain plot presented in Fig. 7-15 describes the vertical motion of this rotor with
respect to time. In this diagram, time progresses from left to right (same as an
oscilloscope); and the vertical displacement is shown to vary between 1.0 Mil,
which is equal to the differential DC voltages divided by the probe sensitivity. In
accordance with the convention established at the beginning of this text, the
total shaft motion would be expressed as 2.0 Mils,p-p. Furthermore, connecting
the eight data points into a continuous curve reveals the repetitive (cosine)
nature of this simple example.
The variation in DC voltage is generally considered to be an AC voltage
that is proportional to the vibration of the observed surface. Some individuals
will argue that this is not a true AC voltage, but an oscillation of a DC level.
Although this distinction may be technically rigorous, it is both convenient and
appropriate to consider a proximity probe output voltage as consisting of an AC
voltage superimposed upon a DC voltage. In all cases, the AC voltage is proportional to the relative vibration of the observed surface. The DC voltage is proportional to the average distance between the probe tip and the observed surface.
Other dynamic transducers such as accelerometers generate an AC voltage
that is proportional to the motion of the vibrating surface. These transducers
may exhibit an output DC voltage, but that is usually associated with the transducer power supply. Within the typical suite of vibration transducers, the proximity probe is the only sensor that provides static position information via the
DC voltage. For additional discussion of specific transducer characteristics, the
reader is referred back to chapter 6.
The AC portion of a proximity probe signal is directly proportional to the
relative vibration between the probe support element (e.g., bearing housing) and
the observed surface (e.g., rotating shaft). The AC portion of the signal produced
by a casing mounted vibration transducer (i.e., velocity coil or accelerometer) is
318
Chapter-7
proportional to the absolute vibration of the surface to which the transducer is

attached (e.g., bearing housing). Within the time domain, the dynamic signal
characteristics of the various transducers may appear to be very similar, and the
same concepts of time domain observation would apply. Please note that this conclusion does not insinuate that shaft and casing motion are identical. On the contrary, there are normally discernible differences between the two measurements.
For instance, a simple machine with residual unbalance and fluid film bearings
will display casing signals that lag behind the shaft motion. In addition, the casing vibration amplitudes will normally be attenuated by the bearing housing. On
a machine with rolling element bearings, the relative shaft motion will be minimal, and casing vibration will be dominant. Obviously, machines with multiple
excitations will exhibit more complicated deviations between shaft and casing
vibration measurements.
Although machinery shafts and casings move with distinctly different characteristics, it is common knowledge that all machines vibrate vertically, horizontally, and axially. Typically, the axial vibration is minimal, and the majority of
the vibration is observed in the lateral (i.e., radial) planes. This natural vertical
and horizontal vibratory motion is typically observed with orthogonally mounted
vibration probes. For instance, the inclusion of a horizontal proximity probe into
the arrangement discussed in Fig. 7-15 results in the X-Y configuration shown in
Fig. 7-16. The additional horizontal probe now provides the capability to view a
two-dimensional response of the shaft surface. As discussed earlier in this text,
the probes may be in any circumferential location. However, they must be 90
apart to provide an accurate two-dimensional representation of the shaft motion.
Considering the horizontal displacement in Fig. 7-16, it is clear that when
the shaft rotation allows the high spot to be in a true horizontal position, the output DC voltage from a horizontal proximity probe will be minimal. This condition
Rotating Shaft
Orientation
at 45
Increments
A
C
T im e
Unbalance
Mass
D
E
1.5
0
-0.5
Max.
DC
0.5
-1.0
Shaft Orbit
at
2.0 Mils,p-p
Horizontal Displ. (Mils)
R ot atio
Min. DC Voltage
0.5
1.0
Stationary
Observation
Points at 45
Increments
1.0
Min.
DC
0.5
0
-0.5
-1.0
0.5
1.5
Max. DC Voltage
Fig. 716 Output From Orthogonal Proximity Probes Observing A Rotating Shaft
319
is defined by point C in Fig. 7-16. As the shaft continues to turn in a clockwise

direction, reference points D, E, and F, expose a different shaft surface to the
probe. In each case, the high spot/heavy spot continues to move further away
from the probe until position G is attained. At this point, the shaft is at the
greatest distance from the probe tip, and the proximity probe DC gap voltage will
be at a maximum.
Continuing in 45 increments through stationary observation positions H,
A, and B, the shaft returns back to the starting point C. Again, time progresses
from left to right, and the horizontal displacement varies between 1.0 Mil, or
2.0 Mils,p-p. Connecting the eight data points into a continuous curve reveals the
repetitive (sine) nature of this example. If these vertical and horizontal signals
were viewed on an oscilloscope, they could be converted into an orbit by selecting
the X-Y sweep position, and the display shown in Fig. 7-17 would appear.
Fig. 717 Simple Circular

Orbit From Orthogonal
Proximity Probes
Not surprisingly, this orbital display is identical to the original shaft motion
of 2.0 Mils,p-p. Since this example was established as a forward and circular
motion, the resultant shaft orbit must also be both forward and circular. By
visual inspection, it is clear that the resultant orbit is circular. By virtue of the
fact that this orbit was developed in a clockwise direction, it is therefore moving
with increasing time (i.e., forward). However, on a real machine the determination of the precession of an orbit is somewhat more complicated.
Specifically, the time and orbital domain signals are superimposed with a
pulse signal originating from a once-per-rev Keyphasor. As previously discussed in chapter 6, the Keyphasor timing pulse will produce a blank-bright
pattern for a negative going signal pulse such as a notch or a drilled hole in the
shaft (e.g., Fig. 6-5). If the Keyphasor probe is looking at a section of key stock
or other projection, the resultant trigger pulse will be a positive going signal, and
the oscilloscope blanking will appear as bright-blank sequence (e.g., Fig. 6-6).
In order to explain the interaction of the Keyphasor pulse with the time
domain traces, it is useful to expand upon Fig. 7-16 by adding a Keyphasor
probe to the diagram. Assume that a notch is milled into the shaft at an axial
location that is several inches away from the radial probe measurement plane.
320
Chapter-7
315
Stationary
Observation
Points at 45
Increments
Rotating Shaft
Orientation
at 45
Increments
A
H
R ot atio
C
T im e
D
E
Keyphasor
Probe
0
-0.5
1.5
-1.0
Shaft Orbit
at
2.0 Mils,p-p
Horizontal Displ. (Mils)
0.5
0.5
1.0
1.0
0.5
45
0
-0.5
0.5
1.5
-1.0
Fig. 718 Time Domain Output Signal Outputs From Orthogonal Proximity Probes And
Radial Keyphasor Probe Observing A Rotating Shaft
Further assume that this shaft notch is located diametrically opposite (180
away) from the identified heavy/high spot. Physically, the Keyphasor probe is
added to the lower left hand corner of the angular reference frame as depicted in
Fig. 7-18. Note that the timing probe is positioned at the F location, but the only
portion of the rotational cycle that exposes the notch to the Keyphasor probe
occurs when the shaft is at the B position. Hence, the Keyphasor probe will
generate a negative going pulse as the shaft rolls past position B. If this timing
pulse is superimposed upon the vertical and horizontal vibration signals, a
blank-bright sequence will appear at the B location. This situation is graphically
shown in Fig. 7-18.
Based upon the standard phase convention presented in chapter 6, the
phase angles for vertical and horizontal probes are identified in Fig. 7-18. In each
case, the peak of the vibration signal is used as the zero point, and the angle is
determined by moving backward in time to the trigger point of the Keyphasor
pulse. In this specific case, each time division is equal to /4, or 45 (derived from
1 revolution = 2 radians = 360). It is clear that the difference between the peak
of the horizontal vibration signal and the K is one division, or 45. The vertical
probe has seven divisions from the peak amplitude to the K pulse. Thus, the
vertical phase angle is 315 (= 7 Divisions x 45/Division).
The physical relationship between these phase angles and the rotating system is illustrated by the drawings in Figs. 7-19 and 7-20. Specifically, Fig. 7-19
describes the relationship between the vertical probe, the Keyphasor probe, the
rotating notch, and the heavy/high spot. When the leading edge of the notch is
321
located under the Key probe, the negative trigger pulse is initiated. At this
exact point, the vertical probe 315 phase angle may be used to locate the heavy/
high spot on the rotor. Since shaft rotation and time are moving in a clockwise
direction, the phase lag (backward time) must be in a counterclockwise direction
as established in Fig. 2-3. Furthermore, since zero degrees (0) is always located
at the probe, the phase angles are always measured in counter rotation from the
probe. Hence, moving 315 counterclockwise from the vertical probe locates the
heavy/high spot in the upper right hand quadrant of Fig. 7-19.
Probe
@ 0
High
Spot
High
Spot
90
Tim e
Time
45
90
Heavy
Spot
CW
R o t a ti o
270
Heavy
Spot
180
CW
R o t a ti o
Probe
@ 0
315
180
270
Keyphasor
Keyphasor
Fig. 719 Phase Angle For True Vertical

Proximity Probe to Heavy/High Spot
Fig. 720 Phase Angle For True Horizontal Proximity Probe to Heavy/High Spot
The diagram in Fig. 7-20 displays an identical relationship for the horizontal probe. For this transducer, the 45 phase angle is counter rotation from the
horizontal probe (again 0 reference for each probe). The same logic applies, and
the same heavy/high spot is identified on the rotor. Naturally, this will only be
precisely the same point when the orbit is forward and circular. This condition
would be visually apparent, and the phase angles would exhibit a 90 difference.
In cases when the specific mechanics of the Keyphasor installation are
unknown, the pulse signal can always be viewed on an oscilloscope to determine
if the trigger is negative or positive going. If the blanking characteristics of the
oscilloscope are unknown, then the superposition of the Keyphasor signal upon
a time base wave form will identify the actual sequence. In all cases, oscilloscopes sweep from left to right, and the identification of blank-bright or brightblank with respect to time may be visually determined on an analog scope.
As applied to a shaft orbit, knowledge of the timing pulse sequence will
allow correct identification of the orbit precession. Again, the real key to this
identification is the fact that the measured parameters are moving forward in
time. Thus, when the time sequence is identified, the orbit precession is also
defined. For instance, consider Fig. 7-21 that shows an array of four circular
322
Chapter-7
Fig. 721 Variation Of Keyphasor Blank-Bright Sequence With Rotation And Pulse Type
orbits combined with negative and positive going Keyphasor pulses for both
clockwise and counterclockwise shaft rotations.
Plots A and C in Fig. 7-21 exhibit the influence of a negative going Keyphasor trigger pulse. In Plot A, the blank spot is followed by the bright spot in a
clockwise direction. This sequential behavior is fully descriptive of a clockwise
shaft precession, which in this simple example is the same as the shaft rotation.
It should be mentioned that some complex machinery instabilities display a nonsynchronous frequency that has a precession that is opposite to shaft rotation.
Without the Time-Orbital-Keyphasor sequencing concepts presented herein,
these reverse precession malfunctions could not be properly diagnosed.
Plot C in Fig. 7-21 reveals a reversal of the blank-bright sequence, which
must be interpreted as a change in rotation to counterclockwise. Once more, the
first event is the blank spot (negative slope), followed in a counterclockwise
direction by the bright spot (positive K slope). Hence, the precession (rotation)
for the orbit shown in Plot C must be counterclockwise.
The two orbits shown in Plots B and D of Fig. 7-21 represent the shaft rota-
323
tional directions viewed with a positive going Keyphasor pulse. As shown in

Fig. 6-6, this type of trigger pulse will display a bright spot followed by a blank
spot. Again the same logic applies, and the orbit in Plot B may be easily identified as clockwise. Clearly, the bright-blank sequence identifies Plot D as a shaft
rotating in a counterclockwise direction.
As a final comment on this four orbit array in Fig. 7-21, note that the diagonal orbits appear to be identical. That is, the clockwise rotating shaft with a negative Keyphasor pulse (Plot A) looks just like the counterclockwise rotating
shaft with a positive pulse (Plot D). Similarly, the counterclockwise rotating
shaft with a negative pulse (Plot C) is identical to the clockwise rotating shaft
with a positive Keyphasor (Plot B). This type of visual similarity, with totally
different mechanical implications is common throughout this machinery business. Additional examples of similar appearances combined with different
mechanical characteristics will be presented in the remainder of this chapter.
However, the machinery diagnostician should not take things for granted, and
must always verify that the output information is totally consistent with the
physical machinery configuration.
In all cases, determination of the specific trigger point for the phase angle
measurements is dependent upon the trigger slope (positive or negative), and the
direction of the trigger pulse (positive or negative going). This actual trigger
point for the Keyphasor probe is especially important during activities such as
field balancing. Hence, the relationship between the rotor trigger point and the
Keyphasor probe will be discussed in greater detail at the beginning of the
rotor balancing chapter 11.
The example under discussion has consisted of a forward, circular orbit,
that occurs at a single frequency. The circular orbit implies equal amplitudes,
plus a 90 phase differential between vertical and horizontal signals. The
described response is representative of a machine operating with a single excitation (e.g., unbalance), and with an equal support stiffness in both the vertical
and horizontal planes. Unfortunately, process machinery seldom behaves in such
an idealistic manner. As stated in chapter 6, a dynamic signal is characterized by
the following basic parameters:
Amplitude (Magnitude or Severity)

Frequency (Rate of Occurrence)
Timing (Phase Relationship)
Shape (Frequency Content)
These fundamental characteristics were discussed for a single dynamic signal. It is understandable that the combination of two dynamic signals into an
orbit may significantly complicate the resultant pattern due to variations
between the signals. Hence, it is desirable to examine the interaction between
variations in amplitude, frequency content, and phase difference between signals. For instance, if the previous horizontal response was reduced by a factor of
two, the vertical orbit displayed in Fig. 7-22 would appear. This 2:1 ratio in vertical versus horizontal machine response could be indicative of a 2:1 stiffness ratio.
324
Chapter-7
Fig. 722 Vertical Orbit With 2:1 Ratio

Between Vertical & Horizontal Amplitudes
Fig. 723 Horizontal Orbit With 2:1 Ratio

Between Horizontal & Vertical Amplitudes
A horizontal preload (e.g., misalignment) that restricts the horizontal motion

would also produce a predominantly vertical orbit. In either case, the resultant
orbits would appear to be similar, but the mechanics responsible for the deviation in horizontal amplitudes would be considerably different.
If the amplitude ratio is reversed, such that the horizontal vibration is
twice the size of the vertical response, the orbit shown in Fig. 7-23 emerges. This
type of orbital pattern is quite common for large turbines equipped with elliptical bearings. In these units, the horizontal clearance is often twice the vertical
clearance, and the associated differential in bearing stiffness allows the horizontal vibration to exceed the vertical vibration. Again, a vertical preload from a
cocked bearing, or a misaligned shaft, could also restrict the vertical motion, and
produce a horizontal ellipse. In all cases, the resultant vibration or motion at a
bearing is dependent upon the applied force(s) and the associated stiffness. This
general relationship between the vibration Response, the applied Force, and the
Restraint (stiffness) was previously stated in equation (3-31) as:
Force
(7-1)
The smaller the restraint, the larger the vibration response for a given unit
force input. Similarly, increased force with a constant restraint will result in an
increased vibration amplitude. However, for a two-dimensional mechanical system, there will be cross-coupling between horizontal and vertical directions.
Thus, an applied force in one direction may result in motion in a perpendicular
plane. This mechanical reality does complicate the interpretation of the vibration
response data, but the orbital presentation provides quantification of actual journal motion within a bearing.
Clearly, vibration amplitudes in the time and orbital domains will be influenced by a combination of the system forces and restraints. It should also be recognized that the system parameters are often vector quantities. These vectors
may appear as constant amplitudes, and mechanical changes may influence or
325
Fig. 724 Variation Of Orbits With Changes In Phase Holding Vertical And Horizontal
Amplitudes and Frequencies Constant
cause variations in the relative timing between signals. This will manifest as
phase angle changes between the measured vertical and horizontal vibration
response vectors. The time domain data will be shifted, and the orbits may be
subjected to significant changes. In order to fully appreciate the potential variation in orbital patterns due to phase angle changes, consider the array of four
orbits shown in Fig. 7-24. In each case, the amplitudes and the frequencies are
identical for both the vertical and the horizontal transducers. The only difference
between signals is the time domain phase differential. The orbit presented in
Plot C is identical to previous examples where a cosine function is plotted
against a sine function (90 difference). The elliptical orbit displayed in Plot B
was constructed with a 45 phase difference between the vertical and horizontal
signals. If the differential signal timing was changed to 135, the elliptical orbit
would lean to the left instead of to the right.
The same type of characteristic is displayed by the straight line orbits. The
display in Plot A of Fig. 7-24 has zero phase difference between channels, and the
326
Chapter-7
orbit shown in Plot D has a 180 offset. In each case the closed orbital loop has
collapsed into a straight line. When the line leans to the right the signals are
directly inphase, and when the line leans to left, the signals are exactly out of
phase. For this case of identical frequencies and amplitudes between channels
the orbits will unfold from a straight line into a circle as a series of ellipses as the
phase angle rolls by 90.
The next example in Fig. 7-25 is complicated by the fact that the horizontal
frequency occurs at precisely twice the vertical frequency. In the case of a heavy
radial preload (e.g., severe misalignment), the vertical probe signal might be
completely dominated by 1X running speed motion, and the horizontal probe signal might exhibit a major frequency component at twice running speed (2X).
Considering this 2:1 ratio of frequencies, the orbits in Fig. 7-25 are generated as
the phase angle differential is varied between 0 and 90. Three of the orbits for
this 2:1 frequency ratio are various Figure 8 shapes, which is consistent with a
vertical preload. This concept of preloads will be discussed in more detail in
chapter 9. Note that the 45 phase offset in Plot C of Fig. 7-25 has resulted in
Amplitudes Constant and Setting A 2:1 Ratio Between Horizontal and Vertical Frequencies
327
another line orbit that resembles a parabola. Again, this is a closed orbit similar
to the previously discussed straight line orbits. It appears as a parabola strictly
due to the time domain relationship of the two signals. As the differential phase
increases past 45 a series of mirror images are generated. For instance at a 135
phase difference between signals, the parabolic orbit becomes inverted, and the
apex points downward.
Frequency ratios of 2:1 are occasionally observed on process machines. But
pure frequency ratios of 3:1 between orthogonal probes viewing the same shaft
surface are seldom encountered. However, for academic purposes, the four orbits
presented in Fig. 7-26 were produced with a horizontal frequency equal to three
times the vertical frequency. Plot C with a phase difference of 30 describes a
three lobed shape with two crossover points. This symmetrical orbit becomes distorted as the phase angle varies. In fact, the Lazy S patterns shown in Plots A
and D are the equivalent of the 1:1 ratio straight line, and the 2:1 ratio parabola
previously discussed. Further extensions in frequency ratio between vertical and
horizontal probes would increase the number of crossover points, and the com-
Amplitudes Constant and Setting A 3:1 Ratio Between Horizontal and Vertical Frequencies
328
Chapter-7
plexity of the closed orbits. However, higher order frequency ratios between
orthogonal probes are rarely encountered on real machines. The more common
variety of frequency variation in a shaft vibration signal is the appearance of a
frequency component that appears in both vertical and horizontal channels. For
example, Fig. 7-27 displays sub rotative speed vibration components set at 50%
of rotative speed. These half speed components were superimposed upon both
horizontal and vertical synchronous signals.
Fig. 727 Variation Of Orbits With Subsynchronous Frequency At 50% Of Rotative Speed
Combined With Changes In Amplitude Ratio Between Subsynchronous & Rotative Speed
Plot A in Fig. 7-27 was generated with a 1:1 amplitude ratio between the
synchronous, and the subsynchronous components. This type of orbit with an
inside loop would be representative of a half speed excitation that is equal in
magnitude to the rotative speed. If the amplitude of the subsynchronous component increased, the single inside loop would fade away, and a generally circular
orbit would eventually appear. For instance, the orbit in Plot B of Fig. 7-27 shows
the influence of increasing the 50% component size to four times the amplitude of
329
the fundamental rotational speed. A slight flat spot appears at the bottom of this
orbit, however further increases of the subsynchronous amplitude will produce
an increasingly circular appearing orbit.
If the half speed component decreased in amplitude, and the rotational
speed motion remained constant, overall vibration amplitudes would drop.
Clearly, the relative size of the inside loop would expand with respect to the size
of the outside loop. An example of this condition is shown in Plot D, where the
magnitude of the half speed frequency was decreased to be only one quarter
(25%) of the amplitude of the running speed component. An intermediate condition of a half amplitude, half frequency component is presented as Plot C in Fig.
7-27. Hence, the appearance and relationship between inside and outside loops is
really dependent upon the amplitude ratio between the constituents.
During orbital analysis a Keyphasor pulse should always be incorporated.
This Z-axis input is invaluable for determining precession of the overall orbit
plus each of the filtered components. The Keyphasor will also help establish
frequency ratios. For instance, if a pair of X-Y probes display an orbit with two
fixed Keyphasor dots, the subsynchronous excitation occurs exactly at 50% of
rotative speed. However, if the dots are moving around the orbit, the subsynchronous frequency is not locked in at exactly one half of running speed. Under this
condition, the subsynchronous frequency could be 55%, 49%, 43%, or any noninteger fraction of rotative speed. This specific behavior is significant, and it will
be examined in greater detail in chapter 9 of this text.
Additional combinations of subsynchronous amplitude and frequency ratios
are presented in Fig. 7-28. Plot A depicts a condition of equal synchronous to subsynchronous vibration amplitudes (i.e., 1:1). However, in this case, the low frequency component is set at 75% of the rotative speed. Note that multiple loops
are generated, and direct interpretation of this behavior might be very difficult.
In addition to the need for a synchronous Keyphasor pulse, the machinery diagnostician should also employ a Dynamic Signal Analyzer (DSA) to assist in specific component identification.
Plot B in Fig. 7-28 displays a 50% subsynchronous frequency combined with
a 2:1 amplitude ratio between the subsynchronous and the rotative speed frequency. By direct observation of Plot B on an oscilloscope, without the benefit of a
Keyphasor pulse, the viewer might conclude that only one frequency was
present in the vibration signals. This type of erroneous conclusion may be
avoided by careful examination of the data using orbital, time, and frequency
domain analysis with a once-per-revolution trigger pulse. It should be recognized
that the examination of vibration signals with various data formats provides the
diagnostician with better visibility of the machinery behavior, and reduces the
possibility of missing one or more key elements in the response characteristics.
Plots C and D in Fig. 7-28 describe the orbits resultant from increasing the
amplitude of the sub-harmonic component, plus decreasing the frequency ratio.
In Plot C the subsynchronous frequency was set at one quarter (25%) of the rotative speed frequency. The sub to 1X amplitude ration was maintained at 2:1 to be
consistent with Plot B in this same Fig. 7-28. Note that with the same amplitude
ratio, the apparently simple orbit of Plot B has evolved into a complex orbit with
330
Chapter-7
Fig. 728 Variation Of Orbits With Different Amplitude And Frequency Ratios Between
The Subsynchronous Component And The Rotative Speed Component
three defined inside loops (Plot C). Furthermore, if the subsynchronous excitation is reduced to one tenth (10%) of the running speed, and the amplitude ratio
is adjusted to 5:1, Plot D in Fig. 7-28 evolves. The nine inside loops appear well
defined, and there might be a tendency to establish hard and fixed rules for the
number of loops versus the subsynchronous frequency. Specifically, it has been
touted that the number of inside loops plus one is equal to the ratio between the
subsynchronous and synchronous vibration frequency. Although this is correct
for Plots C and D, this type of general observation can lead to some significantly
wrong conclusions. The validity of this statement will be substantiated when
reviewing of the final group of orbits in Fig. 7-29.
This last set of calculated orbits in Fig. 7-29 addresses the appearance of
supersynchronous vibration components combined with the fundamental running speed response. In general, low frequency subsynchronous vibration compo-
331
nents have a tendency to exhibit amplitudes that are larger than running speed
motion. In some instances, the subsynchronous activity may occupy the entire
bearing clearance. Conversely, frequency components that occur above shaft
rotative speed tend to display amplitudes that are smaller than the fundamental
1X running speed vibration levels. For discussion and demonstration purposes,
the orbits in Fig. 7-29 were constructed with supersynchronous amplitudes that
are a fraction of the running speed 1X vibration amplitude.
Fig. 729 Variation Of Orbits With Different Amplitude And Frequency Ratios Between
The Supersynchronous Component And The Rotative Speed Component
The orbit in Plot A of Fig. 7-29 exhibits the influence of a frequency at twice
rotative speed (2X), with an amplitude set at one half of the fundamental. Note
that the orbit shows signs of this potential preload frequency distribution by
indentation at the bottom of the orbit. Similar patterns will emerge when orbits
with a 50% frequency are compared against orbits with a 2X frequency. Again,
the machinery diagnostician must be properly equipped with a Keyphasor signal plus an oscilloscope and a frequency analyzer to be absolutely sure of the
332
Chapter-7
proper frequency and amplitude relationships of the individual components

within each dynamic signal.
Plot B in Fig. 7-29 displays a supersynchronous frequency of five times rotative speed, combined with an equal amplitude ratio. The symmetrical rosette
shape is interesting, but difficult to interpret. Again, a Keyphasor pulse plus a
DSA would be mandatory for proper analysis. This type of behavior may also be a
candidate for eliminating the running speed vibration from both signals with a
band-reject filter set at rotational speed. Another approach for analysis of this
type of orbit would be the application of a band-pass filter set at five times rotative speed. In either approach, the 1X signal must also be examined, and the
interaction between frequencies must be documented.
The orbit display Plot C of Fig. 7-29 describes a traditional pattern for a
machine with a synchronous high frequency excitation. In this case, the high frequency motion was set at ten times rotative speed, and the amplitude was
reduced to one tenth of the fundamental running speed vibration. This combination produces a nice scalloped pattern, with a series of nine external loops. It is
easy to be fooled into believing that external loops are indicative of high frequency components, and internal loops are representative of low frequency components. In actuality, neither conclusion is totally correct.
The appearance of outside versus inside loops on an orbit is strictly dependent upon the amplitude ratios, and the frequency ratios of the components
within each dynamic signal. To demonstrate this point, the 10X frequency in Plot
C was replotted with an amplitude increase from 0.1 to 0.2. The resultant data is
shown in Plot D of Fig. 7-29. Note the significant difference between the two
orbital plots presented with a 10X component. In fact, the plot with the 10%
supersynchronous excitation in Plot D of Fig. 7-29 is visually identical to the plot
with a 10X subsynchronous component shown in Plot D of Fig. 7-28. Hence, a
casual examination of the two orbits (without Key) does not tell the diagnostician whether they are dealing with a high amplitude subsynchronous problem,
or a low amplitude supersynchronous vibration problem.
Clearly, the frequency content of the dynamic signals, the ratio between the
dominant component amplitudes, and the phase relationship between signals,
plus the timing (phase) between individual components will determine the time
domain characteristics of the vibration signals. These specific time domain characteristics will govern the shape and pattern of the final orbit(s). In virtually all
cases, the complexity of the dynamic signals will dictate the amount and kind of
diagnostic instrumentation necessary to properly document, analyze, and understand the measured vibration characteristics.
In addition, it is highly desirable for the machinery diagnostician to fully
comprehend and understand the sometimes subtle relationships between the
fundamental dynamic signal characteristics of amplitude, frequency, and phase.
Although the preceding pages have attempted to describe several different situations, there are a myriad of potential combinations to be explored. Often, it is difficult to simulate the necessary interaction over a wide range of variables with
devices such as function generators or rotor kits. However, it is possible to
develop mathematical relationships to examine a wide range of variables. This
Time and Frequency Domain
333
exploration is highly recommended, and it may avoid costly mistakes due to the
misinterpretation of the machinery orbital data.
Historically, mathematical simulation of periodic signals has been a difficult task requiring careful equation structure, combined with extensive computer code for input, display, and plotting of the results. However, since the
evolution of desktop computers, this task has been significantly simplified. For
instance, the calculated orbits within this chapter were produced with a program
entitled Mathematica3. This program provides a highly flexible software package for symbolic mathematical plus graphical computations. Mathematica runs
on a variety of platforms with a common Kernel, and various front end interface
programs for each specific computer system. Since the program Kernel is the
same for all machines, this allows interchangeability of code and statements. It
is recommended that the diagnostician acquire this type of computational capability to enhance the ability for self-training. Other math programs provide similar functions, and the final program selection should be based upon the available
operating system and specific software requirements.
TIME AND FREQUENCY DOMAIN

Time domain analysis is representative of the actual sequence of events
occurring on process machinery. As discussed in the previous section, time and
orbital domain analysis depict the fundamental history of the machinery. However, there are situations when the complexities of the dynamic signals exceed
the data processing capabilities of an oscilloscope. In these cases, frequency
domain analysis is required for detailed dissection of the signals.
Over the past twenty-five years significant improvements have been made
in the instruments used for frequency analysis. The original swept or tuned filter
analyzers, and the initial real time analyzers were strictly analog devices that
primarily displayed amplitude as a function of frequency. These instruments
have been replaced by analyzers that use digital signal processing techniques,
and offer enhanced data acquisition and manipulation.
These current instruments are commonly referred to as Dynamic Signal
Analyzers (DSA), or Fast Fourier Transform (FFT) Analyzers. In either case, the
basic operation of the unit remains the same. DSAs usually incorporate additional functions such as signal sources for network gain and phase measurements, internal power supplies for driving ICP transducers, programing
capability, plus a variety of signal processing, manipulation, and display options.
Typically, these types of instruments operate across a frequency range that
extends from less than 1 Hz (60 CPM) to an upper end that exceeds 100 KHz
(6,000,000 CPM). Certainly, the vast majority of meaningful machinery vibration
response measurements fall well within this frequency domain.
The Fast Fourier Transform (FFT) refers to a fast and efficient method to
3 Stephen Wolfram, The Mathematica Book, 3rd edition, software version 3.0 (Champaign,
Illinois: Wolfram Media/Cambridge University Press, 1996).
334
Chapter-7
calculate the Discrete Finite Fourier Transform. This is the mathematical algorithm used for transforming amplitude versus time data into the amplitude versus frequency data. In practice, the input analog time domain signal is converted
into a digital equivalent. The appropriate windowing is applied, the user selected
number of samples are acquired, and the resultant frequency spectrum of the
time record is displayed.
Advanced instruments such as the Hewlett Packard HP-35670A offer a
variety of additional data processing and display options. However, the fundamental FFT concept converts the time domain dynamic signals into frequency
domain spectra. It is also significant to note that the sampled and digitized signals may be displayed and massaged in the time domain. Thus, a four channel
unit such as the HP-35670A may be used as a digital oscilloscope for time and
orbital domain analysis, as well as for frequency analysis.
The key to understanding frequency domain analysis is to recognize that
machines move and vibrate as a function of time, and frequency domain analysis
is another way of observing the machinery time record. In fact, barring any data
processing errors, the time and frequency domains may be considered as complementary and interchangeable ways of looking at the same data. Due to this close
relationship between the time and frequency domain, it is essential for the diagnostician to understand how common signals appear in both formats. The ability
to visualize how dynamic signals that are viewed in one format will appear in
the other format can save considerable time during data processing and analysis.
To assist in this appreciation of format translation, the following discussion of
four common signal types is presented for consideration.
The easiest dynamic signal to convert is a simple sine or cosine wave. As
discussed earlier in this text, this type of signal occurs at a single frequency, with
an easily definable amplitude. For instance, Fig. 7-30 is a time domain plot of two
complete cycles of a 100 Hz sine wave that has an amplitude of 1.0 Volt,o-p.
In many cases, dynamic signals may be electronically produced, and analyzed on an FFT. For example, a HP-33120A function generator was set to a sine
wave output with a frequency of 100 Hz (6,000 CPM). The voltage level was
Fig. 730 Calculated Time Domain Plot Of A

100 Hz Sine Wave
Fig. 731 Measured Frequency

Domain Plot Of A 100 Hz Sine Wave
335
adjusted to nominally 1.00 Volt,o-p, which is equal to 2.00 Volts,p-p. This AC signal was directed to an HP-35670A DSA, and the data processed in a 0 to 800 Hz
(48,000 CPM) frequency span. Resolution was set to 800 lines, and a flat top window was used to maximize the amplitude accuracy. The spectrum plot shown in
Fig. 7-31 depicts the results of this FFT analysis of the sine wave.
It is understandable that the analysis bandwidth is dominated by a single
component occurring at a frequency of 100 Hz (6,000 CPM), and displaying an
amplitude of 1.002 Volt,o-p. These characteristic values are identical to the input
signal frequency and amplitude. From a mathematical standpoint, the time
domain signal displayed in the spectrum may be evaluated for the instantaneous
voltage (Ysin) at any point in the cycle with the following expression:
Y sin = A sin ( t )
where
(7-2)
A = Maximum or Peak Voltage (Volts,o-p)

= Frequency (Cycles/Second or Hertz)
t = Time (Seconds)
This is the same general equation that was used for describing a simple
periodic motion in chapter 2. The calculated time base plot in Fig. 7-30 displays
the time history of this function for two complete cycles. Since the time required
for one cycle is 0.01 seconds (Fig. 7-30), that is equivalent to a frequency of 100
cycles per second. The spectrum plot (Fig. 7-31) of the same function reveals the
maximum voltage, plus the same frequency. Thus, the characteristic parameters
for this periodic motion may be calculated or measured, and plotted in two distinct but interrelated formats.
Although a sine wave is a simple example, it is common knowledge that
periodic functions may be expressed as a series of sines and cosines. This concept
was originally proposed by the French mathematician and physicist Baron Jean
Baptiste Joseph Fourier (1768-1830). His fundamental theories have formed the
foundation for wave analysis, and his name has carried through to the Fourier
series for basic periodic functions, the Fourier Transform, and the previously
mentioned Fast Fourier Transform (FFT).
An improved understanding of the time and frequency domain relationship
may be gained by an examination of the Fourier concepts. Certainly the previously discussed sine wave falls into this category, but it is a periodic motion that
represents the simplest case. For additional complexity, consider a Fourier series
of a triangular wave. This type of information may be obtained from reference
books such as Marks Handbook4, or the CRC Standard Math Tables5. Based on
these references, a convenient formula describing a Fourier series for a triangular wave may be expressed in the following manner:
4 Eugene A. Avallone and Theodore Baumeister III, Marks Standard Handbook for Mechanical Engineers, Tenth Edition, (New York: McGraw-Hill, 1996), pp.2-36.
5 Daniel Zwillinger and others, CRC Standard Mathematical Tables and Formulae, 30th edition (Boca Raton, Florida: CRC Press Inc., 1996), pp 49.
336
Chapter-7
8 A
cos ( n t )
Y tri = ----------------------------------------- 2
2
n
n = 1, 3, 5, 7.
(7-3)
In this equation, n is equal to the harmonic order. That is, when n=1, the
frequency component under consideration is the fundamental frequency. When
n=3, the third harmonic is the associated component, etc. If the maximum voltage A is set equal to 1.0 Volt,o-p, equation (7-3) may be expanded as follows:
8 1 cos ( 1 t ) cos ( 3 t ) cos ( 5 t ) cos ( 7 t )

Y tri = ----------- + ------------------------------ + ------------------------------ + ------------------------------ +
2 ----------------------------2
2
2
2
1
3
5
7
This expression for a triangular wave may be simplified slightly as:
8
cos ( 3t ) cos ( 5t ) cos ( 7t )
Y tri = -----2 cos ( t ) + ------------------------ + ------------------------ + ------------------------ +

9
25
49

The magnitude coefficients for each of the harmonics may now be computed. For instance, the first four components are calculated to be:
8
1st tri = -----2 ( 1 ) = 0.811 Volts, o-p

1
8
3rd tri = -----2 --- = 0.090 Volts, o-p
9
8
1
5th tri = -----2 ------ = 0.032 Volts, o-p
25
8
1
7th tri = -----2 ------ = 0.016 Volts, o-p
49
The triangular wave amplitude was initially specified as 1.00 Volt,o-p, and
each component must carry the same units. The general equation may now be
combined with the individual coefficients into the following cosine series:
Y tri = ( 0.811 cos t ) + ( 0.090 cos 3t ) + ( 0.032 cos 5t ) + ( 0.016 cos 7t ) +
The above series of cosine terms may now be plotted for two cycles as shown
in Fig. 7-32. Although the above series listed only the first four cosine terms, the
time domain plot was generated using the first eight terms. Note that the time
domain curve is in fact triangular shaped, with a peak amplitude of 1.00 Volt. In
this case, a frequency of 100 Hz, with a period of 0.01 seconds was used for the
fundamental. However, the same time domain plot would be generated if a nondimensional t term was used, and the curve plotted between 0 and 4 (2 cycles)
As before, an electronic signal was produced with a HP-33120A function
generator. This device was set to a triangular wave output of 1.00 Volt,o-p at a fre-
337
Fig. 732 Calculated Time Plot Of A 100

Hz Triangular Wave With 8 Components
Fig. 733 Measured Frequency Domain

Plot Of A 100 Hz Triangular Wave
quency of 100 Hz. This signal was directed to an HP-35670A, and the data processed in a 0 to 800 Hz bandwidth. Resolution was set to 800 lines, and a flat top
window was used to maximize amplitude accuracy. Fig. 7-33 documents the
results of this FFT analysis of the triangular wave signal. It is meaningful to
extract the amplitudes at each harmonic, and compare the calculated versus the
measured peak voltages in Table 7-2:
Table 72 Comparison Of Calculated Versus Measured Component Amplitudes For A
Triangular Wave With A Frequency Of 100 Hz, And An Amplitude of 1.00 Volt,o-p
Harmonic Order
Frequency
(Hertz)
Calculated
(Volts,o-p)
Measured
(Volts,o-p)
First
100
0.811
0.812
Third
300
0.090
0.090
Fifth
500
0.032
0.032
Seventh
700
0.016
0.016
Excellent agreement is noted between the calculated cosine terms, and the
measured FFT amplitudes. The consistency of this data is attributed to the outstanding signal stability of the HP-33120A function generator, plus the signal
processing accuracy of the HP-35670A Dynamic Signal Analyzer. Although the
higher order harmonics have small amplitudes, the extended dynamic range of
the HP-35670A allowed for an accurate measurement of very low voltage levels.
As another example, consider the Fourier series of a square wave. Accessing
the previously referenced Marks Handbook, or the CRC Standard Math Tables,
a Fourier series for a square wave may be expressed with the following common
expression:
338
Chapter-7
4 A
Y sqr = --------------

sin ( n t )
-----------------------------
n
n = 1, 3, 5, 7.
(7-4)
If the maximum voltage A is again set equal to 1.0 Volt,o-p this general
square wave equation may be expanded into the following:
4 1 sin ( 1 t ) sin ( 3 t ) sin ( 5 t ) sin ( 7 t )

Y sqr = ------------ ------------------------------ + ------------------------------ + ------------------------------ + ------------------------------ +

1
3
5
7
This expression may now be used to compute the magnitude coefficients for
each of the harmonics. For instance, the first four components (1st, 3rd, 5th, and
7th harmonics) are calculated as:
4
1st sqr = --- ( 1 ) = 1.273 Volts, o-p

4
1
3rd sqr = --- --- = 0.424 Volts, o-p
3
4
1
5th sqr = --- --- = 0.255 Volts, o-p
5
4
1
7th sqr = --- --- = 0.182 Volts, o-p
7
This general equation for a square wave may now be combined with the
individual magnitude coefficients, and the following sine series results:
Y sqr = ( 1.273 sin t ) + ( 0.424 sin 3 t ) + ( 0.255 sin 5 t ) + ( 0.182 sin 7 t ) +
The square wave series of sine terms will be plotted for two complete cycles
that is equivalent to an overall time span of 0.02 seconds for the 100 Hz fundamental frequency. For demonstration purposes, the series used for the plot will
consist of only the first four terms. The result of this approach is presented in
Fig. 7-34. This rendition is somewhat representative of a square wave, but the
tops and bottoms of each curve are ragged edges instead of smooth lines. Clearly,
additional higher frequency components are required to improve the simulation.
The diagram in Fig. 7-35 represents an extension of this same Fourier sine series
to twenty-four coefficients. Note that the time domain curve has been drastically
Fig. 734 Calculated

Time Domain Plot Of A
100 Hz Square Wave
Based On The First Four
Fourier Components
339

Hz Square Wave With 24 Components

Presentation Of A 100 Hz Square Wave
improved by the inclusion of the higher frequency components. Clearly, a complex dynamic signal with a structure of significant high frequency components
may be substantially distorted if the high frequency components are removed.
An electronic square wave was again produced on an HP-33120A function
generator. An output of 1.00 Volt,o-p at 100 Hz was directed to the HP-35670A,
and the resultant spectrum plot is displayed in Fig. 7-36. Extracting the amplitudes at each harmonic, and comparing the calculated versus the measured peak
voltages, the following Table 7-3 evolves:
Table 73 Comparison Of Calculated Versus Measured Component Amplitudes For A
Square Wave With A Frequency Of 100 Hz, And An Amplitude of 1.00 Volt,o-p
Harmonic Order
Frequency
(Hertz)
Calculated
(Volts,o-p)
Measured
(Volts,o-p)
First
100
1.273
1.276
Third
300
0.424
0.426
Fifth
500
0.255
0.255
Seventh
700
0.182
0.182
Agreement between calculated sine terms and measured FFT amplitudes

are excellent. Minor deviations are attributable to voltage drifts and imprecise
settings of the function generator. It should also be noted that the square wave
data reveals an amplitude at the fundamental frequency (1.273 Volts,o-p) that is
larger than the overall unfiltered signal (1.00 Volts,o-p). This relationship might
seem peculiar to those individuals who believe that: the unfiltered amplitude
must be larger than any of the components. This casual statement is simply not
true as demonstrated by the square wave example. In all cases, the component
amplitude relationship is a function of the time domain wave form, and the indi-
340
Chapter-7
vidual amplitude components and phase relationships required to reconstruct or

replicate the original wave form.
As noted, the triangular and square wave examples are both represented by
a series of odd numbered harmonics. It is also possible to produce a Fourier
series that consists of even and odd harmonics of the fundamental frequency. For
instance, a pulse wave will exhibit this type of series. From the CRC Math
Tables, a Fourier series for a pulse wave may be expressed as follows:
Y pul
2 A
= --------------

( 1 ) sin ( nK ) cos ( nt )
-----------------------------------------------------------------------------n
(7-5)
n=1
Within this expression, A represents the peak voltage, the n variable now
includes each harmonic (i.e., n=1, 2, 3, 4, 5...etc.), and K is used to represent the
duty cycle of the pulse. This non-dimensional duty cycle is determined by dividing the pulse width by the period. For example, a pulse with a 100 Hz frequency,
and an amplitude of 4.60 Volts,o-p is shown in Fig. 7-37. Although this signal
Fig. 737 Measured Time

Domain Waveform Of A
100 Hz Pulse Signal
originated with the pulse output from a HP-3311A function generator, it is representative of a Keyphasor from an optical pickup, or a proximity probe observing
a key. Fig. 7-37 shows a minor overshoot at the top and the bottom of the signal,
which is due to the ringing of the DSA digital filters. The signal period was measured as 9.997 milliseconds. Since frequency is the reciprocal of the period, it is
desirable to check this measurement as follows:
1
1 Cycle
Frequency = ------------------- = ------------------------------------------ = 100.03 Cycles/Second
Period
0.009997 Seconds
The period is verified to be consistent with the set frequency of 100 Hz, and
it is appropriate to proceed with the following calculation of the duty cycle:
Pulse Width
0.001583 Seconds
K = Duty Cycle = ------------------------------------- = ------------------------------------------ = 0.158
Period
0.009997 Seconds
Based upon this calculated duty cycle, and the maximum voltage A of 4.60
341
Volts,p-p, the general equation (7-5) for a pulse may be expanded as follows:
2 4.6 1 sin ( 0.158 ) cos ( t ) 1 sin ( 2 0.158 ) cos ( 2t )
Y pul = ---------------- ------------------------------------------------------------------------------- + -------------------------------------------------------------------------------------
1
2
1 sin ( 3 0.158 ) cos ( 3t ) 1 sin ( 4 0.158 ) cos ( 4t )

-------------------------------------------------------------------------------------- + -------------------------------------------------------------------------------------3
4
5
1 sin ( 5 0.158 ) cos ( 5t )

-------------------------------------------------------------------------------------- +
5
This complex expression may be simplified slightly, and the number of radians determined for each sine term as follows:
9.2
sin ( 0.9927 ) cos ( 2t )
Y pul = ------- sin ( 0.4964 ) cos ( t ) + ------------------------------------------------------------
2
sin ( 1.4891 ) cos ( 3t ) sin ( 1.9855 ) cos ( 4t ) sin ( 2.4819 ) cos ( 5t )

------------------------------------------------------------- + ------------------------------------------------------------- ------------------------------------------------------------- +
3
4
5
The magnitude coefficients for each of the first five components (harmonics)
may now be calculated in the following manner:
9.2
1st pul = ------- [ sin ( 0.4964 ) ] = 1.395 Volts, o-p

9.2
sin ( 0.9927 )
2nd pul = ------- ------------------------------ = 1.226 Volts, o-p

2
9.2
sin ( 1.4891 )
3rd pul = ------- ------------------------------ = 0.973 Volts, o-p

3
9.2
sin ( 1.9855 )
4th pul = ------- ------------------------------ = 0.670 Volts, o-p

4
9.2
sin ( 2.4819 )
5th pul = ------- ------------------------------ = 0.359 Volts, o-p

5
This general series should now be combined with the individual component
magnitude coefficients, and the following cosine series identified:
Y pul = { 1.395 cos ( t ) + 1.226 cos ( 2t )
0.973 cos ( 3t ) + 0.670 cos ( 4t ) 0.359 cos ( 5t ) + }
This series of cosine terms may now be plotted for two cycles (t=0 to 4) as
shown in Fig. 7-38. Due to the complexity of simulating a pulse curve, a total of
50 Fourier coefficients were required for this plot. The minor extraneous peaks
on this calculated time domain plot could be improved by extending the number
342
Chapter-7

Hz Pulse Signal With 50 Components

Presentation Of A 100 Hz Pulse Signal
of coefficients. However, this plot is considered to be a good simulation of the

actual measured time domain pulse reviewed in Fig. 7-37. Again, the machinery
diagnostician is cautioned against inappropriate filtration on a dynamic signal.
If this pulse signal represented a machine Keyphasor, excessive high-pass filtration would eliminate components that are required to maintain the integrity
of the pulse. This could easily result in a distortion of the pulse shape, and the
loss of phase continuity for the machinery train. Furthermore, if balancing work
was in progress, the alteration of the Keyphasor signal could result in a change
in trigger point, ergo a shift in vector angles. On a large diameter shaft this
might not be significant, but on a small diameter shaft, the resultant phase
change might prove to be devastating.
Finally, the FFT representation of the HP-3311A function generator pulse
output is presented in Fig. 7-39. Again, this is a 4.60 Volt,p-p signal at 100 Hz.
This spectrum plot depicts the series of even and odd harmonics of the fundamental. Extracting the amplitudes at each harmonic, and comparing the calculated versus the measured peak voltages, the summary Table 7-4 evolves:
Table 74 Comparison Of Calculated Versus Measured Component Amplitudes For A Pulse
Signal With A Frequency Of 100 Hz, And An Amplitude of 4.60 Volt,p-p
Harmonic Order
Frequency
(Hertz)
Calculated
(Volts,o-p)
Measured
(Volts,o-p)
First
100
1.395
1.390
Second
200
1.226
1.219
Third
300
0.973
0.962
Fourth
400
0.670
0.656
Fifth
500
0.359
0.343
343
Agreement between calculated terms, and measured FFT amplitudes is

quite acceptable but the comparative amplitudes are not as close as the previous three examples. Basically, this is attributed to the fact that the other signals
were generated from a precision source, whereas the pulse signal was not subjected to rigorous amplitude regulation. Considering the source of the pulse signal, and the minor inaccuracies associated with the determination of the duty
cycle, the results shown in Table 7-4 are certainly acceptable.
Other curves such as ramp or a sawtooth patterns will also generate a
string of consecutive harmonics. Shaft surface imperfections such as scratches or
rust will appear as multiple harmonics on proximity probe output signals. In
addition, combinations between various types of signals may also produce a long
string of 1X harmonics.
In retrospect, the past four examples of a sine wave, triangular wave,
square wave, and the pulse signal produce predictable and consistent results.
The interchangeability of time domain to frequency domain data has been
repeatability demonstrated. The diagnostician must recognize that FFT analysis
of dynamic signals may provide valuable information, and this type of data presentation is a fundamental analysis tool. However, the FFT originates from the
machinery vibration time record, and that original time domain signal must
always be examined to obtain the full measure of information.
Case History 16: Steam Turbine Exhaust End Bearing Dilemma
When problems occur on major machinery trains, various techniques may
be implemented to resolve the difficulty. One popular approach consists of assembling the personnel involved with the machine problem, and discussing the
potential origin, plus the corrective solution(s) for the malfunction. This is an
effective problem solving forum when the assembled personnel have a good
understanding of the equipment, and the abnormal behavior. In some cases the
opinions regarding the machinery problem may not coincide, and an acceptable
solution cannot be agreed upon. For instance, a dilemma of this type occurred on
a turbine compressor set following a routine maintenance inspection.
The physical machinery consisted of a horizontally split centrifugal compressor containing seven impellers, and a rotor weight of 750 pounds. The steam
turbine driver contained a three stage rotor that weighed 800 pounds, and produced nominally 5,000 HP at 9,000 RPM. The machines were connected with a
fully lubricated gear type coupling. Both machines contained 5 pad tilting pad
bearings, and diametrical clearances varied between 5 and 7 Mils. The compressor and the turbine had a good operating history, with only occasional problems
at the turbine exhaust end bearing.
During a scheduled plant shutdown, the maintenance personnel elected to
physically inspect the condition of the steam turbine exhaust end bearing for evidence of damage or deterioration. This work consisted of a vertical lift check,
combined with complete disassembly, and physical inspection of the bearing and
turbine journal. It was determined that the diametrical clearance for this turbine exhaust end bearing was 6 Mils. The bearing and journal were both in excel-
344
Chapter-7
lent condition, and the housing was reassembled using the same bearing. No
other maintenance work was performed on this machinery train, and operations
personnel felt quite confident restarting this equipment. The following startup
was performed without any problems, and the machine was successfully placed
on-line at a constant speed of 8,955 RPM. Overall vibration amplitudes at each
bearing were acceptable, and thrust positions for both rotors had returned to
their previous operating position. The only abnormality was a high temperature
of 210F at the turbine exhaust end bearing. This bearing typically operated
between 160 and 170F, and that temperature was measured by a thermocouple
embedded in the bottom bearing pad.
As usual, the startup was performed at night. During the warmth of the
next day, the bearing temperature increased to approximately 215F, and concern began to develop about the longevity of this bearing. At this point, various
examinations were performed, and the plant personnel used their new spectrum
analyzer to produce FFT plots of the proximity probes on this machine. A typical
set of spectrum plots across the coupling are shown in Fig. 7-40.
Fig. 740 Spectrum Plots

from Y-Axis Shaft Proximity Probes Across The
Gear Coupling Assembly
This data was acquired from the Y-Axis proximity probes installed at the
turbine exhaust bearing, and the coupling or discharge end of the compressor.
Note that both FFT diagrams reveal fairly low amplitudes at rotational speed,
plus a string of running speed harmonics (i.e., 2X, 3X, 4X, etc.). On the turbine
exhaust, the second harmonic of running speed has an amplitude that is twice as
large as the fundamental rotational speed motion. This data was of concern to
the plant personnel, and two different opinions were soon openly debated.
Some people believed that the series of multiple harmonics were due to
345
mechanical looseness of the turbine exhaust end bearing. Others had the opinion
that the measured vibratory behavior was primarily due to misalignment across
the gear coupling. The proposed solution was to closely monitor the machinery
vibration as the compressor hold down bolts were loosened, and the compressor
allowed to move into a state of satisfactory hot alignment.
For the uninitiated, it must be mentioned that this type of correction is
extraordinarily dangerous to the people as well as the machinery. If one considers the energy contained in a 800 pound rotor rotating at 8,955 RPM, and the
potential implications of releasing that energy by unloosening the hold down
bolts, the danger in this type of move is obvious.
Following several days of continuous operation, the vibration amplitudes
remained constant, and the turbine exhaust end bearing temperature continued
to cycle between 210 and 215F. By virtue of this consistent behavior, the previous two theories were discredited. Clearly, if the turbine exhaust bearing was
loose, the behavior would tend to degenerate with time, and that did not occur.
The second theory of misalignment was also disproved. Specifically, if heat generation in a bearing was truly due to misalignment it is logical to believe that
the bearing would either fail, or relieve some clearance due to the applied preloads. It is hard to believe that any significant misalignment would appear as
constant vibration and elevated temperature without any change. Furthermore,
the maintenance inspections performed during the previous shutdown did not
disturb the shaft alignment. Clearly, it was necessary to examine the machinery
behavior in more detail in order to resolve this problem.
Initially, the vibration signals shown in the frequency domain on Fig. 7-40
were examined in the time domain as displayed in Fig. 7-41. At both measure-

Plots from Y-Axis Shaft
Proximity Probes Mounted
Across The Gear Coupling
Assembly
ment planes, it is clear that the time base signals are corrupted by a series of
spikes that are indicative of shaft surface imperfections. There is a distinctively
different pattern between the turbine exhaust probe signal, and the compressor
discharge probe signal. However, both cases are representative of rough shaft
surfaces below the proximity probes. Comparison of this Y-Axis probe data with
the associated X-Axis probes at each measurement location (not shown) reinforces the fact that the vertical spikes are shaft surface scratches that are
observed by both probes at each bearing. Based on the previous discussion of the
Fourier components of a pulse wave, it is logical to conclude that the majority of
346
Chapter-7
the harmonic activity shown in Fig. 7-40 is simply due to the shaft scratches.
Some people will accept this explanation for the running speed harmonics
on the shaft vibration signals, and other individuals will not. For additional
proof, it is reasonable to obtain a set of casing velocity measurements across the
coupling as displayed in Fig. 7-42. This data clearly shows that higher order harmonics do not exist on the bearing housings. This is particularly meaningful on
the turbine exhaust bearing. As with many steam turbines of this general size,
the exhaust end bearing is a fairly simple unit. The outer shell of the bearing
housing retains the tilt pad bearing assembly. There is normally a close relationship between the frequency components measured on the shaft versus the casing. More specifically, if the turbine shaft was really subjected to a strong twice
rotational speed component, this frequency component would also appear on the
turbine bearing cap vibration data.

From Vertical Casing
Velocity Coil Measurements Across The Gear
Coupling Assembly
It is very reasonable to conclude that the predominant motion on this

machinery train occurs at the rotational speed, and the proximity probe signals
are influenced by shaft surface imperfections. The time domain scratches that
are clearly visible in Fig. 7-41 are converted into the Fourier components initially
presented in Fig. 7-40. There is absolutely no reason to believe that misalignment or mechanical looseness of the turbine exhaust bearing are contributors to
the high bearing temperature. However, it is apparent that the bearing pad temperature is higher than normal, and external pyrometer measurements of the
turbine exhaust bearing housing also reveals hotter than normal temperatures.
The issue then resolves back to the original question of: How come the exhaust
end turbine bearing runs hotter than normal?
Signal Summation
347
To properly answer this fundamental question, it is necessary to leave the

comfort of the air conditioning, and actually examine the operating equipment.
This approach goes back to the old days of go out and look, touch, feel, smell, and
listen to the machinery. In many instances you do not know what you are specifically looking for, and the best advice is to thoroughly examine the machinery for
any peculiarities. On this train it was observed that the oil flow leaving the turbine exhaust bearing was minimal. There is an old adage that states: for every
ten drops of oil, only one drop is for lubrication, and the remaining nine drops are
for cooling. On this particular bearing, there was barely enough oil flow for lubrication, and not much left over for cooling.
It was concluded that oil flow to the turbine exhaust bearing was restricted,
and this resulted in the elevated bearing temperature. To test this hypothesis,
the lube oil supply pressure was increased from 20 to 25 Psig. This change was
carefully monitored to insure that there were no detrimental effects to the other
machine train bearings. As the oil supply pressure was gradually increased, the
bearing temperature dropped. At an oil pressure of 25 Psig, the 215F bearing
temperature was reduced to 203F. In addition, the oil flowing through the drain
sight glass did perceivably increase. The machinery train was successfully operated in this manner for the next six months. At that time, the bearing was
opened during a short plant outage. It was discovered that Permatex was blocking the oil inlet to the turbine exhaust bearing. After this blockage was removed,
and the bearing properly assembled, temperatures returned back to normal.
SIGNAL SUMMATION
The previous discussions have generally assumed that the vibratory motion
is associated with some fundamental frequency. Typically, this frequency would
be the running speed of the machine, and the various harmonic components
would track this fundamental frequency. Specifically, if the rotational speed
increased, the frequency of the harmonic components would increase proportionally. This type of relationship would apply on supersynchronous excitations such
as 2X, 5X, 10X, as well as subsynchronous frequencies that are locked into fractions of the fundamental rotative speed such as X/3, or X/2.
Within the industrial environment, multiple excitations are emitted by different machines, and it is normal to encounter various interactions between frequencies. Excitations are transmitted through fluid streams, through piping
systems, and they are also conveyed through foundations and other support
structures. This multiplicity of forced and natural resonant excitations are often
combined into complex and interesting mechanical movements. Vibration measurements of these machine elements reveal the myriad of excitations, and it
then becomes necessary to sort out the significant from the inconsequential nonvibratory components. In this regard, the diagnostician must be able to distinguish the common types of signal interactions, and be able to relate these patterns back to the behavior of machine elements.
The three most common types of characteristics encountered are signal
348
Chapter-7
summation, amplitude modulation, and frequency modulation. The two types of

signal modulation will be discussed in the following sections of this chapter. At
this point in the text, the discussion will center on the commonly observed
behavior that will be referred to as signal summation within this book.
Signal summation can, and does, occur between two or more frequencies
that may be closely spaced, or they may be quite divergent in their fundamental
frequencies. In general terms, the combination or summation of a series of
dynamic signals may be expressed as follows:
V sum = F ( t ) 1 + F ( t ) 2 + F ( t ) 3 +
(7-6)
Each separate excitation source is identified as a time variable function

such as F(t)1, F(t)2, F(t)3, etc. The time summation of these different excitations
is given by Vsum. As a specific example, the following equation (7-7) may be used
to identify the parameters associated with summing two different frequencies:
V sum = { V 1 sin ( 2F 1 t ) } + { V 2 sin ( 2F 2 t ) }
where:
Vsum
V1
F1
V2
F2
t
=
=
=
=
=
=
(7-7)
Instantaneous Summation Voltage (Volts)

Peak Voltage of Sine Wave #1 (Volts)
Frequency of Sine Wave #1 (Hertz)
Peak Voltage of Sine Wave #2 (Volts)
Frequency of Sine Wave #2 (Hertz)
Time (Seconds)
Within this equation it is assumed that two periodic excitations are combined, and each excitation is defined as a simple sine function. These could also
be other trigonometric functions, exponential functions, ramps, or a constant
amplitude across the frequency domain. However, for convenience during this
explanation, a sine function was selected for both signals. The first component
was assigned a frequency of F1, and a peak amplitude of V1. Similarly, the second
part of the signal is defined by a frequency of F2, and a voltage amplitude of V2.
Since both signals vary with time, the time t is included in equation (7-7).
For demonstration purposes, assume that equation (7-7) represents the
vibration characteristics of two adjacent machinery trains. Presume that one
machine was running at 100 Hz (6,000 RPM), and producing a vibration signal
with a voltage magnitude of 1.0 Volt,o-p. Further assume that the second
machine was operating at 105 Hz (6,300 RPM), and emitting a vibration signal
with a peak voltage of 1.5 Volts,o-p. If these values are used to define the sine
functions in equation (7-7), and if a time span of 0 to 0.4 seconds was examined,
the calculated plot shown in Fig. 7-43 is easily developed to describe the summation of these two sine waves occurring at different frequencies.
To be perfectly clear, this calculated plot in Fig. 7-43 reflects the interaction
of the two signals based upon equation (7-7). The signals obviously add together
to form the hump in the plot, and they cancel each other to provide the low
amplitude portions. Physically, this would be sensed or interpreted as alternating periods of high and low vibration. The frequency differential between the two
Signal Summation
349
Fig. 743 Calculated

Time Domain Plot Of Signal Summation Between
Two Independent Sine
Waves Of Different Frequencies And Amplitudes
conditions would be a simple beat frequency identified as Fbeat. This frequency,

and the associated period may be stated by the following:
F beat = F 2 F 1
(7-8)
1
Period beat = -------------F beat
(7-9)
In this example, the differential beat frequency is 5 Hz (300 CPM), and the
associated period is 0.2 seconds from equation (7-9). The period of this beat is
consistent with low frequency period of 0.2 seconds displayed in Fig. 7-43.
Signal summation is common in virtually all types of machinery systems. It
most instances, the summation encompasses multiple frequencies into the final
vibration signal. This is normal behavior, and it is one of the common occurrences that the diagnostician must address. One note of caution that should be
mentioned, is that this behavior is not an amplitude modulation. Although Fig.
7-43 looks like an amplitude modulated signal, this is not amplitude modulation.
Some characteristics of summed versus AM signals are similar, but the mechanical implications are quite different. Following the next case history on ID fans, a
discussion of amplitude modulated signals will be presented.
Case History 17: Opposed Induced Draft Fans
A classic example of signal summation was displayed by the pair of induced
draft fans depicted in Fig. 7-44. These two machinery trains are mounted on top
of a primary reformer furnace at an elevation of 240 feet above grade. Each
steam turbine driver rotates in a counterclockwise direction, and the pair of
speed reducing gear boxes provide clockwise rotation to each fan. These rotational observations are based upon standing at the governor end of either turbine. However, when both machinery trains are viewed from the outboard end of
either fan it is clear that one fan rotates clockwise, and the other fan turns in
a counterclockwise direction.
The fans are mounted on a common set of support I-beams that run from
east to west across the top of the furnace. Each turbine is under speed control to
350
Chapter-7
CCW
Gear
Box
East
I.D. Fan
CW
Steam
Turbine
CW
Gear
Box
Steam
Turbine
West
I.D. Fan
CCW
Fig. 744 Induced Draft Fans Mounted On Top Of A Primary Reformer Furnace
maintain a specific draft in each respective section of the furnace. Inlet dampers
are installed on each fan, but the most of the draft control resides with changes
in fan speed (via turbine speed control). The gear box ratio for both trains is constant with 95 bull gear, and 31 pinion teeth. Fan operating speeds typically vary
between 1,000 and 1,400 RPM. These fan speeds translate to a turbine operating
range of 3,065 to 4,290 RPM.
Historically, both ID fans have experienced multiple problems, and a variety of structural braces and supports have been installed over the years. In some
locations, such as the outboard fan pedestals, the braces were supported by additional braces. Both fan rotors are fairly long, and susceptible to bowing. Hence,
operations personnel have implemented strict procedures for slow roll and cooling during a normal reformer shutdown sequence.
Unfortunately, a plant emergency forced a trip of the furnace and both fans
late one evening. Due to the preoccupation of all available personnel with the crisis situation, the fans were left unattended for several hours. During this time,
both rotors experienced substantial shaft bows. In addition, the west fan impeller was damaged, probably during the emergency trip. During the ensuing outage, the furnace tube damage was repaired, but minimal attention was paid to
the rotating equipment. However, the extent of the fan shaft bows was quite evident during the next startup. High vibration levels were exhibited by both fans,
and the entire superstructure of the furnace seemed to be in sympathetic vibration with the fans.
Operations recognized that full speed and load operation was unattainable,
and both east and west units were slowed down to maintain tolerable vibration
amplitudes. A vibration analysis of the fans plus the upper structure of the furnace revealed three fundamental frequencies. The two fan running speeds dominated the machinery trains, and the beat frequency between the fan running
speeds appeared throughout the furnace superstructure. In fact, the low frequency beat was so strong, it was clearly audible from the front gate.
During this investigation, structural plus bearing housing vibration data
was collected on both the east and west machinery trains. For example, the vertical and horizontal time domain motion at the inboard, coupling end bearing of
the east fan are presented in Fig. 7-45. In this time domain plot, the maximum
vertical amplitude was 6.1 Mils,p-p, and the horizontal motion was slightly
Signal Summation
351
Fig. 745 East Fan - Coupling End Bearing Housing Time Domain Vibration Data
higher at 6.7 Mils,p-p. The beat frequency was clearly visible, and it should be
mentioned that the other fan bearing housings displayed similar characteristics.
A frequency analysis of this data from the east fan is presented in Fig. 7-46.
The vibration measurements were made with casing velocity pickups, and
the data was integrated from velocity to casing displacement for both Figs. 7-45
and 7-46. Note that the FFT data also includes the axial casing motion of this
east fan coupling end bearing housing. Although the axial vibration appeared at
lower amplitudes, it displayed the same frequency distribution as the two lateral
transducers.
Fig. 746 East Fan - Coupling End Bearing Housing

Vibration Spectra
There was no apparent vibration at other frequencies, and the spectrum

plot in Fig. 7-46 was expanded to provide improved visibility of the dominant
running speed components. In this case, a frequency window of 900 to 1,500
CPM was examined. As noted in Fig. 7-46, the east fan speed was 1,071 RPM,
and the impeller damaged west fan was operating at a more conservative speed
of 1,013 RPM. The beat frequency is computed from equation (7-8) as:
352
Chapter-7
F beat = F 2 F 1 = 1, 071 1, 013 = 58 CPM

This frequency may now be used to calculate the beat period as follows:
Sec.
1
1 Cycle
Period beat = -------------- = --------------------------------------- = 0.0172 Minutes 60 ----------- = 1.03 Seconds
Min.
58 Cycles/Minute
F beat
The period of 1.03 seconds coincides with the low frequency envelope of the
time domain plot in Fig. 7-45. There was no question that either a signal summation or an amplitude modulation was occurring between the fan running speeds
on top of the furnace. Although this FFT information was definitive in terms of
the respective amplitudes and frequencies, it contained minimal information
regarding the relative motion between frequencies.
Additional perspective of the fan behavior was provided by examination of
this data in the orbital and time domains. For example, Fig. 7-47 displays the
orbit time base plots of the east fan inboard bearing housing. This information
was filtered precisely at the rotative speed of the east fan. This was accomplished
by installing a temporary optical pickup on the east fan shaft, and filtering the
data at the Keyphasor (i.e., running speed) frequency with a DVF. The information shown in Fig. 7-48 was extracted from the same database, but this time the
signals were filtered with a temporary optical pickup installed on the west fan.
From the orbital data in Fig. 7-47, the east fan 1X motion was clockwise,
and with rotation. This is normal and expected behavior for this type of machine
mounted on a compliant support structure. However, the east fan data filtered at
the west fan speed reveals a counterclockwise orbit precession in Fig. 7-48. This
rotational direction is consistent with the west fan rotational direction when
viewed from east to west. Hence, the east fan was excited by running speed
vibration from both east and west fans. Similar measurements on the west fan
indicated that both fan speeds were likewise driving the west fan structure.
Fig. 747 East Fan Coupling End Bearing

Housing Vibration At East Fan Speed
Fig. 748 East Fan Coupling End Bearing

Housing Vibration At West Fan Frequency
Additional tests were performed by holding one fan speed constant, and
varying the other fan speed. In all cases, the beat remained fully active, and the
frequency was always equal to the differential between fan operating speeds. It
was noted that the vibration severity was largest when the beat frequency was in
the vicinity of 60 Cycles/Minute (1 Hz). It was speculated that this frequency
Amplitude Modulation
353
might be in the vicinity of a furnace structural resonance.

Based upon this analysis, the east fan rotor was adequately straightened,
and this rotor was subjected to a field trim balance. The west fan rotor was damaged beyond a reasonable level of repair, and it was replaced. The new west fan
rotor was also field trim balanced. At the conclusion of the field balancing work,
casing and shaft vibration amplitudes were all below 1.0 Mil,p-p, and the interaction between fans was virtually eliminated. The nominal 1 Hz beat frequency
was also eradicated, and the furnace structure resulted in better than normal
vibration characteristics. Finally, the multiple braces and supports were
removed from each of the outboard fan pedestals.
AMPLITUDE MODULATION
Another form of commonly observed signal patterns is known as amplitude
modulation, and this is abbreviated as AM. The physics of amplitude modulation
varies somewhat from radio broadcasting to machinery interactions. For
instance, an AM radio signal is generated by the encoding of a carrier wave by
variation of its amplitude in accordance with an input signal. Radio signals usually display high frequency carrier signals, combined with lower frequency modulating signals. Similarly, AM signals on process machines are signals that
experience a change in amplitude of one signal due to the amplitude of the second or modulating signal. The machinery AM signals may have large differences
between the carrier and the modulating frequencies, or these frequencies may be
fairly close together.
Due to the multiple applications of AM, there are different definitions of
how two or more signals interact to produce an amplitude modulated signal. For
instance, the following equation (7-10) may be used to mathematically describe a
common representation of amplitude modulation:
V am = V o { sin ( 2F m t ) } { sin ( 2F c t ) }
where:
Vam
Vo
Fc
Fm
t
=
=
=
=
=
(7-10)
Instantaneous Amplitude Modulated Voltage (Volts)

Peak Voltage of Wave (Volts)
Frequency of Carrier Wave (Hertz)
Frequency of Modulating Wave (Hertz)
Time (Seconds)
In many respects, equation (7-10) for amplitude modulation is similar to

the previously discussed equation (7-6) for signal summation. Whereas two independent frequencies were specified for the summation case, the AM expression
identifies a carrier frequency Fc plus a modulating frequency Fm. In (7-6) the two
sine terms were summed in the time domain. For AM, the sine of the carrier and
sine of the modulating frequencies are multiplied together as in equation (7-10).
The summation of two components in equation (7-6) resulted in a peak
amplitude that was equal to the sum of both independent signals. With an AM
signal, the base amplitude is generally considered to be some constant value
354
Chapter-7
(e.g., Vo) that is attenuated to various levels throughout the periodic cycle. In certain cases, the similarities between signal summation and amplitude modulation results in time domain signals that are difficult to distinguish.
For example, equation (7-10) was used to compute the AM signals shown in
Figs. 7-49 and 7-50. Within Fig. 7-49, a 5,000 Hz carrier frequency Fc was modulated by a 200 Hz frequency Fm. Note the physical similarities between the true
AM signal in Fig. 7-49, and the pair of summed signals previously discussed in
Fig. 7-43. Based on this visual similarity, it is no wonder that many cases of simple signal summation are often referred to as amplitude modulation.
Fig. 749 Calculated AM Signal With

5,000 Hz Carrier And 200 Hz Modulator
Fig. 750 Calculated AM Signal With 200

Hz Carrier And 190 Hz Modulator
The time domain plot shown in Fig. 7-50 is another common form of an AM
signal. This plot was computed with a carrier frequency of 200 Hz, and a modulating frequency of 190 Hz. Although the time domain pattern has been altered,
this is still a pure AM signal. Note that the low frequency period for this signal is
nominally 100 milliseconds, or 0.1 seconds. This period is equivalent to the beat
frequency of 10 Hz, or the differential between the carrier and the modulating
frequencies (i.e., 200-190=10 Hz).
For demonstration purposes, an amplitude modulated signal equivalent to
the calculated data in Fig. 7-49 was generated with an HP-33120A function generator. The carrier frequency was set at 5,000 Hz (300,000 CPM), and the modulating frequency was adjusted to 200 Hz (12,000 CPM). The synthesized data
was directed to HP-35670A, and the processed time base and FFT data are
shown in Figs. 7-51 and 7-52. The similarities between the calculated time plot
in Fig. 7-49 and the measured data in Fig. 7-51 are self-evident. Within the frequency domain, the wide band FFT revealed a low frequency component at 200
Hz (not shown), and a high frequency component at 5,000 Hz.
For improved visibility, the FFT was zoomed in (translated spectrum) to a
frequency range of 4,200 to 5,800 Hz. The data from this 1,600 Hz span is presented in Fig. 7-52. Note that the carrier frequency of 5,000 Hz is bracketed by
two strong components at 4,800 and 5,200 Hz. For a true amplitude modulated
signal, the side bands at 4,800 and 5,200 Hz represent the sum and difference
frequencies between the carrier and the modulating frequency. Stated in another
Fig. 751 Measured AM Signal With 5,000

355
Fig. 752 Measured FFT of AM Signal

With 5,000 Hz Carrier & 200 Hz Modulator
way, for a true AM signal, the visible frequency components should include the
carrier and modulating frequencies, plus sidebands equal to the sum and difference of the two fundamental frequencies.
As another example, the same procedure may be applied to the closely
spaced carrier and modulating signals previously computed for Fig. 7-50. If this
200 Hz carrier plus the 190 Hz modulating signal are synthesized and sent to
the DSA, the plots shown in Figs. 7-53 and 7-54 may be generated. Once again,
the time domain similarity between the calculated signal of Fig. 7-50 and the
measured data shown in Fig. 7-53 is quite clear. The spectrum plot in Fig. 7-54 is
quite interesting since it encompasses all of the AM frequencies. Specifically, this
includes the carrier frequency of 200 Hz, the carrier plus the modulating frequency at 390 Hz, the differential beat frequency at 10 Hz, and a small component at 190 Hz that is the modulating signal.
In most cases, signal summation consists of two independent frequencies
that originate from totally different sources. These two frequencies interact in
Fig. 753 Measured AM Signal With 200

Fig. 754 Measured FFT of AM Signal

With 200 Hz Carrier And 190 Hz Modulator
356
Chapter-7
such a manner that at times they cancel each other out, and at other times they
add together to produce high vibration amplitudes. On these summed signals,
the beat frequency is often visible or even audible, and a frequency analysis
would exhibit two individual components separated by the beat frequency. However, in an amplitude modulated signal there is often a direct physical link
between the carrier and the modulating signal. That is, AM signals are usually
associated with interactive excitations on one shaft, or rotating elements that
are in direct physical contact (e.g., mating gear sets). The distinguishing or defining frequency characteristics of a true amplitude modulated (AM) signal are the
presence of the following frequency components:
Carrier frequency
Modulating frequency
Sum of the carrier and modulating frequencies
Difference between carrier and modulating frequencies
The mutual coexistence of these four frequency components is illustrated by

the following case history of a centrifugal compressor that displays all of the
characteristics of a true amplitude modulation.
Case History 18: Loose and Unbalanced Compressor Wheel
A five stage high pressure centrifugal compressor operated successfully for
over 20 years at speeds between 9,000 and 9,600 RPM. The compressor was in
refrigeration service, and it experienced minimal mechanical problems throughout its entire operating history. During a scheduled overhaul, the unit was disassembled for mechanical inspection, and the spare rotor installed. The rotor that
had been running during the previous 4 years was sent to the shop for a routine
set of assembled runout, outer diameter, plus axial position measurements of the
wheels with respect to the thrust collar. At the conclusion of these rotor measurements, an over zealous shop supervisor had the rotor un-stacked. At this time
they began runout measurements of the bare shaft, keyway inspections, plus the
acquisition of bore dimensions on the impellers, thrust disk, and balance drum.
The rotor inspection work was progressing in a timely manner, when the
overhaul was completed, and the compressor was restarted with the spare rotor.
At a speed of about 3,000 RPM the compressor ingested some type of foreign
object that literally destroyed the first stage impeller. The train was not
equipped for automatic shutdown, and the compressor operated for several minutes in a seriously distressed condition. Subsequent inspection revealed that the
suction end journal bearing was damaged, and the shaft was severely scored.
At this point, the only available option was to split the case, pull the spare
rotor, and reinstall the other rotor with the best available set of bearings and
seals. Unfortunately, the other rotor was scattered all over the shop, and the
responsible shop supervisor was completely embarrassed. In virtually all operating plants, the rotor(s) removed from the machines during a turnaround are not
disassembled until the plant has been up and running for a week or more. In this
357
case, the shop supervisor jumped the gun, and put the plant startup in jeopardy
when he ordered the un-stacking of the compressor rotor.
Options were limited, and the only thing to do was reassemble the old rotor,
do a quick check in the shop balancing machine, and stuff this rotor back into the
compressor. There was a lot of pressure on the shop personnel, and they ignored
the fundamental rotor assembly rule of: do it right, or do it over. They rushed the
job, and although the residual unbalance of the stacked rotor was different from
the incoming inspection report, they went ahead and reinstalled the old rotor
back into the compressor casing.
During the ensuing startup, the compressor exhibited fluctuating vibration
amplitudes from slow roll up to minimum governor speed. As the unit continued
to operate, and temperatures began to stabilize, the fluctuations in radial shaft
vibration tended to increase, and then stabilize into clean smooth shaft orbits. As
time progressed, the periods of high vibration increased in severity, and the low
vibration condition was also degenerating. An example of the shaft vibration just
prior to trip of the machine is displayed in Fig. 7-55.
Fig. 755 Shaft Displacement Orbit And Time Base

Plots With Loose And
Unbalanced Middle Compressor Wheel
Note the classical amplitude modulation characteristic of the time domain

plot, plus the radial pulsation of the shaft orbit. In essence, the shaft orbit varied
from a minimum amplitude of 1.0 to a maximum of 5.4 Mils,p-p. The multiple
Keyphasor dots shown on the orbit indicated the presence of a forward and circular subsynchronous vibration component. Examination of the synchronous, 1X
shaft vibration revealed a forward circular precession, with amplitudes that varied between 1.0 and 2.5 Mils,p-p.
Both inboard and outboard compressor bearings revealed similar characteristics. The suction, thrust end, of the compressor was slightly lower than the discharge end bearing. This high vibration condition appeared and subsided at both
ends of the compressor in unison. Even the experienced hands who had worked
on this machine for many years agreed that something was really wrong when
the floor grating began to pop out of the retaining channels. Vibration data was
recorded, and the unit was finally shutdown to minimize any further damage to
the only available compressor rotor assembly.
Following a safe shutdown of the machinery train, the recorded vibration
data was examined in greater detail. For instance, the spectrum plot in Fig. 7-56
358
Chapter-7
Fig. 756 Spectrum Plot

of Vertical Shaft Displacement With Loose And
Unbalanced Middle Compressor Wheel
documents the frequency behavior sensed by the vertical proximity probe at a

time equivalent to the orbit and time base data shown Fig. 7-55. The two dominant frequency components are the 1X running speed vibration at 9,420 RPM
and 2.4 Mils,p-p, plus a larger component at 8,940 CPM and 3.1 Mils,p-p. The differential between these two major components provides a lower sideband at 480
CPM (=9,420-8,940). Similarly, summation of the two major components yields
an upper sideband at 18,360 CPM (=9,420+8,940). By the definitions listed in the
previous section, it is reasonable to conclude that this observed behavior is a
clearly defined amplitude modulation.
Since it is an AM interaction between running speed and the larger component at approximately 95% of running speed, it is reasonable to concentrate on
the compressor rotor as the culprit. If this was a signal summation condition, the
search for potential problems would have to be expanded to a much broader base.
However, since this is an amplitude modulation, the efforts were fully directed
towards the rotating assembly.
Further examination of the full speed data revealed that the 95% component periodically increased in frequency, and locked into the running speed
motion. Under this state, shaft vibration amplitudes subsided, and the compressor appeared fairly normal. Then, for no apparent reason, the subsynchronous
component would reappear, and drift back to 5% below running speed with
increasing vibration amplitudes. Again, both ends of the compressor exhibited
the same type of behavior, with very similar vibration levels. Since this was a
reasonably symmetric rotor, this suggested that the malfunction was associated
with a problem in the middle of the compressor rotor.
A variation in compressor load would have been beneficial in analyzing this
problem, but that option was not achievable. In the final analysis, it was concluded that a midspan compressor impeller was loose. It was speculated that
during a free spinning condition, the loose impeller dropped to 95% of operating
speed. As it ran under this condition, the residual mass unbalance plus eccentricity of this wheel provided a forward circular forcing function to the rotor.
Since this occurred at a different frequency from the shaft running speed, it
resulted in two distinct excitations. One excitation occurred at shaft rotational
Frequency Modulation
359
speed, and the other excitation was due to the rotating frequency of the loose
wheel. Furthermore, as the loose impeller continued to rotate, friction between
the impeller bore and the shaft outer diameter would result in localized heating.
This could cause the impeller bore to shrink back onto the shaft, and result in a
temporary elimination of the 95% loose wheel frequency. As the refrigerant flow
cooled off the shaft and impeller, the wheel would eventually re-initiate the
entire cycle by loosening up, and slowing down to 95% of operating speed.
After shutdown, and disassembly of the compressor, plus un-stacking of the
rotor it was determined that the middle compressor impeller was indeed loose.
The primary reason for this loose wheel was that the impeller key was never
reinstalled during the hasty shop rebuild of this rotor. There was indication of
this problem on the slow speed shop balance machine, but the responsible individuals choose to ignore this information. Once again, the machinery responds in
accordance to the laws of physics, and the associated human beings tend to manage by emotion instead of good engineering practice.
FREQUENCY MODULATION
The final type of common signal interaction commonly encountered around
machinery analysis is known as frequency modulation (FM). This type of signal
was developed between 1925 and 1933 by Edwin Howard Armstrong (1890 to
1954). The original intent of his invention was to eliminate static on radio transmissions. Today, FM signals are used in a variety of different applications. In all
cases, these signals originate with a high frequency, constant amplitude carrier
signal. The frequency of this carrier wave is varied (or modulated) by the lower
frequency modulating wave or signal. A simple form of frequency modulation is
presented in the following expression:
V fm = V o cos { ( 2F c t ) + sin ( 2F m t ) }
where:
Vfm
Vo
Fc
Fm
t
=
=
=
=
=
(7-11)
Instantaneous Frequency Modulated Voltage (Volts)

Peak Voltage of Carrier Wave (Volts)
Frequency of Carrier Wave (Hertz)
Frequency of Modulating Wave (Hertz)
Time (Seconds)
Once again, Fc is defined as the carrier frequency, Fm is the modulating signal, Vo is the carrier voltage, and Vfm represents the instantaneous FM voltage
over time t. Note that the time dependent sine of the modulating frequency is
combined with the time dependent carrier frequency in equation (7-11). The
resultant combination is subjected to a cosine trigonometric function. Overall,
this type of FM expression appears as a normal cosine function with a constant
peak-to-peak amplitude, and a distortion of the frequency with the passage of
time. The diagram presented in Fig. 7-57 was computed with a carrier frequency
Fc of 5,000 Hz, a modulating frequency Fm of 4,000 Hz, and a base voltage of 1.0
360
Chapter-7
Fig. 757 Computed Time

Domain Plot Of FM Signal
Showing Accordion Effect
Volt,o-p. If this type of signal is viewed live on an oscilloscope, the time base wave
will appear to move back and forth horizontally across the screen. This type of
time domain motion appears to oscillate like the bellows of an accordion. Hence,
time base observation of a FM signal is sometimes referred to as an accordion
effect. In many instances this is a subtle condition that may be easily missed if
an appropriate sweep rate is not selected on the oscilloscope.
A broad band frequency analysis might also miss an FM signal if the modulating frequency is small compared to the carrier frequency. For example, if a
5,000 Hz carrier is frequency modulated by a 100 Hz component, a wide 5,000 Hz
component would be visible in a 0 to 12,800 Hz spectrum. If the DSA frequency
span was translated to a center frequency of 5,000 Hz, and a bandwidth of 800
Hz, the spectrum shown in Fig. 7-58 would appear. In this example, the carrier
frequency appears at the center of the array at 5,000 Hz. In addition, a series of
side bands appear at the carrier frequency, plus and minus multiples of the modulating signal (i.e., FcFm, Fc2Fm, Fc3Fm, etc.).
This same behavior occurs in Fig. 7-59, where the modulating frequency
has been reduced to 50 Hz. Note that the quantity and size of the sidebands have
been altered. In addition, the magnitude of the carrier frequency at 5,000 Hz has
been reduced. In some cases, the carrier frequency might be reduced to very low
Fig. 758 Measured FFT Of FM Signal

With 5,000 Hz Modulated By 100 Hz
Fig. 759 Measured FFT Of FM Signal

With 5,000 Hz Carrier Modulated By 50 Hz
361
levels. Overall, the distinguishing characteristics of a frequency modulated (FM)

signal are summarized as follows:
Carrier frequency at the center of the component array

Multiple sidebands at the carrier frequency minus orders of the modulator
Multiple sidebands at the carrier frequency plus orders of the modulator
Constant signal amplitude when viewed in the time domain
No distinct component occurring specifically at the modulating frequency
In the field of machinery analysis, FM signals are encountered in three distinctly different ways. This includes FM tape recorders, measurement of torsional vibration using FM techniques, and mechanical phenomena that generate
FM signals. In the first application of FM magnetic tape recorders, a carrier frequency is generated, and the dynamic signals are superimposed upon the carrier.
During data reproduction, the high frequency FM signals are demodulated, and
the resultant signals provide excellent reproductions of the original dynamic
data. FM tape recorders provide high signal to noise ratios, combined with the
ability to record data down to zero frequency (DC) of the dynamic signal.
The characteristics of FM recorders are specified by ISO 3615. For example,
consider a tape transport speed of 15 inches per second (38.1 centimeters/second). A Wide Band Group 1 recorder will have a carrier center frequency 54.0
kHz, and a recording bandwidth extending from DC to 10.0 kHz. This varies
with speed, so that a decrease to a tape speed of 7.5 inches per second (19.05 centimeters/second) will cut the center frequency in half to 27.0 kHz. It will also
reduce the recording bandwidth by a factor of two to DC to 5.0 kHz. This topic
will be discussed in greater detail in the following chapter 8.
The second common utilization for FM signals consists of torsional vibration measurements. In this application, a transducer (e.g., a proximity probe or
magnetic pickup) is positioned over a gear or a precision cut slotted wheel. As the
machine rotates, the gear tooth passing, or slot passing frequency, acts as a carrier signal. Any torsional vibration of the machinery will result in angular oscillations that frequency modulate the carrier signal. Demodulation of the FM
signal provides an electronic signal that is proportional to torsional vibration.
This application was previously discussed in chapter 6 of this text.
It should be mentioned that the carrier frequency for torsional measurements generally occurs at a fairly high frequency. For instance, if a 3,600 RPM
turbine generator set is equipped with a common 60 tooth wheel, the resultant
carrier frequency at synchronous speed will be 216,000 CPM (3,600 Hz). This frequency exceeds the normal operating range for good amplitude accuracy from a
proximity probe (e.g., Fig. 6-35). However, it should be recognized that FM measurements vary the frequency of the carrier, and amplitude is not important. As
long as the transducer provides a clear pattern of the passing teeth, any changes
in timing are detectable, and may therefore be demodulated into torsional displacement data.
The third category of FM signals consists of actual machinery excitations
that generate FM signals. This can occur in machines such as gear boxes, electri-
362
Chapter-7
cal machines such as generators and motors, as well as bladed machines such as
steam and gas turbines. When multiple excitations are tied into the same fundamental rotational speed frequency on the same rotor, or on mating rotors, the
opportunities for generation of FM signals are extensive. In many cases, machinery excitations may appear as amplitude or frequency modulation, or a combination thereof. The machinery diagnostician should be fully aware that AM and FM
signals are possible on even simple machines. This basic recognition will save a
lot of time identifying sidebands, and allow the diagnostician to examine the
potential physical reasons for the AM or FM signals. It should also be mentioned
that machines can also exhibit other types of interactions such as phase modulation. These are less common physical occurrences, and they are beyond the current scope of this text.
Case History 19: Gear Box with Excessive Backlash
A simple two element speed reduction gear box consisted of an 80 tooth
high speed pinion driving a 220 tooth low speed bull gear. This was a down mesh
arrangement with a normal pinion operating speed of 4,228 RPM, and an associated bull gear speed of 1,537 RPM.
Fig. 7-60 documents the normal bull gear spectrum plots obtained from a
vertical casing accelerometer at the outboard bearing. The top plot covers a nominal frequency range of 0 to 500,000 CPM. This plot reveals a gear mesh frequency at 338,220 CPM, with an amplitude of 1.3 Gs,o-p. The lower plot displays
a translated FFT of the frequency domain immediately surrounding the gear
mesh frequency. It is clear that the gear mesh frequency is not influenced by any
other significant excitation.
The FFT plots in Fig. 7-61 describe the normal pinion characteristics. This
acceleration data was obtained in the vertical plane at the blind end pinion bearing cap. Gear mesh frequency occurs at a slightly lower amplitude of 0.7 Gs,o-p,
and a slight modulation of the mesh by bull gear speed (Xls) is apparent. Since
the bull gear is substantially larger than the pinion, it is common to see the
influence of the large bull gear element upon the small pinion. This type of activity is visible in the low frequency domains around running speed, and around the
identified gear mesh frequency.
These high frequency vibration response characteristics remained quite
consistent until the gear box was disassembled for inspection during a routine
maintenance overhaul. Unfortunately, the pinion bearings were damaged beyond
repair when a 6 300# weld neck flange was inadvertently dropped on the bearings. This necessitated the installation of a set of spare pinion bearings from the
warehouse. Due to time restraints, the spare bearings were installed with no
consideration of proper fit, or relative position of the pinion versus the gear.
Subsequent startup and loading of the unit resulted in the generation of a
significantly different high frequency whine from the box. The vertical casing
accelerometer mounted on the bull gear displayed the FFT data in Fig. 7-62. It is
noted that speeds are only 0.5% higher than the earlier data set, and that the
Fig. 760 Bull Gear Casing Acceleration

With Normal High Frequency Behavior
363
Fig. 761 Pinion Casing Acceleration With

Normal High Frequency Behavior
gear mesh amplitude is similar to the previous data. However, the post-overhaul
bull gear plots revealed a modulation of the gear mesh (GM) frequency by the
low speed bull gear (Xls), and the high speed pinion (Xhs).
These upper and lower side bands of the running speeds were also evident
on the pinion data as shown in the Fig. 7-63. Although the relative amplitude of
the pinion sidebands appeared to be lower than the bull gear, the absolute values
Fig. 762 Bull Gear Casing Acceleration

With Abnormal High Frequency Behavior
Fig. 763 Pinion Casing Acceleration With

Abnormal High Frequency Behavior
364
Chapter-7
were actually higher on the pinion. In addition, the gear mesh frequency on the
pinion bearing had increased to over 6.0 Gs,o-p. In the final analysis, a mesh discrepancy was apparent, and the unit was shutdown. Further inspection revealed
that the backlash was excessive due to improper lateral position of the pinion
bearings. Once again, production pressures are combined with human error and
poor judgment to render a good machine inoperable. Fortunately, in this case the
correction was easily achievable, with minimal expense and downtime.
BIBLIOGRAPHY
2. Harris, Cyril M., Shock and Vibration Handbook, Fourth edition, chap. 14, New
York: McGraw-Hill, 1996.
3. The Fundamentals of Signal Analysis, Application Note 243, (Hewlett Packard,
printed in USA, 1995).
4. Wolfram, Stephen, The Mathematica Book, 3rd edition, software version 3.0
Champaign, Illinois: Wolfram Media/Cambridge University Press, 1996.
5. Zwillinger, Daniel and others, CRC Standard Mathematical Tables and Formulae,
30th edition, p. 49, Boca Raton, Florida: CRC Press Inc., 1996.
Data Acquisition and Processing 8
ndustrial processes operate over a wide

range of pressures, temperatures, and flow rates. This variety of operating conditions dictates a similarly large distribution of machinery types and configurations. Although the variables involved are extensive, a common focal point
appears when reliability and mechanical integrity are discussed. In virtually all
cases, it is economically favorable to evaluate the onstream condition of the
machinery. Furthermore, when problems do appear, early detection and proper
diagnosis are mandatory. In the pursuit of parameters to access mechanical condition, the analysis of machinery vibration characteristics has consistently
proven to be a powerful tool.
Successful vibration analysis requires an intimate familiarity with various
types of measurements, application and characteristics of transducers, plus
capabilities and limitations of diagnostic instrumentation. Ultimately, the data
must be reduced to hard copy for evaluation. The format used for data presentation can enhance the information, yielding a direct identification of the occurring
mechanism, or it can submerse the relevant data in a sea of inclusive information. In previous chapters, the basic measurements and transducer characteristics have been discussed. In addition, dynamic signal characteristics and signal
manipulation techniques have been reviewed. At this point, it is appropriate to
address the integration of a contemporary suite of industrial transducers with
the instrumentation required to properly acquire and accurately process high
resolution data. Finally, the traditional formats used for documentation of steady
state and transient data will be reviewed.
VIBRATION TRANSDUCER SUITE

Since the late 1960s, the eddy current proximity probe has been widely
accepted for obtaining relative shaft vibration and position measurements. In
most cases, these transducers are used to measure the motion of a rotating shaft
with respect to a stationary element such as a bearing. This type of sensor is eminently suitable for machinery protection and trend analysis, due to its capability
to measure both vibration and relative changes in position between elements.
Proximity probe calibration is generally consistent with API specifications
365
366
Chapter-8
of 200 mv/Mil, 5%. The total linear operating range is specified to be a minimum of 80 Mils by the same API1 specifications. Frequency response for proximity probes are often stated as DC (zero frequency) to 600,000 CPM (10,000 Hz).
However, at high frequencies, displacement amplitudes are quite small, and typically fall below the noise floor of the measurement system. For this reason, the
most significant probe data occurs between DC and approximately 90,000 CPM
(1,500 Hz). Certainly this range is more than adequate for addressing the anticipated vibratory behavior of most large industrial machinery trains.
A pair of mutually perpendicular proximity probes are typically installed at
each journal bearing on a machinery train. Normally these probes are mounted
at 45 from the true vertical centerline as shown in Fig. 8-1 of a three bearing
turbine generator set. Thrust bearings are protected by dual axial probes as indicated on the turbine. To provide synchronous tracking and filtration capability, a
once-per-revolution Keyphasor probe is usually installed on each train. Since
these timing probes are permanently mounted in a fixed location, the resultant
phase data is considered as an absolute measurement.
Casing vibration transducers may be installed on a temporary basis for
diagnostic measurements, or they may be permanently installed and connected
to vibration monitors. Historically, two fundamental types of casing vibration
transducers have been applied: the velocity coil, and the piezoelectric accelerometer. Both types of transducers provide relative motion of the casing with respect
to free space. These types of sensors are also referred to as seismic, or inertially
referenced transducers.
Velocity transducers contain a spring mounted mass enclosed by an outer
coil that responds to machine vibration. The motion of the spring mass system is
either damped electronically or with an internal viscous fluid. Since this is a
mechanically activated system, it is limited in overall frequency response. Typically, a velocity coil will exhibit a low frequency rolloff of 600 CPM (10 Hz), with
a high frequency limitation in the vicinity of 90,000 CPM (1,500 Hz). This frequency range is suitable for measurement of casing vibration occurring on
medium and high speed machinery. It should be recognized that a velocity coil is
generally inappropriate for very low speed measurements. This is due to the
attenuation of amplitudes, and phase distortion inherent with the low frequency
damping of the spring mass resonance. At high frequencies, this transducer cannot respond mechanically, and the upper frequency limitation appears.
Velocity coils are self-generating devices that do not require an external
power source. Output sensitivity for these transducers varies between 500 and
1,080 mv/IPS. Due to these high scale factors, the resultant electronic vibration
signals are strong, and easily observed with a variety of readout devices. The
major difficulties with this type of transducer are the limitations associated with
the sensing mechanism, i.e., the spring mass. This type of pickup is susceptible
to spring breakage, and as previously mentioned, it is limited in overall fre1 Vibration, Axial Position, and Bearing Temperature Monitoring Systems API Standard
670, Third Edition, American Petroleum Institute, (Washington, D.C.: American Petroleum Institute, November 1993).
Vibration Transducer Suite
367
Inlet
Exhaust
Normal
Thrust
Solid Coupling
Steam
Turbine
1ADa
1ADb
1VD
1HD
Generator
3 - 60 Hz
2VD
45 45
CW
2HD
3VD
45 45
2VA
2HA
45 45
3HD
45 45
3VA
3HA
45 45
Rotation and Probe Orientation

Viewed from the
Turbine Governor
Fig. 81 Typical Machinery Arrangement With Shaft And Casing Vibration Transducers
quency response. In addition, most industrial velocity coils are fairly large, and
this can restrict installation in many locations.
Due to these velocity coil limitations, accelerometers have emerged as the
preferable vibration transducer for seismic measurements. In general, industrial
accelerometers are smaller, lighter, more durable, and they cover a significantly
larger frequency range than velocity coils. In most accelerometers, the transducer contains an internal mass, a piezoelectric crystal, and an integrated circuit
(ICP). The crystal is generally quartz or ceramic, and the application of force by
the internal transducer mass produces an electrical charge. This charge sensitive
signal is converted to a voltage signal proportional to acceleration in the ICP
circuit, or in an external charge amplifier. The final conditioned output signal
carries the engineering units of millivolts per G of acceleration.
Accelerometers are manufactured in a multitude of configurations. Transducers are available that can successfully operate in cryogenic environments at
temperatures below -350F, and other units are designed for sustained high temperature operation at well over 1,200F. Some accelerometers are designed for
low frequency measurements, with operating ranges of 6 to 1,800 CPM (0.1 to
300 Hz). These transducers are low noise devices with output scale factors that
range from 500 to 10,000 mv/G. At the opposite end of the frequency domain,
miniature, high frequency accelerometers can reach 1,200,000 to 1,800,000 CPM
(20,000 to 30,000 Hz), with typical scale factors ranging from 1.0 to 5.0 mv/G.
The accelerometers used for casing measurements on process machines are
less exotic. The normal frequency range for these industrial transducers extends
from 600 to 600,000 CPM (10 to 10,000 Hz). The output sensitivity for these
368
Chapter-8
accelerometers is generally in the vicinity of 100 mv/G. These probes are not limited by the mechanical movement characteristics of velocity coils, and the top
end frequencies are significantly higher. As such, accelerometers are suitable for
measuring high frequency machinery excitations such as gear meshing and
Blade Passing frequencies.
For demonstration purposes, two sets of mutually perpendicular accelerometers are shown in the machinery diagram in Fig. 8-1. In this example, the bearings on each side of the generator are equipped with X-Y accelerometers
mounted at 45 from the true vertical centerline. These transducers are directly
in-line with the shaft sensing proximity probes. Typically, the accelerometers are
mounted on the same machine element (e.g., bearing housing) as the shaft displacement probes.
The acceleration signals may be double integrated to yield bearing housing
displacement with respect to free space. The shaft proximity probes measure relative motion between the shaft and structure upon which the probes are
attached. If the proximity probes are mounted on the bearing housing, then the
probes detect relative motion between shaft and the housing.
Performing an electronic summation between the shaft relative and the
casing absolute displacement provides the useful measurement of shaft absolute
motion. For machines with flexible supports or foundations this can be an
extraordinarily important measurement. Conversely, machines with very rigid
support structures will display minimal casing motion, and the shaft absolute
vibration will be closely approximated by the shaft relative motion.
The vibration transducers are generally designated with some type of logical probe identification scheme. For example, the probes mounted on the
machine train shown in Fig. 8-1 are identified with a common three character
code. Within this code, the first character refers to a specific bearing. The 1 identifies the turbine governor end, 2 denotes the turbine exhaust, and 3 designates
the outboard generator bearing. The second letter refers to probe orientation.
Specifically, the letter V refers to a vertical or Y-axis probe, H specifies a horizontal or X-axis probe, and A is used for the axial (thrust) probes. The third character of D or A identifies the transducer as a displacement proximity probe, or an
accelerometer. Thus, a three character identification code is used to uniquely
describe the location, orientation, and type of vibration transducer. The only
exceptions to this code are the Keyphasor probe (Key), and the a and b designations added to the axial probes to define the two channels directed to the
thrust monitor.
Other static or dynamic transducers may also be installed on a machinery
train to continuously monitor a specific parameter, or measure a particular
dynamic characteristic during an investigative test. Additional static devices
such as thermocouples, LVDTs, valve position, or any number of process or load
measurements may be recorded. The supplementary dynamic transducers
include devices such as pressure pulsation pickups, force transducers, strain
gauges, and torsional vibration transducers. Many of these probes contain ICP
electronics, and share a common power supply and signal conditioning.
Recording Instrumentation
369
RECORDING INSTRUMENTATION
The machinery transducer systems discussed in the previous section are
generally terminated at one or more racks of permanent monitors. Although
some racks are field mounted close to the machinery, the majority are installed
in a local or a central control room. The monitor racks are normally configured to
provide digital or analog outputs to a Distributed Control System (DCS). These
monitor output signals are generally used for trending of overall values, plus correlations with process conditions.
The machinery monitoring system may also provide digital outputs to a
separate dynamic or transient data acquisition system such as a Bently Nevada
Transient Data Manager (TDM). This type of computer-based system provides
improved visibility of vibratory characteristics. Whereas the DCS will trend
overall unfiltered vibration amplitudes, this auxiliary system will trend characteristics such as FFT data and vectors at various frequencies (1X, 2X, etc.). This
system may also be equipped to capture transient startup or coastdown data.
Although these are very useful systems, it should be recognized that these types
of systems have limitations in terms of resolution and sampling characteristics.
For situations where an automated data processing system does not exist,
or where improved data resolution is required, or in cases where further signal
manipulation is anticipated the use of separate recording instrumentation is
mandatory. Fortunately, the tools and techniques to accurately record dynamic
data with low noise levels have been available since the advent of the FM tape
recorder in the early 1970s. This device allows accurate recording and reproduction of multiple channels of complex dynamic signals in a continuous time
record. The recording media is magnetic tape, and configurations ranging from
reel-to-reel, cassette, VHS, and DAT formats have been successfully employed.
During the evolution of tape recorders, size and weight have decreased, and
capabilities have increased. Although it is tempting to use a tape recorder as a
stand-alone instrument, it is much more effective if it is integrated into a complete data acquisition system. For example, consider the diagram in Fig. 8-2 of a
typical multichannel analog field data acquisition package.
In Fig. 8-2, the raw transducer signals are terminated at a multichannel
switching box. This device allows AC coupling of signals such as proximity
Transducer Signals
Multichannel
Switching Box
TEAC XR-510
14 Channel
Tape Recorder
HP
3560A
5,250
9.75
TK83
DMM
Tektronix
2205
Scope
Fig. 82 Typical Multichannel Analog Field Data Acquisition System
370
Chapter-8
probes or accelerometers (via internal coupling capacitor). It also provides a way

to switch and compare each transducer signal with the corresponding tape
recorder output signal. This comparison is most effectively performed visually on
an oscilloscope. A check of signal integrity is repeated many times during a data
acquisition session to insure accuracy of the recorded and reproduced information. Thus, a primary function of the field Tektronix oscilloscope is to continually verify the validity of the TEAC tape recorder signals.
Voltage amplifiers may be included as separate units, or the amplifiers may
be incorporated as an integral part of the tape recorder. For separate amplifiers,
the signal gain is usually indicated by a rotary or toggle switch. Common voltage
amplifier settings include gains of x1, x2, x5, x10, x20, x50, and x100. These are
often combined with attenuation settings of x0.5, x0.2, and x0.1. For tape recorders that include built-in amplifiers, the determination of signal gain is not necessarily defined directly by the tape recorder. In most cases, the user establishes an
input voltage level, plus an output voltage range. The signal gain across the
recorder is then determined by equation (8-1).
Output Voltage Range
Gain recorder = ------------------------------------------------------------------Input Voltage Range
(8-1)
On many recorders the input voltage range covers values such as 0.1, 0.2,
0.5, 1.0, 2.0, 5.0, and 10.0 volts. Depending on the specific tape recorder,
the output voltage range may include a similar set of steps. It may also be fixed
at a level of 1.0 volt, or it may be adjustable with a potentiometer. In any case,
the ratio of output to input will determine the signal gain across the tape
recorder. There are also occasions where external amplifiers are used in conjunction with the internal recorder voltage amplifiers. In these situations, the overall
signal gain is given by equation (8-2).
Gain overall = Gain external Gain recorder
(8-2)
In the general application, voltage amplifiers are used to amplify low level
signals, and reduce the magnitude of large signals. For example, assume that an
input signal is amplified by a factor of 5 with the internal (or external) tape
recorder amplifier(s). If the direct transducer output signal is viewed on an oscilloscope at 0.2 volts per division, the tape recorder output signal should be
checked with an oscilloscope setting of 1.0 volt per division. Conversely, presume
that a strong Keyphasor pulse is recorded on tape at a gain of 0.1. The raw
input signal might be viewed at 5 volts per division, and the recorder output signal would be correctly observed at a scope setting of 0.5 volts per division. In
either case, the tape recorder gain or attenuation value must be equal to the
ratio between the voltage setting on the two oscilloscope channels. Clearly, a
visual comparison of dynamic data on the oscilloscope must always be adjusted
for any gain or attenuation of the recorded signals.
The internal record and reproduce amplifiers are supplied in two configurations. The most useful type for machinery analysis is the frequency modulated
(FM) amplifier. In this type of recording, the signal to be recorded modulates a
371
high frequency carrier. This FM amplifier allows recording of low frequency (DC)
voltages, and it provides a flat bandwidth of frequencies that are accurately
recorded and reproduced. Specific performance of a typical FM system on a VHS
tape recorder are summarized in Table 8-1.
Table 81 Typical FM Tape Recording Characteristics In A VHS Format
Tape Speed
(Inches/Second)
Tape Speed
(Cm/Second)
Bandwidth
(Hertz)
Signal to Noise
Ratio (dB)
Record Time
(Minutes)
30.0
76.2
0 to 20,000
50
5.4
15.0
38.1
0 to 10,000
50
10.8
7.50
19.05
0 to 5,000
50
21.7
3.75
9.52
0 to 2,500
48
43.4
1.875
4.76
0 to 1,250
47
87.0
0.938
2.38
0 to 625
46
173.
0.469
1.19
0 to 313
42
347.
From Table 8-1, it is clear that tape speed and the recording bandwidth are
directly related. That is, as tape speed changes by a factor of 2, the bandwidth
responds in a similar fashion. For example, at a recording speed of 15 inches per
second, an FM channel will have a frequency range extending from 0 (DC) to a
top end of 10,000 Hz (600,000 CPM). Doubling the tape speed will increase the
bandwidth to 20,000 Hz (1,200,000 CPM). Moving in the opposite direction, cutting the tape speed in half to 7.5 inches per second will reduce the recording frequency range in half to a value of 5,000 Hz (300,000 CPM).
It is noted from Table 8-1 that the signal to noise ratio is defined in terms of
decibels (dB). This is just another way of expressing a voltage ratio, and the following equation may be used to convert the voltage ratio to decibels:
Voltage 1
dB = 20 log -------------------------
Voltage 2
(8-3)
This expression is computed on a log to the base 10. If the dB value is

known, then equation (8-3) may be reconfigured to calculate the voltage ratio as
shown in the next expression (8-4).
dB
-------Voltage 1
20
------------------------- = 10
Voltage 2
(8-4)
From Table 8-1, the signal to noise ratio at the higher tape speeds is 50 dB.
If the full scale, or reference Voltage1 is set at 1.0 volt, then the tape channel can
372
Chapter-8
ideally resolve voltage levels Voltage2 per equation (8-4) as follows:

Voltage 1
1.00 Volt
1.00 Volt
1.00 Volt
Voltage 2 = ------------------------- = --------------------- = --------------------- = --------------------- = 0.0032 Volts
2.5
50
316.2
dB
10
------------ 20
20
10
10
Thus, the noise floor is nominally 3 millivolts. A properly calibrated amplifier should be able to accurately record and reproduce voltages above this level,
and below full scale voltage. In actuality, a conservatively accurate tape recording should not approach the bottom, and it should not exceed the top of the voltage range. One should strive to adjust recording gains so that the minimum
signal level is at least ten times the noise floor, and the maximum signal does not
exceed the full scale setting. In this example, accurate data could be expected if
the recorded signal was between 0.03 and 1.00 volts. If the dynamic signal falls
below the noise floor, it will be lost due to the small amplitudes. On the other
hand, if the signal significantly exceeds the amplifier full scale voltage setting,
the signal will be lost due to overdriving of the amplifier. In either case, the
required dynamic data will not be retrievable.
Clearly, tape recording amplifiers must be carefully adjusted to stay above
the noise floor, and below the full scale voltage. This is generally easy to do on
steady state vibration data where the signals are reasonably constant in amplitude. However, this can become a real challenge on transient data where the
machine must pass through a critical speed, and the maximum vibration amplitudes are unknown. In this situation, the machinery diagnostician must provide
an educated guess of the potentially largest vibration amplitudes that may be
encountered. For instance, if the diametral bearing clearance is 15.0 Mils, it may
be assumed that the peak amplitude through the critical speed region would be
approximately 50% of the total bearing clearance. This would indicate that the
highest shaft vibration amplitude might be in the vicinity of 7.5 Mils,p-p. This
would be equivalent to 1.5 Volts,p-p if probe sensitivity was 200 millivolts per Mil.
Thus, if the voltage amplifier was set at 1.0 Volt, (or 2.0 Volts,p-p) there is good
confidence that accurate data could be acquired as the rotor passed through the
critical speed range.
Alternatively, the control room operators could be questioned as to the typical maximum vibration amplitudes that are observed during transient speed
conditions. This information may be obtained by observations of the vibration
monitors, or it may be documented within the DCS or TDM system. It should
also be noted that startup versus coastdown vibration amplitudes may be quite
different. Hence, the tape recorder amplifier gain settings that were fully acceptable during a startup condition may be totally inappropriate for coastdown data.
From a frequency response standpoint, it was previously stated that displacement and velocity signals are generally limited to frequencies below 1,500
Hz (90,000 CPM). This frequency range is well within the domain of FM amplifiers as shown on Table 8-1. However, if higher frequencies (e.g., accelerometers)
must be recorded, Table 8-1 reveals the major weakness of an FM recording sys-
373
Table 82 Typical Direct Tape Recording Frequency Characteristics In A VHS Format

Tape Speed
(Inches/Second)
Tape Speed
(Cm/Second)
Bandwidth
(Hertz)
Signal to Noise
Ratio (dB)
Record Time
(Minutes)
30.0
76.2
0.1 to 150,000
30
5.4
15.0
38.1
0.1 to 75,000
30
10.8
7.50
19.05
0.1 to 36,000
30
21.7
3.75
9.52
0.1 to 18,000
30
43.4
1.875
4.76
0.1 to 9,000
30
87.0
0.938
2.38
0.1 to 4,500
30
173.
0.469
1.19
0.1 to 2,250
30
347.
tem on VHS tape. That is, the time duration of the recording is severely limited
by the high tape speeds, combined with the 246 meter length of the VHS tape
cartridge. For instance, at full speed of 30 inches per second, the entire VHS tape
will only last for 5.4 minutes.
If it is necessary to record high frequency data for extended periods of time,
the FM record and reproduce electronics must be replaced with direct record and
reproduce amplifiers. This type of recording system is virtually the same technology that has been used for audio tape recorders for many years. A set of typical
characteristics for direct recording on VHS are tabulated in Table 8-2. From this
summary, it is clear that the frequency bandwidth of the recording has been substantially increased, but the signal to noise ratio has suffered. At all speeds the
direct signal to noise ratio is 30 dB. This is equivalent to a voltage ratio of 31.6. If
the full scale voltage is set at 1.0 Volt, then the tape channel can ideally resolve
voltage levels of:
Voltage 1
1.00 Volt
1.00 Volt
1.00 Volt
Voltage 2 = ------------------------- = --------------------- = --------------------- = --------------------- = 0.032 Volts
1.5
dB
30
31.62
--------
------
10
20
20
10
10
Thus, the noise floor is nominally 32 millivolts, and a properly calibrated
amplifier should be able to record and reproduce voltages above this level. However, it must always be recognized that the direct electronics are 20 dB, or a factor of ten, noisier than FM amplifiers. In actual practice, a tape recorder may be
configured with a combination of FM and direct record and reproduce amplifiers.
In all cases, the FM cards are used for high resolution, accurate phase coherence
between channels, and the inclusion of low frequency characteristics. The direct
cards are used for high frequency transducers, and it is understood that voltage
accuracy on direct channels is traded for an extended frequency range.
The characteristics presented in Tables 8-1 and 8-2 are typical for analog
tape recorders operating with VHS tapes that are 246 meters (807 feet) long.
Other reel-to-reel, and cassette recorders provide similar performance, but the
374
Chapter-8
total recording time must be adjusted for the actual length of the magnetic tape.
These conventional FM and Direct recorders use stationary record and
reproduce heads that contain a defined spacing between tracks. This standardization of tape heads allows interchangeability of tapes between recorders. Various other types of recording configurations are also available. For instance, if FM
extra wide band record and reproduce electronics are used, the frequency
response ranges in Table 8-1 will be doubled. However, the dynamic range (signal
to noise ratio) will be reduced by approximately 3 dB. Again, there are many possible tape recorder configurations, and the machinery diagnostician must be
fully aware of the performance of any tape recorder used for serious data acquisition work.
Another approach to acquiring and storing dynamic data consists of using
DAT technology. In these types of recorders, an analog to digital converter is
used to digitize the data prior to recording. A rotating record head (typical speed
of 2,000 RPM) uses a helical scan to record the digital data on small cassettes.
During reproduction, the dynamic data is converted from digital to analog format, and directed to the output connectors. DAT recorders typically have a signal to noise ratio of greater than 70 dB. This is 20 dB, or 10 times greater than
the previously discussed FM recorders using VHS tape. Tape speed on DAT units
is constant at values such as 0.321 inches/second, which provides a two hour
recording time on one 60 meter cassette.
Initially, it might appear that DAT recorders have significant technical
advantages over traditional FM recording electronics. However, this conclusion
is rapidly altered when the recording bandwidth characteristics of DAT units are
considered. In this type of magnetic tape recorder, the recording bandwidth is
dependent on the number of data channels. For instance, Table 8-3 describes the
DAT frequency response characteristics for various channel quantities:
From this tabular summary it is clear that the frequency response characteristics are totally based upon the number of data channels. Hence, 2 channels
may be recorded for the entire length of the tape (i.e., 2 hours), with a bandwidth
of DC to 20,000 Hz. If 16 channels of data are required, the bandwidth drops to
Table 83 Typical DAT Tape Recorder Frequency Response Versus Number Of Channels
Number of
Channels
Bandwidth
DC to 20 kHz
Bandwidth
DC to 10 kHz
8
10
12
16
Bandwidth
DC to 5 kHz
Bandwidth
DC to 2.5 kHz
4
8
8
4
8
16
375
DC to 2,500 Hz (still 2 hours recording). If this is an acceptable frequency span,

then the DAT recorder is a good choice. However, if more recording controls are
mandatory over the full array of available channels, then the diagnostician
should give serious consideration to traditional fixed head recorders.
Regardless of the type or configuration of the tape recorder, calibration and
operation of the unit must be documented at some point during the data acquisition process (usually at the beginning). This is normally achieved by recording
one or more calibration signals on the tape, and logging the specific amplitude
and frequency of the applied sine wave. The calibration signal may be internal to
the recorder, or it may be obtained from an external function generator. Independent of the source of the calibration signals, the tape output must always match
the input calibration signals.
Typically, the frequency of the calibration signal will be set to some reasonable value such as the machine running speed. In other cases, a predetermined
frequency at some constant value may be used. In other situations, it may be
necessary to cover a wide frequency range with FM and direct amplifiers, and
multiple calibration signals should be recorded. For example, consider a machine
that operates at 12,000 RPM (200 Hz), and assume that it is driven by a gear box
that has a mesh frequency of 540,000 CPM (9,000 Hz). If it is important to record
both frequencies, then FM amplifiers would be used for the shaft displacement
signals, and direct amplifiers would accommodate the high frequency acceleration signals. A calibration signal at a frequency of 200 Hz would be used for the
FM channels, and a 9,000 Hz cal signal would be recorded on the direct channels.
In this manner, calibration signals are recorded at the frequencies of interest,
and a high level of confidence may be placed in the final data.
In all cases, accurate records of the calibration signals must be maintained.
In addition, descriptive information of the various data runs must also be generated. In order to provide the proper level of documentation for accurate processing of the data, it is necessary to consider the other instruments included in the
field data acquisition system shown in Fig. 8-2. These additional devices are used
to examine the dynamic signals, and to provide reference information for proper
reproduction of the tape recording.
For instance, synchronous 1X amplitude and phase information may be
obtained with a Digital Vector Filter (either DVF or single channel TK83). This
information should be logged for each data channel on each tape run. When
viewed in conjunction with the overall, unfiltered amplitudes, and the operating
speed, it is possible to perform an initial evaluation of the machinery based upon
the tape log. However, the additional importance of this information lies in the
ability to support the accuracy of the taped data during post processing. For
example, if the hand logged information indicates a 1X vector of 1.6 Mils,p-p at
146, and the data extracted from the tape exhibits 1.57 Mils,p-p at 145, then the
values are equivalent. In this case, the diagnostician will have a high level of
confidence in the tape recorded data.
The proximity probe gap voltages should always be measured and documented for analysis of lateral and axial positions. This measurement can be eas-
376
Chapter-8
ily obtained with a battery operated Digital Multimeter (DMM). It goes without
saying that this measurement must be made on the raw or direct transducer signal before AC coupling. The probe gap voltages should initially be acquired at
zero speed with the lube oil turned off, and with the lube oil warm and circulating. This establishes the initial starting point for the radial and axial position
measurements. Gap voltages should also be obtained at each definitive constant
speed point such as slow roll, minimum governor, full speed, and full process
load. A DMM is also highly useful for routine items such as checking cable continuity, verifying power supply voltages, and gaping proximity probes.
Observation of frequency domain information during data recording can be
quite important, particularly on complex signals. In these situations it is important to use a dual channel DSA (such as the HP3560A) during data acquisition.
Modern units are quite compact, and a full set of features are available in a DSA
that weighs less than ten pounds. Many field DSAs are also equipped with digital storage. With this capability, they may be used for performing routine surveys, or making multiple structural or piping measurements in an effort to
define or locate a particular frequency node or anti-node. A dual channel DSA is
also quite handy for performing simple resonance tests with an accelerometer
and an impact hammer equipped with a force transducer.
Finally, tape logs must be established and maintained throughout the field
tests. A sample log for a 14 channel recorder is shown in Fig. 8-3. This log identifies the machinery, and the magnetic tape number. It also provides necessary calibration information for recording an internal calibration signal. Each data run
is specifically identified, and the date, start and stop times, plus tape recorder
speed and machine rotation speed are listed. Each of the 14 channels is adequately defined with the pickup location (e.g., 1V, 2H, etc.), the transducer type
(e.g., D, V, or A), and the overall recording signal gain. Probe gap voltages, plus
Client/Location:
Machine:
Cal. Date
Tape #
Gear Ratio:
Start Cal. at:

End Cal. at:
Ch
Loc
Calib. Signal Source=

XR-510 Internal
P/U Gain DC Gap
Tape Speed=
Cm/Sec
O/A
1X-Amp
Page
of
Calib. Amplitude=
Calib. Frequency=
2.0 Vp-p = 0.707 Vrms
Hz
1X-
Comments
Run I.D.
Date
Start
Counter
Time
Start:_______
End:________
AM / PM
Tape Speed
Cm/Sec
RPM
End
Counter
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Run I.D.
Date
Time
Start:_______
End:________
AM / PM
Tape Speed
Start
Counter
1
2
3
4
5
6
7
8
9
Fig. 83 Portion Of A Typical Fourteen Channel Tape Recorder Field Log
377
overall vibration levels, and synchronous vectors are also tabulated. This type of
tape recorder log is very useful when different transducers with different configurations are recorded on consecutive runs.
However, there are many situations when the instrumentation is set up
with only one group of transducers, and the system configuration remains constant throughout the entire test. For example, a field balance on a steam turbine
might include four radial proximity probes, two casing transducers, two thrust
probes, and a Keyphasor. If the slow roll and full speed vector information was
hand logged on the tape recorder logs (like Fig. 8-3), the necessary data would be
available, but it would be somewhat difficult to compare. In these situations, a
consecutive field vibration log, as shown in Fig. 8-4, might be more suitable.
Client/Location:
Machine:
Probe ID
Page
of
Gear Ratio:
Date:
Time:
RPM:
Gap
(-Volts DC)
Overall (Mils,p-p)
1X Amp.(Mils,p-p)
1X Phase (Degrees)
Gap
(-Volts DC)
Overall (Mils,p-p)
1X Amp.(Mils,p-p)
1X Phase (Degrees)
Gap
(-Volts DC)
Overall (Mils,p-p)
1X Amp.(Mils,p-p)
1X Phase (Degrees)
Gap
(-Volts DC)
Overall (Mils,p-p)
1X Amp.(Mils,p-p)
1X Phase (Degrees)
Gap
(-Volts DC)
Overall (Mils,p-p)
Fig. 84 Portion Of A Field Vibration Data Log For Varying Conditions
On this type of data log, the emphasis resides on accurate identification of

the transducer type, location, and angular orientation, plus the signal parameters of DC gap, overall vibration, and rotational speed vectors. For balancing
work, or comparative operation at constant speeds or loads, this type of format
provides an easy scan of the machinery vibration and position data.
As the complexity of things increase, and the pressures to acquire and process a lot of data in a limited amount of time the time required to fill in
numerous hand logs may not be available. In addition, complicated machines
often have complex vibration response characteristics, and logging of a single
running speed vector may not be adequate to define the behavior. In these cases,
computer-based systems may be employed in conjunction with the data recording system previously discussed. For example, Fig. 8-5 describes the addition of a
pair of Bently Nevada, 208 Data Acquisition Interface Units (DAIU). These 8
channel units (16 channels total) are wired in parallel with the tape recorder system. The vibration data is acquired and digitized in the DAIUs, and downloaded
to a Compaq laptop computer. The data is processed and manipulated into various formats within the computer, and hard copy results are available by parallel
378
Chapter-8
Transducer Signals
Digital Signals
Multichannel
Switching Box
BNC 208 DAIU

LCD Display
TEAC XR-510
14 Channel
Tape Recorder
BNC 208 DAIU
9.75
DMM
Tektronix
2205
Scope
5,250
TK83
Compaq
Laptop
LaserJet 5L
Printer
Fig. 85 Typical Multichannel Analog Plus Digital Data Acquisition & Processing System
port connection to the HP LaserJet 5L printer.

A variety of flexible options are possible with the data acquisition system
described in Fig. 8-5. For instance, if the primary project objective is field balancing, then the main emphasis would be placed on acquisition and processing of
the hard copy data. In this instance, the tape recorder system would provide a
full analog backup. However, if the fundamental project requirements demanded
the acquisition of high resolution data for additional post processing, the tape
recording might be assigned as top priority, and the field data processing system
used for periodic examination of certain portions of the acquired database. Certainly other combinations are quite probable, and the diagnostician should consider this type of instrument array as complimentary. In all cases, each piece of
this instrumentation package should be used to its full capabilities to properly
document the behavior of the machinery train.
As a final note, it should be mentioned that utilization of a PC-based data
acquisition system can result in the development of numerous data files. These
files may become quite large, and it is not unusual to occupy 10 to 12 megabytes
of storage with one sample run. Certainly this type of digital data may be stored
on the computer hard disk. It is highly desirable to provide a backup set of field
data on a separate mass storage media. The use of another hard drive, magneto
optical disks, or magnetic media of 100 megabytes or more is a good approach.
Procedurally, it is recommended that at the conclusion of each data run, that the
diagnostician store the data on the PC hard disk, and then make a backup copy
on the external mass storage device. In order to keep track of the various data
sets, a computer storage log sheet as described in Fig. 8-6 is recommended.
In addition to the data description, plus the date and time of the acquired
field data, a specific file name is assigned to this set of digital information. For
obvious reasons, the same file name should be used for the hard disk as well as
the backup data file. It is also desirable to include the file size as defined down to
the last byte. This gives an indication of the amount of data, and it provides a
good checkpoint to verify that the backup is identical in size to the original version stored on hard disk.
The first column in Fig. 8-6 is specified as Tab Section. The purpose for this
heading is to allow for organization of the hard copy field data. In many cases,
Data Processing Instrumentation
Tab
Section
Data Description
379
Date
Time
File Name
File Size
(Bytes)
Fig. 86 Portion Of A Computer Data Storage Log For Multiple Sample Runs
the quantity of data plots becomes unmanageable, and a methodical organization of the data is necessary. In these common situations, one successful
approach is to equip a 3-ring binder with a set of divider tabs. As each consecutive data set is sampled and processed to hard copy format, the plots are placed
in tab sections of the binder, and tab section identification listed in the computer
data storage log sheet. This may seem like a minor item, but when cross referencing or checking data in the middle of the night, it can be enormously helpful.
DATA PROCESSING INSTRUMENTATION

Following field recording of vibration and other dynamic signals onto magnetic tape, or digital storage, the next step is to extract the meaningful information from the data. Historically, most of the devices used for data processing were
analog instruments that produced hard copy output with pen plotters, and photographs of instrument CRTs. Today, most instruments are digital devices that
supersede their predecessors by exhibiting higher resolution, expanded capabilities, and improved interfaces between devices. The current data processing systems are often computer controlled, and they generally employ a combination of
instruments from various vendors.
A typical data processing system is shown in Fig. 8-7. This sketch is
descriptive of the system that was used to produce most of the data plots contained within this text. It is clear that a variety of analog and digital devices
have been integrated into this high resolution data processing system. Many of
the instruments within this system are both self-checking and self-calibrating.
However, manual system checks are periodically performed using a function generator (HP-33120A) in conjunction with a full 5 1/2 digit digital multimeter (HP3468A), and various frequency measuring devices (e.g., HP-35670A or DVF).
Analog system input is obtained by directing the TEAC XR-510 tape
recorder output signals through a switching panel into a rack of voltage amplifiers. These step gain amplifiers are used to maintain a suitable and consistent
voltage input range for the various instruments. Following voltage amplification
380
Chapter-8
Analog Signals
BNC Model 29191
Voltage Amplifiers
BNC 208 DAIU
BNC 208 DAIU
Krohn Hite 3323 Filters
Multichannel
Switching Box
Tektronix
5110R
Scope
TEAC XR-510
14 Channel
Tape Recorder
5100
Digital Signals
HP-35670A
Dynamic
Signal Analyzer
Compaq PC
2.1
HP-33120A
9.7504
HP-3468A
BNC Digital
Vector Filter 2
273
HP-IB
HP54600B
Digital
Scope
17 Color Monitor
30.125
Tektronix
5110N
Scope
HP-9153C
Disc Drive
HP-9000
Series 360
Computer
R-G-B
HP-98752A
19" High Res. Color CRT
HP-C2001A
Laser Jet 4M
Printer
Fig. 87 Typical Lab Based Multichannel Data Processing System
or attenuation, the dynamic signals are processed through low-pass and/or highpass Krohn-Hite filters to allow early removal of any extraneous high or low frequency components. Although most data does not require this type of filtering, it
is mandatory in certain data processing situations.
Following the Krohn-Hite filter, the dynamic signals are paralleled into the
Digital Vector Filter (DVF), the Dynamic Signal Analyzer (HP-35670A), and the
four channel Tektronix (5110R) oscilloscope. Typically, two scope channels are
used for the unfiltered X-Y signals, and the other two channels are dedicated to
X-Y signals filtered at rotational speed by the DVF. With this arrangement, filtered and unfiltered orbital and time domain signals can be observed on the four
channel oscilloscope. This analog data is also compared with the final computer
processed plots to insure accurate and consistent results.
A separate two channel Tektronix (5110N) scope is used to monitor the
Keyphasor signal, plus the DVF trigger pulse. Most of the data processing functions are dependent upon a suitable speed trigger. Hence, it is mandatory to continually observe the Keyphasor pulse, and perform any necessary amplifier
adjustments to maintain a clean trigger to the other instruments. In some cases,
the machine Keyphasor signal may be of poor quality due to any number of
transducer difficulties. The cleanup of an inferior Keyphasor signal pulse may
normally be accomplished by using two stages of voltage amplifiers.
As demonstrated in Plot A of Fig. 8-8, the direct Keyphasor signal exhibits
a substantial amount of surface (baseline) variation, and a trigger pulse that has
a height of only 1.0 volt. This Keyphasor signal shape will not trigger the downstream instruments due to the small pulse height, combined with the potential
for false triggers from the erratic upper portion of the curve. Fortunately, the sig-
Data Processing Instrumentation
381
Fig. 88 Manipulation Of Machine Keyphasor Signal With Voltage Amplifiers
nal may be salvaged by applying two voltage amplifiers in series.

The first amplifier is DC coupled, and a DC offset plus a voltage gain are
applied to the direct Keyphasor signal. Specifically, the results of this initial
manipulation are shown in Plot B of Fig. 8-8. The DC bias voltage flattened out
the erratic upper portion of the initial signal, and the voltage amplification of 5
increased the usable pulse height to approximately 3 volts. Although signal clarity has been significantly improved, the pulse in Plot B is still of minimal height,
and the pulse magnitude varies slightly with time. The final cleanup of this signal is provided by the second voltage amplifier in the series circuit. This device is
AC coupled, and the pulse signal is amplified by 5 to yield the time domain curve
shown in Plot C of Fig. 8-8. Note that the final pulse height of 9 volts is combined
with a truncated bottom portion of the pulse. This squared off pulse bottom is
due to overdriving the amplifier, and effectively bounding the trigger pulse.
Returning back to the data processing instrumentation in Fig. 8-7, a separate HP 54600B digital oscilloscope is used for observing data where a rapid
sweep rate, or the capabilities of a storage oscilloscope are required. This type of
digital scope also provides accurate and rapid on-screen measurements of signal
characteristics with the moveable cursors and digital parameter displays.
The HP-35670A Dynamic Signal Analyzer is a four channel FFT spectrum/
network analyzer with a frequency range that extends from nearly DC to slightly
over 100 KHz. This unit has a minimum dynamic range that exceeds 80 dB,
which allows total extraction of any data that originates from a tape recorder.
Although the DSA is primarily a frequency domain analyzer, it is also used to
make amplitude and time domain measurements. In fact this instrument performs most of the analog to digital data conversions. The DSA is used for processing Fast Fourier Transform (FFT) information, frequency response functions
(FRFs), and time history data. This unit has an adjustable resolution that varies
from 100 to 1,600 spectral lines per channel. It incorporates various averagers to
obtain statistical accuracy, plus several windows, and a variable zoom to allow
precise frequency identification. The DSA is used for sampling both steady state
and transient data. Upon confirmation of proper information on the internal
382
Chapter-8
CRT, the data is transferred to the HP 9000 series computer system for final processing and printing.
Most of the plots contained in this text were generated by interfacing the
DSA and the DVF with a dedicated HP computer system. As noted in Fig. 8-7,
the hardware includes an HP-9000 series workstation computer equipped with a
19 high resolution color CRT, a combination hard and floppy disc drive, plus a
LaserJet 4M printer. A dedicated IEEE-488 interface maximizes data transfer
rates between the computer and the instruments. A separate HP-IB interface
was used for rapid communications between the computer and the hard disk.
Finally, a third IEEE-488 interface is employed to transfer data between the
computer and laser printer. This system operates under HP Basic, with improved
speed provided by a binary compiler. The data processing software evolved from
Bently Nevada ADRE software. The original ADRE programs have been subjected to numerous revisions by the senior author to provide higher resolution,
increased speed, and improved hard copy output. In the overview, the DSA digitized and processed data, and the DVF provided synchronous filtration plus
phase and amplitude data. The computer performed sampling, data formatting,
and the production of the final hard copy laser plots.
ADRE for Windows software by Bently Nevada is used during the direct
digital processing path shown on the left side of Fig. 8-7. In this data path, the
analog signals are connected from the tape recorder output to the Bently Nevada
208 Data Acquisition Interface Units (DAIUs). As previously mentioned, two of
these units can simultaneously process 16 data channels of dynamic data. The
digital data is passed to a Compaq PC for manipulation, final formatting, printing and storage. This system operates the same in the shop as it does in the field.
In reality, the all digital processing system shares many common qualities
with the generally analog system. As time progresses, it is reasonable to assume
that more channels and more capabilities will be included in the interface box
between the transducer signals and the computer. It is anticipated that PCbased systems will continue to expand in software sophistication and the ability
to examine dynamic signals in greater detail with a variety of formats. For
instance, the use of Computed Order Tracking on Hewlett Packard DSAs eliminates the need for analog ratio synthesizers and digital order tracking filters
combined with their associated phase accuracy and dynamic range limitations.
Most PC-based systems have built-in math and analysis capabilities that allow
slow roll removal from orbits, Bode, and polar plots. They also contain curve filling software to allow characterization of frequency response functions (FRFs).
This is quite useful in modal analysis, plus the development of polynomial equations to define bearing housing stiffness, as discussed in chapter 4 of this text.
One of the most significant advantages of an all digital sampling system is
that the information may only need to be processed one time. This is particularly
important on machines such as large turbine generators that require one or two
hours to roll from turning gear up to minimum governor. In an all digital system,
the PC samples the entire event for all data channels simultaneously. Thus, at
the end of the startup, the machinery diagnostician is ready to start examining
the sixteen channels of transient data, and committing the necessary informa-
Data Presentation Formats
383
tion to hard copy format. If the T/G startup required two hours, and the data
review and printing required another two hours, the complete startup documentation package would be finished in four hours. However, if a fourteen channel
analog tape recorder is employed, and one channel is devoted to the Keyphasor,
the diagnostician is faced with a two hour startup, fourteen hours of dual channel playback, plus another two hours of data examination and printing. In this
case, the all digital system is the clear winner with four hours of total data acquisition and processing time compared against eighteen hours for the conventional
dual channel processing.
This substantial time savings may lead to the conclusion that analog
recording systems are obsolete, and that all future data will be acquired in an all
digital format. In actuality, this trend is not completely true. The magnetic tape
recorder still provides a true recording of the total dynamic signal, whereas the
digital system acquires samples at predetermined incremental speeds or times.
If an event occurred between digital samples, the digital system would miss it,
and the analog system would capture the event. This ability to do a cycle by cycle
analysis of the analog signals will not be feasible with all digital systems in the
immediate or foreseeable future.
Reliability is also an issue when dealing with the acquisition of unique or
non-repeatable data sets such as a machinery startup. In most operating plants
the startup or shutdown of a major machinery train is a significant event. The
machinery diagnostician has only one chance to acquire this type of necessary
transient data. If that opportunity is lost, it probably cannot be recovered. Specifically, if a tape recorder fails during a startup sequence, the section of data
recorded before the tape failure will normally be preserved. However, on an all
digital system, any type of computer, digital interface, or hard disk malfunction
will probably invalidate the entire data file. This is an embarrassing condition
for all parties involved, and the desirability of a full backup based upon an analog tape recording of the transient events is self-evident.
It should also be mentioned that backup on magnetic media is only a shortterm solution. Magnetic tapes and floppy disks are all subject to damage, and
deterioration. These types of magnetic media should only be considered as temporary or interim storage devices. For long-term storage of 20 or more years, the
diagnostician should invest in Magneto-Optical or CD-R storage devices.
DATA PRESENTATION FORMATS

Machinery vibration characteristics processed by the previously discussed
system are presented on several distinct types of plots. The first format consists
of vibration signals in the time domain as shown in Fig. 8-9. This is normally
combined with a shaft orbit where mutually perpendicular radial probes are
installed, and phase referenced with a Keyphasor mark.
During field data acquisition, this same information is observed on an oscilloscope. The final hard copy data consists of computer-generated orbits and time
base plots that accurately represent the analog signals. Since the radial trans-
384
Chapter-8
Fig. 89 Unfiltered Orbit

And Time Base Plots
ducers are often not installed at true vertical and horizontal positions, the orbits
must be corrected for this angular deviation. The orbits in Fig. 8-9 are rotated to
allow true vertical and horizontal representation. It should also be noted that
the computer processed time base plots have not been corrected for transducer
orientation. Thus, the time domain plots are representative of the signals as
viewed on an oscilloscope.
In order to clearly observe the synchronous 1X vibration of the machinery,
the data is also filtered precisely at rotational speed, and another set of orbits
and time domain plots are produced. Fig. 8-10 is identical to the previous display,
with the inclusion of the 1X running speed filter. On some data sets the influence
of shaft scratches and other frequency components substantially reduces the
analytical usefulness of the orbital data. In these cases, 1X filtered data is almost
mandatory to evaluate the machinery behavior.
Fig. 810 Filtered Orbit

And Time Base Plots
Another type of common steady state data consists of a frequency analysis

of the vibration signal. This is achieved by the DSA that uses Fourier analysis
techniques to separate a time domain signal into discrete frequency components.
Information of this type is presented on an X-Y plot where frequency (Cycles/
Minute) is plotted along the horizontal axis; with linear vibration amplitude on
the vertical axis. The following Fig. 8-11 displays a spectrum plot of a shaft displacement signal. Certainly any dynamic transducer signal such as displacement, casing velocity or acceleration, pressure pulsation, force, strain, torsional
385
Fig. 811 Steady State

Spectrum Plot Of Shaft
Vibration
displacement or velocity may be processed in this manner.

Statistical accuracy in an FFT plot is obtained by employing RMS averaging. This averaging results in a smoothing of random noise variations, but does
not reduce the noise level. Hence, RMS averaging is representative of the actual
frequency distribution within each signal. The spectrum data is normally processed with a Hann window filter. This passband is similar to those found in
swept frequency spectrum analyzers, and combines a good compromise between
amplitude accuracy and frequency resolution across the analysis bandwidth.
As shown in Fig. 8-11, rotational speed (1X) motion generally appears as
the highest spectral peak. However, in certain cases such as severe oil whirl,
other components may dominate. In order to quantify the frequency components,
machine running speed is obtained from the DVF, or from a DSA with a tach
input. Once this speed is established, the frequency of the other components are
divided by speed to determine their respective harmonic order. Often, an even
order will coincide with an internal geometrical configuration, allowing the spectral component to be linked to a machine element. For example, vane passing frequency is the product of the number of impeller vanes times running speed.
In other cases, a particular frequency component is significant towards
quantifying the machinery behavior. An example of this behavior would be an
excitation from an adjacent unit. Typically, the frequency components are identified on each plot for future documentation and reference. For synchronous components the harmonic order is noted; and for asynchronous peaks, the measured
frequency is indicated in Cycles per Minute at the top of each plot. A special designation is generally included when data from a gear box is processed. For example, Fig. 8-11 was obtained from a bull gear, and the associated Pinion identifier,
plus the respective frequency of 9,310 RPM, are shown on the plot. For high
speed pinion signals, the bull gear speed would be similarly displayed.
Another type of steady state data consists of a frequency response function
(FRF) between signals. The example presented in Fig. 8-12 describes a FRF
between proximity probes mounted at opposite ends of a turbine. This hard copy
data consists of three interrelated plots including the Amplitude Ratio (dimensionless Mils/Mils), relative Phase between signals (Degrees), and Coherence
386
Chapter-8
(dimensionless) between signals. This type of information is used to examine the

amplitude and timing characteristics at specific frequencies. Data validity is verified or ignored based upon the coherence function. In general, coherence must
be sufficiently high (e.g., greater than 0.95) to allow confidence in the amplitude
and phase data. In Fig. 8-12, only four frequencies display acceptable coherence,
and these four frequencies represent the only acceptable data in the entire plot.
The machine running speed on Fig. 8-12 is 5,340 RPM, and the amplitude shows
about a 2.2:1 ratio between inlet and exhaust ends of this turbine. The phase difference at this frequency is 188 that might seem to be out of phase shaft vibration. However, the radial probes are physically mounted 180 apart, as shown in
Fig. 4-7. Hence, the FRF phase value of 188 is really indicative of an 8 phase
differential, or an in-phase motion across the gas turbine.
Fig. 812 Steady State

Frequency Response
Function (FRF) Between
Proximity Probes At Opposite Ends Of A Gas Turbine
The behavior at twice rotative speed 10,680 CPM, plus the coupled pinion
speed at 9,300 RPM may also be examined for appropriate amplitude ratios and
differential phase angles. FRF plots are normally performed with RMS averaging, and a Flat Top window to minimize amplitude errors. This type of data presentation is often the only way to accurately relate amplitude and phase
characteristics at frequencies other than rotative speed.
The last type of steady state data to be considered is the change in radial
rotor position with respect to the stationary bearing. For example, Fig. 8-13
depicts the lateral clearance of a plain circular bearing with two axial oil inlet
grooves. In this diagram, the shaft orbit is noted in a normal running position for
387
a counterclockwise rotating machine, and the X-Y proximity probes are also
shown in this sketch. Since these probes produce a DC signal proportional to the
average gap between the probe tip and the observed surface (shaft), this measurement may be used to locate the journal running position within the bearing.
Vertical Probe
Horizontal Probe
Vertical
Clearance
45
Fig. 813 Shaft Centerline

Position Shift In A Circular
Bearing With Counterclockwise Shaft Rotation
45
CCW
Rotation
Bearing
Center
Shaft Orbit
Horizontal
Clearance
In practice, DC gap voltages from each pair of perpendicular X-Y probes are
measured and logged throughout the field tests. This DC voltage data is then
converted to a differential basis of at speed minus the zero speed voltage. The
change in shaft centerline with respect to each probe is obtained by division of
this differential voltage by the probe sensitivity (e.g., 200 mv/Mil). Finally, a vector summation of the changes with respect to each X-Y probe yields the overall
change in shaft radial position. This resultant vector describes the magnitude
and direction of the centerline change from the initial starting, or rest point.
Another, and more descriptive way to display this type of radial position
information is presented in the Fig. 8-14. This plot exhibits a continuous sample
of DC gap voltages that describe the shaft centerline position during a startup
Fig. 814 Measured Shaft

Centerline Change During
Startup Of An Expander
With Clockwise Rotation
388
Chapter-8
condition. Again, this data is based upon the changes in static DC proximity
probe gap voltages. The starting point is at the bottom of the bearing (zero
speed), and changes are plotted with respect to that initial starting point. Information of this gender is used to observe variations in shaft centerline position as
a function of speed, time, temperature, or any other appropriate variable. In the
specific case of Fig. 8-14, the shaft exhibited a smooth transition from slow roll
speed of 170 RPM to full operating speed of 7,940 RPM. Since this expander was
equipped with tilt pad bearings, the documented change in journal position was
indicative of normal and expected behavior for this clockwise rotating journal.
The previously discussed categories of data are obtained under constant
speed, and steady state operating conditions. The next major group of data formats considers the presentation of transient information. Typically, this includes
observation of vibratory changes as a function of speed and/or time. This type of
information is presented as speed change plots, synchronous rotational speed
vector plots, plus frequency spectra, and order tracking from a DSA. In most
instances, all four types of data are necessary to provide proper and complete visibility of machinery transient behavior.
The variation of machine rotative speed with respect to time is an often
ignored piece of information. As shown in Fig. 8-15, this type of data is a linear
plot of shaft rotational speed in RPM on the horizontal axis, versus elapsed time
in seconds or minutes on the vertical axis. The data presented in Fig. 8-15 documents the startup of a gas turbine from slow roll to crank speed, plus the final
acceleration up to minimum governor at 4,600 RPM. Time history data of this
Fig. 815 Speed Ramp

Of A Gas Turbine During A
Routine Cold Startup
389
category allows identification of plateau periods where speed remained relatively

constant, plus sections where rapid speed changes occurred. Precise definition of
both of these conditions are extremely helpful in understanding the machinery
transient characteristics. A comparison of startup versus coastdown behavior
will often help explain the differences between the machine vibration differences
of a controlled startup versus a coastdown.
Note that these types of data plots also compute the rate of speed change at
various portions of the curve (e.g., 25.5 RPM/Second). Thus, it is possible to compare defined speed change rates under different transient conditions or situations. This is also convenient for determination of acceptable speed rate changes,
and the settings of proper speed ramps on electronic governors.
Transient vibration data consisting of synchronous rotational speed 1X vectors are initially accommodated by the Bode plot where rotational speed vibration amplitude (1X) and phase angle are plotted as a function of machine speed.
The example in Fig. 8-16 depicts a typical data set for a shaft displacement
probe. Again, other transducers may be used for examining vibration changes as
a function of machine speed. This type of data is appropriate for both startup
(speed increasing), and coastdown (speed decreasing) conditions. It is meaningful to note that startup versus coastdown behavior will be different due to the
variations of stiffness and damping occurring in the machine bearings. Hence, it
is desirable to document and review both the speed increasing as well as the
speed decreasing data. A variation of this format considers amplitude and phase
versus time. This type of information is useful for evaluating large machines that
Fig. 816 Bode Plot Of A

Startup
390
Chapter-8
require a long time to achieve a full heat soak (e.g., large turbine generator sets).
The Bode is often combined with a polar plot describing the locus of rotational speed vectors during speed changes. Although both of these plots provide
the same data array; the Bode provides excellent visibility of changes with
respect to speed, and the polar yields improved resolution of phase variations.
The diagram presented in Fig. 8-17 is representative of a polar plot for a single
Fig. 817 Polar Plot Of A

Startup
radial probe. Note that the zero degree point is always located at the transducer
angular position. Thus, a comparison of data from a pair of perpendicular probes
mounted at a journal will result in a 90 difference in the zero point on the plot.
If a machine has symmetrical bearing stiffness, vertical and horizontal behavior
will be identical, and the polar plots from a pair of probes will track together.
However, most machines exhibit deviations in stiffness and damping, and this is
reflected and visible in the associated polar plots. This type of data is essential
for identifying rotor critical speeds, and the influence of secondary system resonances. In most cases, the synchronous 1X vectors are plotted with amplitudes in
Mils,p-p, and phase angles expressed as Degrees of phase lag. If other transducer
signals are plotted in a polar format, the engineering units assigned to the vector
magnitude would be appropriately adjusted.
Under machine conditions where significant sub or supersynchronous
vibration components are generated, it is necessary to generate a cascade or
waterfall plot of individual spectra at incremental operating speeds or times.
This type of data presentation provides an excellent overview of the frequency
content of the vibration signals as a function of operating speed or time. Fig. 8-18
391
Fig. 818 Cascade Plot

Of A Gas Turbine Startup
depicts a typical cascade plot with FFT samples acquired at 60 RPM increments.
The harmonic orders for 1X, 2X, 3X, etc. are shown to identify the even order frequency components. In addition, other order lines at 0.5X, and 1.78X are also
exhibited. In the case of this machine, the 1.78X represents the vibration frequency of the pinion in the coupled gear box. In all cases, the correct amplitude
engineering units are displayed on the right hand axis. Thus, the magnitude of
any spectral peak may be compared, and scaled directly with this legend. As
noted with the Bode plot, this type of data is generally sampled and processed
during both startups and coastdowns. It is also appropriate to process this type
of data as a function of time to observe changes that occur with parameters such
as load or casing temperature. These types of stacked spectrum plots acquired
over time are commonly referred to as waterfall plots, whereas the speed variable plots are generally called cascade plots.
Although cascade or waterfall plots provide excellent visibility of the frequency components, they can miss some information due to the time span
between samples. For instance, in Fig. 8-18, the FFT samples were acquired at
60 RPM increments. Although the final cascade plot appeared to be correct, there
could be additional unobserved or undocumented activity occurring between
sampled blocks of data. To address any potential lapse in information, it is desirable to increase the sampling rate for specific frequency components. In some
instances it is also meaningful to observe spectral components as a function of
392
Chapter-8
Fig. 819 Order Tracking

Of A Gas Turbine Startup
speed or time. To address both of these potential situations, the utilization of

Computed Order Tracking2 is highly desirable. An example of this type of information is exhibited in Fig. 8-19, where the first four harmonics of a gas turbine
radial vibration signal are plotted against speed during a normal cold startup.
This type of data allows improved visibility of individual frequency component
behavior over and above the previously discussed cascade or waterfall plot.
Although the same DSA is used for both tasks, the order tracking capability of
the DSA firmware allows substantially more samples for each component, and
thereby provides increased resolution.
Another potentially difficult data processing scenario occurs when the
major excitations are not synchronous. In other words, the largest vibration
amplitudes occur at frequencies that are not integer multiples of running speed.
This type of excitation might be completely missed by an order tracking analysis
of the data. The only positive method to insure that the maximum amplitudes
have been properly identified and documented is to examine the time history
record of each signal. An initial check can be performed by reproducing the taped
data into an oscilloscope. If any high amplitude excursions are detected, then an
additional signal processing technique of transient capture may be employed.
For instance, Fig. 8-20 depicts a transient capture plot of a hydro generator
during load rejection. This data was produced by reproducing the taped data into
the DSA, and configuring the DSA as a high speed digital recorder. Operating in
2 Ron Potter and Mike Gribler, Computed Order Tracking Obsoletes Older Methods, SAE
Technical Paper Series Proceedings of the 1989 Noise and Vibration Conference, Traverse City,
Michigan (May 1989), pp 63-67.
393
this mode, the analog data is digitized and stored into the memory of the DSA.
The data is not processed during the transient capture, it is just digitized and
stored in RAM. In essence, this information replicates a high speed time base
recorder, and an initial plot of vibration amplitude versus time is produced. A
review of the capture buffer display reveals the areas or events of interest that
may be expanded into various formats and examined in greater detail.
Fig. 820 Transient Capture Of A Hydro Turbine

Driven Generator During
Load Rejection
Once this transient capture data resides within the memory of the DSA, the
information may be further dissected, and appropriate post processing applied to
the digital data. Depending on the specific data array, the application of cascade,
waterfall, or Computed Order Tracking plots may be necessary. Another option
would be to perform a time domain cycle by cycle analysis of two or more transient signals to determine relative phase or timing relationships. In any case, the
objective is to fully document the recorded dynamic signals, and allow the execution of a meaningful and accurate analysis of the data.
In the final overview, it is quite clear that no single data processing technique is appropriate to all machine configurations, and to all types of mechanical
malfunctions. In fact, there are substantial differences in the applicability of the
various data processing techniques with different vibration sensors. These differences between shaft displacement proximity probes, casing velocity coils, and
casing accelerometers are summarized in Table 8-4 for steady state operation.
Clearly, proximity probes are suitable for time, orbital, and frequency domain
analysis. Obviously, any of the three transducer outputs may be examined with
Table 84 Applicability Of Various Steady State Data Formats To Different Transducers
Steady State
Data Type
Shaft
Displacement
Casing
Velocity
Casing
Acceleration
Time Domain
Yes
Sometimes
Sometimes
Orbital Domain
Yes
Sometimes
Seldom
Frequency Domain & FRF
Yes
Yes
Yes
394
Chapter-8
frequency analysis. If the signals are not overly complex, casing velocity may be
used for time or orbital domain observation. However, casing acceleration signals
are seldom applied and used for orbital analysis. On some types of mechanical
events, such as the multiple impacts exhibited by rolling element bearings, casing accelerometers may be successfully used for time domain analysis of the
vibration signals.
The applicability of the three different types of vibration transducers to
various types of transient data are summarized in Table 8-5. Shaft displacement
may be used for any of the categories, but casing velocity or acceleration will not
yield position change data. All three transducers may be used with cascade,
waterfall, or trend plots, but the signal complexity on the casing pickups may
negate some of their effectiveness. The use of Bode and polar plots for casing
velocity and acceleration is dependent on the machine type, and the availability
of a rotational speed signal for use with the casing mounted transducer.
Table 85 Applicability Of Various Transient Data Formats To Different Transducers
Shaft
Displacement
Casing
Velocity
Casing
Acceleration
Cascade or Waterfall
Yes
Yes
Yes
Bode or Polar
Yes
Sometimes
Seldom
Position Change
Yes
No
No
Transient Capture or Trend
Yes
Yes
Yes
Transient Data Type
Overall, these tabular summaries reinforce the previous discussions of

using displacement transducers for low frequency measurements, and the need
to use FFT processing for the high frequency measurements produced by accelerometers. As with any general rules, exceptions always occur, and the diagnostician must keep an open mind. In all cases, the machinery diagnostician must
employ the best combination of transducers, instrumentation, and data processing formats to examine and dissect the information extracted from the data set.
Understanding the motion characteristics of the mechanical elements under
examination requires looking at the vibratory behavior from many different
angles, and applying the best available techniques and instrumentation.
BIBLIOGRAPHY
1. Potter, Ron and Mike Gribler, Computed Order Tracking Obsoletes Older Methods,
SAE Technical Paper Series Proceedings of the 1989 Noise and Vibration Conference, Traverse City, Michigan (May 1989), pp. 63-67.
Common Malfunctions
rocess machines are subject to many

malfunctions that range from internal forcing functions, to self-excited mechanisms, to external forces, plus a myriad of physical phenomena that may impose
dynamic loads upon the machinery. Some of the malfunctions (e.g., unbalance)
are common to all rotating machines. Other excitations (e.g., gear mesh forces)
are unique characteristics for a particular type of mechanical device. These various excitations are normal for all moving elements, and they form a significant
portion of the behavioral parameters for the specific piece of machinery.
When the forces responsible for the excitations increase beyond normal or
expected limits, this is often detrimental to the integrity of the equipment. This
physical change in applied forces is often detectable as a change in the machinery vibration response characteristics. In order to provide the diagnostician with
some additional insight into these mechanical relationships, these excitations
and the resultant malfunctions will be examined. Sequentially, the common
types of machinery malfunctions are reviewed, and specific case histories are
presented in this chapter. These common malfunctions are applicable to most
rotating machines, and they include forced as well as free vibration mechanisms.
In addition, a series of unique excitations associated with specific machine types
will be presented and discussed in the following chapter 10.
SYNCHRONOUS RESPONSE
The synchronous, or running speed, or fundamental, or 1X motion of a
rotating element is an inherent characteristic of every machine. It should be recognized that all machines function with some level of residual unbalance. All
machines must operate with some finite clearance between stationary and rotating elements. Since it is physically impossible to produce a perfectly straight and
concentric rotor, another source of synchronous motion is apparent. In addition,
all machines are supported by various compliant structures and foundations.
Vibration response measurements on any machine with virtually any
transducer will reveal a component at rotational frequency. Not surprisingly, this
universally common excitation accounts for the majority of the machinery malfunction mechanisms. Unfortunately, the analysis of 1X vibration is significantly
395
396
Chapter-9
complicated by the fact that many different mechanical malfunctions appear as

changes in the rotational motion. Hence, the machinery diagnostician is faced
with the real dilemma of observing a single frequency, and attempting to diagnose the origin of an increased vibration amplitude.
The first line of attack resides in reviewing the traditional relationship discussed earlier in this text as:
Force
(9-1)
It is clear that the vibration response is directly proportional to the applied

force when the restraint or stiffness is held constant. As force increases, the
resultant vibration response will also increase. This type of relationship is intrinsic with the fundamental concepts associated with activities such as rotor balancing. It is also clear that vibration response is inversely proportional to the
restraint or stiffness when the applied force is held constant. In this condition, as
the restraint decreases, the resultant vibration response will increase. For
instance, a bearing with increasing clearances will typically exhibit a reduction
in support stiffness (i.e., restraint). Assuming a constant unbalance force, the
rotor will vibrate at a higher response level.
The third possibility for changes in vibration response amplitudes must
consider variations in both the force and the restraint. As stated by Donald E.
Bently in an issue of Orbit1 magazine:
Vibration is usually either the result of a bowed rotor or is the result of a
force or moment acting on the stiffness of that machine to that force or moment. As
such, vibration is actually a ratio and frequently not an end objective measurement in itself. Remember that forces and moments flow through, moments are
measured across, and:
Dynamic Forces or Moments
Dynamic Motion ( Vibration ) = -------------------------------------------------------------------------------------Dynamic Stiffness
Thus, to best read the behavior of a machine, it is often necessary to know
BOTH the numerator and the denominator of this simple relationship Obviously, you must make some sort of assumption of stiffness or force in order to have
a knowledgeable vibration measurement. We do this regularly and will continue
to do so. However, to improve our capabilities of operating machinery, the measurement of observed operating dynamic stiffness will become more important in
the future, as either the numerator (Dynamic Forces) may be incorrect, or the
denominator (Dynamic Stiffness) may be incorrect.
In many instances it is extraordinarily difficult to quantify the active forces,
or the associated restraints (stiffness). This inability to define actual machine
parameters often yields to an investigation of changes in response constituents.
For instance, the effective rotor support characteristics were discussed in chapter 4. This discussion concluded that for most types of process machinery that
1 Donald
E. Bently, Vibration levels of machinery, Orbit, Vol. 13, No. 3 (September 1992), p. 4.
Synchronous Response
397
the effective stiffness is related to the oil film stiffness and the overall bearing
housing stiffness in equation (4-16). This expression is restated as follows:
1
1
1
------------ = ----------- + ------------K oil K hsg
K eff
(9-2)
The bearing housing stiffness includes the support pilings and foundation,
the grout and baseplate, the bearings or machinery pedestals, plus the stiffness
of the bearing housing itself. Often a visual inspection of the machinery will
identify the condition of these mechanical elements. For example, it is quite clear
when a bearing housing is loose on a pedestal, or when grout degradation has
occurred. If these support elements remain in good condition, then the oil film
stiffness characteristics should be examined.
One of the most powerful and commonly available tools for evaluating bearing condition is an examination of the journal position within the bearing. This is
performed with radially mounted X-Y proximity probes as discussed in chapter 6
of this text. Specifically, the change in shaft centerline position was determined
with the vector example previously displayed in Fig. 6-20. For purposes of completeness, this same diagram is reproduced in the following Fig. 9-1. Within this
diagram, the change in probe DC gap voltages may be vectorially summed to
determine the overall shift in journal position from an initial stop condition to an
operating position of the shaft within the bearing.
90
Shaft Centerline
Shift Vector
135
ils
@
5
13
Fig. 91 Shaft Centerline

Shift Vector As Measured
With X-Y Proximity Probes
4.
4
ils
2
3.
45
5.44 Mils
81
45
180
Substantial changes in radial shaft position are often associated with bearing damage. This is particularly true for a horizontal machine that has experienced damage to the bottom half of the bearing. In these instances, the probe gap
voltages will reveal a vertical drop of the shaft into the babbitt. This type of damage often results in a change of the synchronous vibration combined with the
position shift. In some cases the 1X shaft vibration will increase, as shown in the
induction motor case history 44. In other situations, the running speed vibration
will decrease, as illustrated by the refrigeration compressor case history 50. In
both cases, the change in rotational speed vibration was associated with a distinct variation in bearing support stiffness. It should also be mentioned that the
398
Chapter-9
analysis of both of these case histories depended heavily on the measured

changes in shaft centerline position data.
Furthermore, variations in nominal bearing clearance (i.e., too big or too
small) represents one of the most common types of mechanical problems on process machinery. These abnormal clearances may be due to poor installation,
attrition of the bearings during operation, or the presence of some forcing function that is hammering out the bearing clearance. Clearly, there are visual and
measurement methods for evaluating changes in rotor support stiffness. If these
techniques do not identify any change in restraint (stiffness), then it is reasonable to conclude that the vibratory change is primarily associated with a change
in the applied force(s). The remainder of this chapter will be devoted to an examination of common forcing functions, and traditional excitation mechanisms.
These malfunctions will also be illustrated with descriptive case histories
MASS UNBALANCE
Mass unbalance represents the most common type of synchronous excitation on rotating machinery. Every rotor consists of a shaft plus a series of integral disks used for turbine wheels, or thrust collars. Turbomachinery rotors may
also include a series of slip on elements such as compressor wheels, pump impellers, thrust collars, spacers, coupling hubs, etc. Although each item is typically
manufactured to high dimensional tolerances, a residual unbalance is present in
each element. It is self-evident that the residual unbalance for a single machine
disk may be satisfactory, but the combined effect for a stacked rotor may be completely unacceptable as described by John East2. To address this issue, a variety
of tools and techniques have evolved to correct mass unbalance problems. Since
this is a fundamental problem with all rotating machinery, chapter 11 of this
text has been devoted to a detailed explanation of mass unbalance response, and
the variety of methods used to determine and correct rotor unbalance.
From a recognition standpoint, mass unbalance will normally produce a
transient Bode plot as shown in Fig. 9-2. This calculated plot for a forced unbalance spring-mass-damper system was extracted from Fig. 2-19 of this text. At
speeds well below the resonance, the vibration response will vary as the speed
squared. The applied centrifugal force may be estimated by equation (9-3):
2
RPM
F cent = Mass Radius ---------------
4, 000
where:
Fcent
Mass
Radius
RPM
=
=
=
=
(9-3)
Centrifugal Force Due To Residual Unbalance (Pounds)

Effective Mass of Residual Unbalance (Grams)
Effective Radius of Residual Unbalance (Inches)
Shaft Rotational Speed (Revolutions / Minute)
2 John R. East, Turbomachinery Balancing Considerations, Proceedings of the Twentieth Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University, College Station,
Texas (September 1991), pp. 209-214.
Mass Unbalance
399
Phase Lag (Degrees)
Low Damping, =0.1

30
60
High Damping,
=2.0
=2.0
90
120
=1.0
150
=0.2
=0.5
=0.1
180
5
Low Damping, =0.1
Amplitude Ratio
4
=0.2
3
2
=0.5
Fig. 92 Typical Mass

Unbalance Response As
Described By A Calculated Bode Plot
0
0.0
=1.0
0.5
1.0
1.5
=2.0
2.0
2.5
3.0
Thus, if the speed is doubled, the centrifugal force will be increased by a

factor of four. If the mechanical system is totally linear, the observed rotor vibration will also exhibit a fourfold increase in magnitude. For units such as turbines
or compressors that operate above a translational critical speed, the shaft
rotates about the mass centerline (principal inertia axis). This behavior produces
a plateau region of constant 1X amplitude and phase as shown in Fig. 9-2. Thus,
moderate changes in speed (within the plateau region) will have minimal effect
upon the synchronous vibration vectors.
When a machine is operating at full speed and load, a step change in the
running speed vectors (amplitudes and/or phase) may be indicative of a mass
unbalance shift. Part of a blade shroud or other minor attachment will manifest
as a 1X vector change in accordance with the deflected mode shape. Major unbalance changes such as a blade loss will also produce a running speed vector
change. In this case, the balance change will be significant, and concern should
be placed on how are we going to shutdown this machine, and pass through the
critical without causing further damage. In this type of situation, a rapid coastdown is desirable to minimize time within the bandwidth of the critical speed
domain. It might be advisable to bring the machinery train down under full load
to slow it down as fast as possible.
Typically, a pure mass unbalance problem will appear as a forward and circular shaft orbit. The orbit could also appear elliptical if the machine contains a
significant difference in vertical versus horizontal stiffness, or if the rotor is sub-
400
Chapter-9
jected to a shaft preload in addition to the unbalance. Finally, the phase relationship across the rotor will be in accordance with the deflected mode shape, and
the location of the probes along the axis. It is easy to be confused by traditional
rules that say things like identical phase angles across a machine are representative of mass unbalance. This statement is only true for a specific set of conditions, for a particular group of machines. Many cases of mass unbalance (e.g.,
turbine generator case history 39) will exhibit a phase relationship other than a
pure inphase motion across the machine.
It is highly recommended that the diagnostician become intimately familiar
with the specific topics of shaft mode shapes in chapter 3, dynamic signal characteristics chapter 5, and rotor balancing chapter 11 before attempting to diagnose
a mass unbalance problem. Furthermore, the study of 1X synchronous behavior
of a rotating system will reveal important information on the specific characteristics of the machine. This type of information will provide significant benefits in
malfunction diagnosis of the rotating equipment.
BENT OR BOWED SHAFT

Bent rotors and shaft bows represent another major class of synchronous
1X motion. It was previously mentioned that all machine parts contain some
finite amount of residual unbalance. In a similar manner, all assembled horizontal rotors (and some vertical rotors), will exhibit varying degrees of rotor bows.
In some cases, such as a light weight pinion with a short bearing span, the midspan deflection will be minimal. In other cases, the rotor will deform due to gravity. That is, the rotor will bend or bow under the influence of its own weight. An
example of this type of rotor would be a large gas turbine or steam turbine rotor.
These types of rotors will display a change in the gravitational bow by simply
lifting the rotor off a set of rollers, and setting it back down.
The shaft bow may be purely a gravitational bow, or it may be a thermally
induced bow. In either case, the force associated with the bent shaft is equal to
the shaft stiffness times the initial bow radius. This is considerably different
from the mass unbalance case, where the initial force is equal to the product of
residual unbalance mass, radius, and speed squared. The bent shaft will exhibit
a variable speed characteristic that is similar to the diagram in Fig. 9-3 that was
extracted from Fig. 2-18. This calculated plot for a forced spring-mass-damper
system was based upon the application of a constant force. This is identical to the
bent rotor condition where the applied force is equal to the deflection times the
shaft spring constant as shown in equation (9-4).
Deflection Shaft Stiffness
F bow = ---------------------------------------------------------------------------------1, 000
where: Fbow = Applied Force Due to Shaft Bow (Pounds)
Deflection = Maximum Midspan Deflection Due to Shaft Bow (Mils)
Shaft Stiffness = Lateral Shaft Stiffness (Pounds / Inch)
(9-4)
Bent or Bowed Shaft
401
Low Damping, =0.1

Phase Lag (Degrees)
30
60
90
High Damping,
=2.0
=2.0
120
=1.0
150
=0.2
=0.5
=0.1
180
5
Low Damping, =0.1
Amplitude Ratio
Fig. 93 Typical Lateral

Response Due To A Shaft
Bow As Described By A
Calculated Bode Plot
=0.2
3
2
=0.5
1
=1.0
=2.0
0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Since shaft stiffness is usually a very large number, a moderate deflection

(i.e., bow) will result in a substantial radial force. From Fig. 9-3, it is noted that
at speeds well below the resonance, the vibration response is equal to shaft bow.
Hence, a symmetrical rotor supported between bearings would display a maximum response at the middle of the rotor. A proximity probe or dial indicator
mounted at this location would reveal a runout vector equal to the bow magnitude and location. At speeds well above the resonance, the magnitude of the bow
would approach zero as the machine tends to rotate about the mass center, which
would be equivalent, or more precisely, coincident with the rotor bow center.
In the vast majority of cases, a thermal or gravitational shaft bow may be
rolled out by extended operation at slow roll speeds. In most instances the 1X
vectors are monitored, and when minimum and constant runout values are
achieved, the machinery may be safely started with a minimal shaft bow. In
other situations, the severity of the bow is of such a magnitude that the rotor
cannot be straightened by slow rolling. Localized heat application may be used to
relieve the bow, or the rotor may be suspended vertically and heated in an oven.
In other cases the rotor must be completely scrapped (e.g., case history 17).
In all cases, the diagnostician must recognize that rotor bows are inherent
with rotating machinery. Furthermore, the shaft bow may consist of complex
curves instead of a simple catenary. Also, the synchronous force from any shaft
bow will vectorially interact with the synchronous forces due to unbalance or any
other 1X forcing function. Hence, the final measured rotational speed vectors
probably include contributions from more than one synchronous excitation.
402
Chapter-9
Case History 20: Repetitive Steam Turbine Rotor Bow

A 35,000 horsepower steam turbine was subjected to an extensive overhaul
that included replacement of the 11 stage rotor. The maintenance work was performed with minimal difficulties, but major problems were encountered when a
routine overspeed check could not get past slow roll conditions. Although initial
uncoupled turbine runout vibration levels were quite acceptable, the 1X vectors
significantly increased at slow roll. At 1,000 RPM, vibration amplitudes reached
2.5 and 3.1 Mils,p-p at the governor and exhaust ends respectively. This manifestation of high vibration was combined with in-phase deflection of the turbine
rotor. Specifically, the governor end horizontal probe had a 261 phase angle,
with 265 displayed by the exhaust end horizontal pickup.
This unusual bowed rotor behavior was repeated on multiple runs covering
a time period of two days. At this point, historical rotor records were examined,
and the following general conclusions and observations were reached:
All of the attempted turbine solo runs were aborted due to high rotor vibra-
tion. This was indicated by the control room monitoring instrumentation, as

well as the physical sensation of excessive deck vibration.
During the occurrence of the high turbine vibration, a pure translational
(inphase) shaft bow was clearly evident across the rotor.
For this particular turbine, shaft vibration amplitudes should remain essentially constant between 300 and 1,400 RPM.
Initial appearance of the shaft bow was independent of speed.
Initial appearance of the bow was a function of time and temperature. Specifically, the shaft bow appeared following approximately two hours of operation in a warm (>200F) casing.
Cold shaft runout vectors were repeatable, and consideration of any cracked
shaft malfunction was discontinued.
Rotor inspection records revealed acceptable runout along the length of the
turbine. Furthermore, the shop balance was performed to low levels of
residual unbalance. Hence, the observed behavior was not associated with a
cold rotor bow, or a mass unbalance problem.
Records revealed that the spare turbine rotor installed during this overhaul
had not operated in the turbine casing for five years. At that point in history, the plant suffered a catastrophic fire, and this spare rotor was essentially baked in the casing for several days as the fire burned itself out.
A turbine rotor is machined from a solid forging, and all wheels and thrust
collars are integral with the shaft. This type of rotor assembly is heat treated
and tempered as part of the manufacturing process. These heat treatments of the
alloy forging are performed to obtain specific mechanical properties, and they are
implemented by controlled heating and cooling of the rotor. Hence, the turbine
rotor is constructed of a steel alloy that was subjected to various heat cycles during fabrication. It is reasonable to believe that such an assembly might be sensi-
Bent or Bowed Shaft
403
tive to the heating and cooling anomalies associated with the previous plant fire.
Based upon these measurements and observations, it was reasoned that the
turbine rotor contained a residual stress that was probably inflicted during the
fire. This residual stress manifested as a shaft bow whenever the rotor was
heated above ambient temperature. Conversely, the shaft bow was not apparent
when the rotor was cold. It is logical to assume that the emergency shutdown
produced a thermal bow as the hot rotor rested between bearings, on the interstage labyrinths. As the fire subsided, the turbine cooled over a period of several
days. It was postulated that as the rotor cooled, it returned to a straight condition at ambient temperature, and locked in the residual stress from the thermal
bow. Since all rotor repairs, inspections, runout checks, and shop balancing were
performed with a cold rotor, this type of internal stress would be undetectable.
If this hypothesis was accurate, then correction of the residual thermal
shaft bow would require additional heat treatment, combined with continuous
rotation of the turbine rotor. Obviously, this type of repair would be difficult to
perform in most shop repair facilities. However, the turbine casing provided a
means to heat the rotor with inlet steam, plus the ability to turn the rotor at controlled speed. Thus, the opportunity existed to perform an on-line stress relief of
the rotor within the actual turbine casing. The ASM International defines stress
relieving as: Heating to a suitable temperature, holding long enough to reduce
residual stresses, and then cooling slowly enough to minimize the development of
new residual stresses.
From this common definition, both heating and cooling must be combined to
stress relieve the rotor. During original manufacturing of this rotor, the heat
treatment temperatures are quite high. It must be recognized that the exhaust
casing has much lower temperature limits (circa 300F). Thus, any field stress
relieving of the rotor in the turbine casing must be limited in the heat soak temperature. Based upon the mechanical parameters of the installed turbine system, the following on-line rotor pseudo-stress relieving procedure was developed:
1. Operate the turbine at a slow speed of nominally 300 RPM for approximately 30 minutes with the sealing steam off, and a cool turbine casing.
This is the cooling portion of the cycle.
2. Apply shaft sealing steam, and allow turbine speed to increase to approximately 500 RPM with the improved vacuum.
3. Increase speed, and monitor the radial shaft vibration at both journal bearings. Continue to increase speed until the unfiltered radial vibration amplitudes approach a maximum of 4.0 to 5.0 Mils,p-p (based on conservative use
of the 12 to 14 Mil diametrical turbine bearing clearance).
4. Operate the turbine at this heat soak condition for approximately 60 minutes. During this time period, the rotational speed and steam flow should
be adjusted to maintain a maximum unfiltered radial vibration amplitude
between 4.0 and 5.0 Mils,p-p, combined with a maximum exhaust casing
temperature of 250F, plus a maximum speed of 1,400 RPM.
5. Following 60 minutes of high speed, high temperature, and high vibration
404
Chapter-9
operation; the turbine speed should be reduced back to 500 RPM. Seal
steam should then be removed, and rotational speed returned back to 300
RPM for a repeat of the cooling cycle, Step 1.
6. The previous Steps 1 through 5 should be repeated until the shaft vibration
amplitudes remain essentially constant between 300 and 1,400 RPM. When
this consistency of slow roll vibration data is achieved, the turbine may be
started, and operated normally.
This procedure was implemented, and the first 60 minute hot run was limited to 654 RPM. A total of six additional cold to hot runs were completed, and
the results of these consecutive runs are summarized in Fig. 9-4. This diagram
consists of 1X radial vibration amplitudes measured at the end of each cold 30
minute run, plus each hot 60 minute run. Note that these 1X filtered amplitudes
are slightly less than the unfiltered, overall vibration levels mentioned in the
procedure. For simplification, only the horizontal probes at governor and exhaust
bearings are shown. The vertical probes exhibited identical characteristics.
B
J
946 Rpm
785 Rpm
Exh Cool
J
B
1
Slow Speed at 300 Rpm
Fig. 94 Variation Of 1X
Vibration Amplitudes During Multiple Heating and
Cooling Cycles
Gov Hot
Gov Cool
B
J
0
1
1,508 Rpm
Exh Hot
1,203 Rpm
980 Rpm
J
B
J
740 Rpm
702 Rpm
4
654 Rpm
1X Shaft Vibration (Mils,p-p)
J
8
Consecutive Run Number
Within Fig. 9-4, the 1X vibration amplitudes at the governor journal are
depicted by the solid lines and circular plotting symbols. The exhaust end amplitudes are defined by dotted lines and square plotting symbols. Data points
acquired at the end of a cold run at 300 RPM are identified by the open plotting
symbols. The 1X vibration amplitudes measured at the end of a hot run are represented by the solid symbols. For each of the hot runs, the rotational speed at
the end of the run is listed for each pair of hot data points. From this summary
diagram, it is apparent that the maximum attainable speed during each hot run
successively increased from run to run. In addition, the hot vibration amplitudes
across the turbine tracked up and down in unison.
A significant portion of this plot is noted in the lower right hand corner.
Within this region, the measured amplitudes at the conclusion of the seventh
Bent or Bowed Shaft
405
run show a close agreement between hot and cold readings. Continued operation
at 1,508 RPM resulted in the convergence of hot and cold amplitudes into a common value. Under this condition, the 1X shaft motion at the governor journal
was 0.2 Mils,p-p, and the exhaust end converged to a 1X amplitude of 1.0 Mil,p-p.
The difference in vibration magnitudes between ends of the turbine is primarily
attributed to the chrome overlay sensed by the proximity probes on the turbine
exhaust shaft.
These vector amplitudes should be supplemented by the phase data to gain
a better appreciation of the bow characteristics and subsidence. Fig. 9-5 summarizes the running speed phase angle from the same horizontal probes during
each of the cold to hot cycles. Note that the cold phase angles are divergent. However, as the shaft warms up during the first five runs, the phase angles across the
turbine snap together as the bow becomes active. During the last three runs, the
hot phase angles become increasingly coincident with the cold values. At the conclusion of the seven cold to hot pseudo-stress relieving runs, the 1X vector amplitudes and phase angles remained constant. That is, between the cold condition at
300 RPM, and the warm operation at 1,500 RPM, the 1X vectors have not
changed. It is concluded that the application of the cold to hot runs allowed a
relaxation of the internal rotor stress, and the shaft bow has been relieved.
B
J
300
1
946
J
B
260
280
Fig. 95 Variation Of 1X
Vibration Phase Angles
During Multiple Heating
and Cooling Cycles
Exh Hot
Exh Cool
Gov Hot
Gov Cool
1,508 Rpm
1,203 Rpm
240
J
B
980 Rpm
220
J
B
785 Rpm
200
702 Rpm
180
654 Rpm
1X Phase Angle (Degrees)
160
740 Rpm
140
Consecutive Run Number
This procedure proved to be quite effective to prepare the rotor for the overspeed trip runs. Following this work, the turbine was coupled to the load compressor, and two days later the machinery train placed on slow roll. It was noted
that some of the residual bow activity was reappearing following two days of
non-rotation. Four more pseudo stress relieving runs were performed, and the
proper turbine slow behavior was re-established. The coupled startup was quite
smooth, and the machinery train displayed normal transient vibration characteristics. This bow eliminating procedure is still in use on this turbine, and it is
successfully applied after every extended outage on this unit.
406
Chapter-9
ECCENTRICITY
Eccentricity of one machine part with respect to another represents a less
common category of rotational speed excitations. Normally, shafts and most rotor
elements are ground on-centers. The material center is thereby concentric with
the initial center of rotation, and eccentricity is generally not a problem. However, there are occasions when a machine part is bored off-center. Although the
majority of the rotating assembly may be straight and concentric, the presence of
an eccentric element can impose a significant rotational speed force.
If the eccentric element is a minor part of the rotor assembly, the resultant
1X forces may be insignificant compared to the other active synchronous forces.
However, if the eccentric element represents a substantial portion of the rotating
assembly, or if it is located at a modally sensitive location (e.g., the coupling hub),
then the eccentricity may be a problem. The actual forces associated with an
eccentric element may be determined from the following equation (9-5):
2
RPM
F ecc = Weight element Eccentricity ---------------
5, 930
where:
Fecc
Weightelement
Eccentricity
RPM
=
=
=
=
(9-5)
Radial Force Due to Eccentricity of a Mechanical Element (Pounds)

Weight of the Eccentric Mechanical Element (Pounds)
Radial Eccentricity of Machine Element (Mils)
Shaft Rotational Speed (Revolutions / Minute)
Large machine elements or high rotational speeds are the most susceptible
to high forces due to an eccentric element. In many respects, an eccentric element appears similar to a shaft bow at low rotational speeds. Both mechanisms
provide large shaft displacement amplitudes at slow speeds. However, the forces
from a bowed rotor may remain constant at all speeds in accordance with equation (9-4). The radial forces from an eccentric element will vary with the speed
squared as described by expression (9-5). Naturally this all becomes much more
complicated when machines with flexible rotors and multiple mode shapes are
discussed. In all cases, eccentric machine elements on a rotor should be avoided,
and one source of potential synchronous excitation removed from consideration.
From a detection standpoint, shaft bows and eccentric elements can be
determined in the shop with accurate runout checks as described by John East3.
Once the machine is assembled, runouts can be detected at low speeds with relative shaft sensing proximity probes. Casing velocity coils and accelerometers will
probably not detect either mechanism at slow roll speeds. However, the casing
vibration transducers will pick-up the influence of a bow or an eccentricity at
higher speeds when the radial forces are significant. For instance, case history
21 considers a situation where a pinion coupling hub was bored off-center, and
the resultant eccentricity had a considerable influence upon the machinery.
3 John R. East, Turbomachinery Balancing Considerations, Proceedings of the Twentieth Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University, College Station,
Texas (September 1991), pp. 209-214.
Eccentricity
407
Case History 21: Seven Element Gear Box Coupling Bore

The machinery train in question consists of a lime drying kiln driven by a
variable speed synchronous motor through a seven element speed reducing gear
box. The drive portion of this train is depicted in Fig. 9-6. The kiln dryer itself
consisted of a long cylindrical tube that is mounted on rollers, and it is inclined
with respect to grade. Wet lime is loaded into the top end of the kiln, and dry
lime is extracted from the lowest elevation. This massive cylindrical kiln rotates
slowly, with a maximum speed of 1.8 revolutions per minute.
Output to Ring Gear
on Kiln OD
1.8 Rpm Maximum
Last Bull Gear
High Casing Vibration

on Motor IB
Fig. 96 Machinery
Arrangement For Large
Kiln Drive
Synchronous
Motor
3 - 50 to 500 HP
120 to 1,200 Rpm
Seven
Element
Speed
Reducing
Gear Box
First Pinion
The kiln drive system is located on an elevated platform. Initial operation

revealed unacceptable vibration amplitudes on the platform, and various structural modifications were implemented. Following these support improvements,
the drive train continued to exhibit uncomfortable vibration amplitudes. The
machinery was not equipped with any vibration monitoring instrumentation,
and initial readings were limited to casing measurements. The large seven element gear box exhibited low vibration amplitudes, and the problem appeared to
be confined to the variable speed motor. The outboard motor bearing displayed
casing vibration levels of 3.0 to 4.0 Mils,p-p. At the inboard, or coupling end motor
bearing, the unfiltered horizontal motion varied between 4.0 to 5.0 Mils,p-p.
It was clear that the kiln and the gear box are massive structures, and any
casing vibration would be substantially suppressed by their respective steel
structures. Conversely, the motor driver consisted of a fairly light frame, and it
was susceptible to excitation from a variety of sources. A complete survey was
performed with triaxial casing velocity measurements at each main bearing,
plus various locations on the support structure. It was quite evident that the primary excitation frequency occurred at motor running speed. Axial vibration of
408
Chapter-9
the motor, gear box, and structure was minimal at all locations. The major vibration appeared to be running speed motion of the motor in a radial direction. The
casing orbits presented in Fig. 9-7 depict the unfiltered and 1X filtered behavior
at the motor inboard bearing housing. Operating at an average speed of 1,190
RPM, the motor coupling end bearing housing exhibited a forward and elliptical
orbit at motor running frequency. As noted in the unfiltered plots, the maximum
horizontal motor casing vibration approached 4.0 Mils,p-p. Simultaneously, the
companion casing orbit on the pinion input bearing housing (not shown) revealed
casing displacement amplitudes of less than 0.5 Mils,p-p.
Fig. 97 Motor Bearing Housing Casing

Vibration At Full Operating Speed
Fig. 98 Pinion Relative Shaft Vibration At

Full Operating Speed
Further investigation, and visual inspection revealed excessive vibration of

the coupling assembly. Since the normal threshold for visual observation of
vibrating surfaces is in the vicinity of 10 Mils,p-p, it was considered significant
that the coupling could be vibrating twice as much as the motor bearing housing.
This observation prompted further investigation into the absolute motion of
the coupling. To implement this measurement, a separate stand was constructed
from steel plate and angle iron to support a pair of X-Y proximity probes on
either side of the coupling. One pair of probes observed the 8.25 inch diameter
coupling hub on the motor shaft. The other pair of proximity probes were positioned over the 9.75 inch diameter pinion coupling hub. All four proximity probes
were supplemented by velocity coils to allow for potential correction of the shaft
vibration signals for motion of the probe supporting structure. As it turned out,
this correction was not required, and the direct probe signals from the pinion
coupling hub are presented in Fig. 9-8.
This data reveals a distinctive forward circular pattern to the pinion coupling hub displacement. Since the hub outer diameter was not machined or pre-
Eccentricity
409
pared for proximity probes the hub surface imperfections were visible, and
overall amplitudes in the vicinity of 12.0 Mils,p-p were documented. The hub signal filtered at the rotational speed (1X) of 1,190 RPM exhibited an average circular amplitude of 10.0 Mils,p-p. It should also be mentioned that the coupling hub
on the motor side displayed circular amplitudes of nominally 5.0 Mils,p-p. This
steady state data at full speed suggested that the high vibration might be originating at the pinion instead of the motor. In support of this preliminary conclusion, it was understood that uncoupled, and unloaded, motor vibration was quite
low. That does not necessarily give the motor a clean bill of health, since many
motor problems only appear under load. Nevertheless, it did suggest that perhaps the pinion might be the culprit, and the light motor might be just responding to a forced vibration condition.
Additional perspective on this problem was gained by the acquisition and
analysis of variable speed information. Specifically, vibration data was recorded
during a shutdown of the kiln, and the Bode plot shown in Fig. 9-9 obtained from
the X-Y proximity probes positioned over the pinion coupling hub. Note that the
synchronous 1X amplitudes and phase angles remained essentially constant
from the top speed of 1,190 RPM to the minimum sample point of 186 RPM. This
type of behavior is certainly representative of an eccentric mechanical element.
In this case, the pinion coupling hub was the primary suspect.
Another perspective of the kiln shutdown was gained from Fig. 9-10, that
documents a time history plot of the coastdown. In this diagram, the 1X amplitude and phase are plotted against time from 0 to 60 seconds. In addition, the
Fig. 99 Bode Plot Of Pinion Vertical Hub

Vibration During A Typical Coastdown
Fig. 910 Time History Coastdown From

Vertical Prox Probe On Pinion Coupling
410
Chapter-9
rotative speed is included at the top of the same plot. The various humps in this
summary plot are due to the fact that the main kiln cylinder carries a tremendous amount of inertia from the rotating kiln, plus the internal lime. During a
routine coastdown of this machine, the kiln slows down from full operating speed
of 1,190 RPM, and comes to a stop in approximately 6 seconds. The kiln tube
(with the moving lime) then begins a reverse rotation, and drives the gear box
and motor in a reverse rotation. The input pinion and motor reach a peak speed
of 800 RPM before the train starts to slow down. As shown in Fig. 9-10, the unit
experiences two more forward, and one more reverse rotation sequence before
the train comes to a final stop.
Note that the rotational speed amplitude peaks at about 8.0 Mils,p-p irrespective of a forward or a reverse cycle. When the unit is rotating in a forward
direction, the vertical probe phase angle is approximately 250. During reverse
rotation, the 1X phase angle is in the vicinity of 80. Thus, a nominal 170 reversal in the high spot occurs as the pinion hub rotates in a forward or a reverse
direction. This is close enough to 180 to conclude that there was a complete
reversal of the phase relationship between forward and reverse rotation. This
documented behavior also helps to substantiate the hypothesis of an eccentric
coupling hub.
Additional data at various loads provided no other useful information, and
it was finally concluded that the coupling hub was bored off center. The physical
configuration of the pinion extension did not allow simultaneous dial indicator
measurements of the pinion hub versus the pinion shaft. However, when the coupling hub was removed, it was determined that the shaft bore was indeed off center by approximately 8 to 10 Mils. Naturally the coupling supplier was slightly
embarrassed, and they provided a concentrically bored coupling assembly in a
short period of time. The installation of this correctly bored coupling half on the
pinion shaft solved the problem.
SHAFT PRELOADS
Another category of potential malfunctions that are generally applicable to
all rotating machinery is the topic of shaft preloads. The presence of various
types of unidirectional forces acting upon the rotating mechanical system is a
normal and expected characteristic of machinery. Just as residual unbalance,
rotor bows, and component eccentricity are inherent with the assembly of rotating elements, the presence of shaft preloads are an unavoidable part of assembled mechanical equipment.
From an initial categorization standpoint, shaft preloads may be divided
into two fundamental groups. The first group would address the preloads that
originate within the machinery. These internal preloads may be due to any or all
of the following common mechanisms:
Shaft Preloads
411
Gravitational Preloads
Bearing Preloads
Internal Misalignment Preloads
Gear Mesh Forces
Fluid Preloads
Gravitational preloads on horizontal rotors are responsible for the rotor

bow or sag that was discussed earlier in this chapter. Again, this is part of the
normal and unavoidable characteristics of process machinery. This is generally
not a design correctable problem. It is a physical phenomena that must always be
considered, and dealt with on a machine by machine basis.
Bearing preloads represent one of the machinery design considerations. As
discussed in chapter 4, bearing preloads are typically expressed as a non-dimensional number between 0 and 1. A bearing preload of 0 indicates no bearing load
upon the shaft. Conversely, a bearing preload of 1 indicates a shaft to bearing
line contact (i.e., maximum preload of 1). The computation of bearing preload is
based upon the difference in curvature between the shaft and the individual
bearing pad. As expressed by equation (4-6), bearing preload is determined by:
Cb
Preload = 1 -------
C p
(9-6)
Where Cb is the bearing clearance which is equal to the bearing radius

minus the journal radius. The pad clearance Cp is equal to the pad radius minus
the journal radius. On segmented bearings it is common to find that the pad
radius is greater than the bearing radius. Thus, the pad clearance Cp is greater
than bearing clearance Cb, and their ratio is less than one. From equation (9-6)
the bearing preload must therefore be less than 1. Typically, the bearing designers will employ preloads that vary between 0.1 and 0.4.
Clearance reduction at the center of the pad forces the oil to converge into
an oil wedge. The operational characteristics of bearing stiffness, damping, and
eccentricity position will vary with the preload, and the actual bearing configuration. Obviously, the characteristics of a fixed pad bearing, such as an elliptical
bearing, will be different from a five shoe tilting pad bearing. However, in either
case an oil wedge will be developed, and that oil wedge will have an associated
pressure profile (Fig. 9-20). The direct action of this pressure profile upon the
journal is a preload force. Again, this is part of the normal behavior of the rotating machine, but the amount of preload, and the associated journal force is
adjustable (within limits) by varying the pad radius and bearing geometry.
The third type of internal machine preload is attributed to internal misalignment of elements. This can vary from offset or cocked seals or bearings, to
distorted diaphragms or stators, to a variety of rub situations. It is virtually
impossible to quantify all of the potential combinations of misaligned internal
machine elements. However, the common characteristic that they all share is the
generation of a load or force against the shaft. Some of these preloads may
412
Chapter-9
relieve themselves during normal operation. For instance, a laby seal rub might
occur during initial startup on a machine, and the expanded shaft to seal clearances may never rub again. Other preloads such as a distorted stator will generally remain constant, and will continue to provide a force upon the rotor.
The fourth type of internal preload is associated with gear mesh forces.
These are significant loads that must always be considered. To demonstrate the
magnitude of gear contact forces, the values calculated in case history 24 are
repeated. These loads are for a simple pinion - bull gear arrangement. The forces
were computed for a transmitted load of 4,000 horsepower, a pinion speed of
5,900 RPM, and a bull gear speed of 1,920 RPM. The significant element weights
and forces for this gear box are summarized as follows:
Pinion Weight ......................... 220 Pounds
Bull Gear Weight ................. 1,630 Pounds
Separation Force.................. 4,550 Pounds
Tangential Force ................ 10,880 Pounds
From this summary, it is clear that the major forces within a gear box are
the gear contact forces. The magnitudes of the separation and the tangential
forces place the gear weights into the role of a secondary influence. These gear
forces are used in the development of gear box bearing coefficients, and in the
initial estimation of the journal running position. It is important to consider the
journal operating position during alignment of the gear box, and recognize that
the bull gear and pinion bearings are subjected to significant radial preloads
from the gear forces.
The fifth type of fluid preloads is applicable to many types of rotating
machines. For instance, the unbalanced radial force in a volute pump is an obvious case of fluid forces acting directly upon a rotor. A less obvious example of
fluid forces would be the behavior of a multistage and multilevel turbine during
startup. It has been documented that partial steam admission to the first stage
nozzles may cause a lifting force on the rotor when the first nozzle segment is
located in the bottom half of the turbine casing. This radial force may be sufficient to lift the rotor, and allow the governor end bearing to go unstable. Hence,
the vertical shaft preload would work against the stabilizing gravitational force
to drive the machine into another type of malfunction. In all cases, the machinery diagnostician must be aware of these types of physical interactions, and must
strive to understand and address the fundamental forces behind the observed
vibratory motion.
The second major category of shaft preloads considers the array of potential
external preloads or forces. For example, the following short list identifies some
of the common external shaft or machinery preloads:
Coupling Misalignment
Locked Coupling
Thermal or External Forces
Shaft Preloads
413
The problem of coupling misalignment is common to most types of rotating machinery. Fortunately, the machinery community has devoted considerable
time and effort to develop solutions and techniques for execution of correct shaft
alignment. In fact, many operating facilities have applied these tools and techniques to develop very successful machinery alignment programs. In these operating plants, misalignment has ceased to be a problem. Nevertheless, when
misalignment between machines is present, the shaft preload forces may be substantial, and may result in premature mechanical failure.
A related mechanism to shaft misalignment is the problem of a locked
coupling. Primarily this occurs on oil lubricated gear type couplings. Since
these couplings are designed as flexible members with a tolerance to misalignment, the abnormal condition of locked coupling teeth will violate the intended
behavior of the flexible design. A locked coupling may behave similarly to a misaligned coupling. In may cases the coupling is locked due to excessive misalignment. In other cases, the locked coupling may develop during operation due to
the accumulation of sludge between the teeth. In either case, the resultant forces
on both machinery shafts are unwelcome preloads that may damage machinery
components on either side of the locked coupling.
The third category of external preloads is associated with the presence of
any number of potential external forces or moments on the machinery. This kind
of preload could be due to baseplate strain imposed by a degrading grout or foundation. External preloads could also arise from piping strain upon the machine.
For instance, the case history 49 describes the effect of a piping moment upon a
compressor, and the coupled turbine driver. External preloads may influence the
coupling alignment, or they may distort bearing housings, casings, or other
mechanical attachments.
In essence, shaft preloads are a normal part of rotating equipment that
must be addressed. It is useful to recognize that preloads have different levels of
severity. For instance, some preloads such as gravity or fluid based forces may be
classified as soft preloads that are generally non-destructive. Other preloads,
such as misalignment or gear contact forces, may be considered as hard preloads
that can be damaging to the machinery. A third severity classification for shaft
preloads would be the destabilizing variety. This type of preload may oppose the
normal rotor or bearing forces, and it may act to destabilize the rotor. The severity of these destabilizing preloads may also vary from soft to hard, depending on
the final influence upon the process machinery.
Preload detection is predicated upon the recognition of abnormalities in
radial vibration. The following three characteristics are used to identify the presence of radial shaft preloads:
Normal Orbital Motion
Abnormal Shaft Centerline Position
Abnormal Shaft versus Casing Motion
Abnormal orbital motion is demonstrated in Fig. 9-11, that depicts an array

of four shaft orbits with different levels of radial preloads. The orbit in Sketch A
414
Fig. 911 Changes In

Shaft Orbits With Increasing Levels of Radial Shaft
Preloads
Chapter-9
Sketch A - No Preload
Sketch C - Moderate Preload
Sketch B - Slight Preload
Sketch D - Heavy Preload
displays a normal forward circular pattern with no shaft preload. Application of

a slight preload to the shaft in Sketch B forces an elliptical shape to the shaft
motion. The moderate preload of Sketch C drives the orbit into a banana shape.
Finally, the heavy preload case shown in Sketch D results in a Figure 8 orbit.
This type of analysis is quite appropriate to machines with fluid film bearings,
and shaft sensing proximity probes. In the vast majority of the cases, the direction of the preload is perpendicular to the major axis, and the severity of the preload is dependent on the ellipticity of the orbit.
As a cautionary note, the machinery diagnostician should not confuse the
preloaded orbit with the normally elliptical orbit. Machines with significant differences between vertical and horizontal bearing stiffness will naturally display
an elliptical orbit. This type of orbit is strictly a function of the bearing geometry,
and must be considered as normal and proper behavior.
For example, Fig. 9-12 displays a diagram of an elliptical bearing with a
counterclockwise rotating shaft. The normal elliptical orbit is noted in a proper
running position in the lower right hand quadrant. This same diagram also
Vertical
Clearance
Vertical Probe
Fig. 912 Normal Versus

Abnormal Shaft Centerline
Position In An Elliptical
Bearing Assembly
Horizontal Probe
45
45
Normal
Orbit
Position
Abnormal
Orbit
Position
Horizontal
Clearance
CCW
Rotation
Shaft Preloads
415
depicts another elliptical orbit residing in the lower left hand quadrant. The second orbit is obviously in the wrong place for a CCW rotating machine. This
improper radial position represents the second method of detection of shaft preloads. That is, the calculated journal centerline position should be in the proper
location within the clearance of the bearing. If the shaft centerline position
resides at an abnormal location, the possibility of a damaged bearing, or the
presence of a radial preload should be suspected.
Three cautionary notes should be added to this type of evaluation. First, the
diagnostician must know the proper running position of the journal within the
specific bearing before attempting to pass judgment on any field data. For
instance, a five shoe tilting pad bearing will display a vertical attitude angle, and
normal position for this type of bearing is considerably different from the previously discussed elliptical bearing. If the eccentricity position and the attitude
angle are not known, then the diagnostician should consider an FEA analysis of
the specific bearing configuration as discussed in chapter 4.
Second, the machinery diagnostician must be working with a properly calibrated proximity probe system to measure the true running position of the journal within the bearing. This includes accurate probe scale factors plus a correct
probe orientation diagram. In some instances, it may be necessary to install four
radial proximity probes at 90 increments. In this application, the diametrically
opposed probes are summed to determine an average shaft position in each
orthogonal direction. This is more work, but it does enhance accuracy of the
radial position measurement.
Third, the initial proximity probe DC gap voltages must be accurately
known to allow a confident calculation of the shaft centerline position. It is difficult to generalize on the precise condition to obtain the at stop gap voltages. Normally, this data is obtained prior to startup with warm oil circulating. To be safe,
it is recommended that DC voltages be tracked with a computer-based system
that will identify time, speed, and other useful information such as oil supply
temperature, ambient temperature, etc.
Finally, on machines with accessible bearing housings, it is desirable to
acquire X-Y casing vibration response measurements. The casing probes should
be placed in the same angular orientation as the shaft sensing proximity probes.
In addition, the casing probes should be located as close as possible to the mounting point of the proximity probes. The casing data must be integrated to displacement, and the 1X synchronous vectors compared directly against the runout
compensated shaft displacement 1X vectors. Under normal conditions, the casing
motion should be smaller than the shaft vibration, and the casing phase angles
should lag behind the shaft vibration angles.
For a machine with a radial preload, and a compliant support, the shaft
vibration may be suppressed. In this condition, the normal shaft vibration within
the bearing is transmitted to the casing and surrounding structure. From a measurement standpoint, the casing 1X vibration amplitudes may exceed the shaft
motion, and the shaft to casing phase relationship may appear abnormal. This
final criteria is not a totally conclusive test, but it does provide additional insight
into the mechanics of the machinery.
416
Chapter-9
RESONANT RESPONSE
Machines and structures all contain natural frequencies that are essentially a function of stiffness and mass. As described in previous chapters, the fundamental relationship may be described by the following expression:
Stiffness
Natural Frequency -------------------------Mass
Recall that this expression was developed for a simple spring-mass system,
and it basically identified the lowest order resonant frequency. For more complex
mechanical systems an entire family of resonant responses must be addressed.
For example, consider a turbine compressor set mounted on a mezzanine structure, and connected with a flexible coupling. The potential array of anticipated
natural, or resonant frequencies are summarized as follows:
Lateral Critical Speeds
Turbine Translational (1st)
Turbine Pivotal (2nd)
Turbine Bending (3rd)
Compressor Translational (1st)
Compressor Pivotal (2nd)
Compressor Bending (3rd)
Torsional Critical Speeds
Turbine (1st)
Turbine (2nd)
Compressor (1st)
Compressor (2nd)
Rotor Element Resonances
Coupling Natural Axial
Coupling Lateral
Turbine Blades
Compressor Impellers
Acoustic Resonances
External Piping Systems (including stubs and branches)
Internal Passages Within Casings
Structural Resonances
Piping Systems
Structural Steel Systems
Machinery Pedestals
Baseplate, Foundation, or Ground Support System
This list of potential resonant frequencies can be intimidating. The range of

natural frequencies may vary from 60 CPM (1 Hz) for the foundation and support systems, to 1,800,000 CPM (30,000 Hz) for turbine blade natural frequen-
Resonant Response
417
cies. With this extended range of natural frequencies it is somewhat amazing

that process machinery can be designed to operate without violating many of
these system natural frequencies. One of the tools used by machinery designers
is the Campbell diagram for plotting natural frequencies versus excitations. This
was initially introduced in chapter 4. For reference purposes, the Campbell plot
from Fig. 5-4 is reproduced in Fig. 9-13. Please note that the natural frequencies
(rotor lateral modes) are shown as horizontal lines off the vertical axis, and the
various machine excitations are depicted as the slanted lines. These slanted lines
are the 1X, 2X, 3X, etc. multiples of rotational speed that the machine will produce as speed increases from slow roll to the normal operating point.
12,000
1X
2X
2X
5X
5X
4th
8,000
Excitations
10,000
1X
10X
6,000
Rotor Bending - 3rd

10X
4,000
5th
4th
3rd
Pivotal - 2nd
2,000
2nd
Translational - 1st
1st
Natural Frequencies
Natural Frequency (Cycles/Minute)
5th
0
10,000
20,000
30,000
40,000
50,000
Excitation Frequency (Cycles/Minute)
Fig. 913 Campbell Diagram Describing Interference Between Several Forced Machinery
Excitations And A Partial Group of Turbine Natural Lateral Resonances
An intersection between curves identifies points of potential resonant

response. That is, a natural frequency exists, and an excitation source is present
at the same frequency. This coincidence results in a stimulation or activation of
the resonance. The Campbell plot is an interference diagram that is commonly
developed during the design stages of a new machinery train. The actual number
of system natural frequencies may vary from 20 to 50 or more. Furthermore, the
actual number of machinery excitations may also be significant, and 10 to 30 significant excitations are not unusual.
The general characteristics of each type of resonance must always be considered. For example, a structural resonance has minimal damping, and it will
exhibit a sharp response with a high Q amplification factor. Resonances of this
type have a narrow bandwidth, and the excitation frequency must be precisely
equal to the resonance for any action to occur. On the other hand, rotor lateral
resonances include damping from the bearings and seals, and they are active
over an appreciable frequency range (bandwidth). These type of rotor resonances
may be activated by an excitation force that falls anywhere within the bandwidth of the damped rotor resonance. The resonant frequency may also change
418
Chapter-9
with machine speed. For instance, bearing stiffness and damping characteristics
vary with rotational speed. In most cases, the bearing parameters that apply as
the machine passes through a critical speed are different from the normal operating condition. Hence, the same resonance may appear at different frequencies
due to the influence of other associated machine elements (e.g., bearings).
Additionally, the exciting mechanism does not have to occur at rotating
speed, or any even order harmonic. The excitation may occur at a subsynchronous (below running speed), or a supersynchronous (above running speed) frequency. For instance, a machine may startup with a locked 43% oil whirl
instability. Presumably, the frequency of this instability would increase in direct
proportion to the rotational speed. If the machine operated above twice the critical speed, the 43% whirl would eventually coincide with the rotor balance resonance during startup. This coincidence between the rotor resonance and the oil
whirl would probably result in a re-excitation of the critical. The oil whirl would
evolve into an oil whip, with potentially damaging implications to the machinery.
Due to the severity of this problem, additional discussion and explanation will be
presented later in this chapter.
As discussed throughout this text, machinery systems exhibit a wide variety of natural resonances. In the vast majority of cases, these resonances remain
dormant, and their presence goes undetected. However, when an excitation does
appear, or when the mechanical characteristics of the system undergo a change
(due to failure or attrition); the idle resonance may become adversely excited.
The solution to these occurrences typically resides in identification of the
changes in physical machine parameters. With respect to resonance problems,
the machinery diagnostician should always examine the mechanical system for
evidence of variation of mass, stiffness, or the application of a new force. When
the variant is discovered, the solution is close at hand.
Finally, there are groups of natural frequencies that may be discounted
from a design standpoint, but they may have to be examined during a detailed
machinery analysis. For instance, torsional resonances on a turbine driven compressor would probably not be a major cause for concern. However, if mechanical
failures indicated the existence of twisting forces, torsional vibration should be
considered. Torsional modes have low damping (high Qs) and they have caused
many failures on large turbines.
Conversely, torsional resonances on a reciprocating engine would be of significant interest during initial design, and acceptance testing. However, if the
main bearings displayed babbitt failures, the lateral vibration should be evaluated. Thus, a group of potential resonances and/or excitations should not be eliminated from possibility just because they may not apply. The other extreme of
performing detailed examinations on mechanisms that have no possible relationship to the immediate problem should also be avoided. As always, the machinery
diagnostician must exercise good engineering judgment when selecting or eliminating potential resonances for a machinery problem.
Resonant Response
419
Case History 22: Re-Excitation of Compressor Resonance

Many refrigeration compressors in operating facilities such as ammonia or
ethylene plants are configured with multiple side streams plus one or more levels of extraction. These refrigeration compressor systems are designed to maintain specific flow rates and refrigerant temperatures. In many cases, the process
requirements are quite stringent, and the compressor is optimized to operate
within a very limited performance envelope. These requirements pose some challenging system and machinery design considerations that often results in
machines that are difficult to operate at non-design conditions.
For instance, consider a large six stage propylene refrigeration compressor
that draws in excess of 30,000 horsepower at design operating conditions. The
rotor weighs 7,100 pounds, and the normal operating speed for 100% plant load
is 3,860 RPM. Obviously this speed will be subjected to control system adjustment dependent on the actual number of operating cracking furnaces, and the
associated refrigerant load. A typical startup of this machine is described by the
Bode plot shown in Fig. 9-14, plus the polar plot shown in Fig. 9-15. This transient data was acquired from the horizontal proximity probe at the coupling end
of the compressor. On this machine, the horizontal vibration is normally larger
than the vertical motion, and the discharge bearing vibration is slightly higher
than the outboard suction end bearing. Hence, the data presented in Figs. 9-14
and 9-15 represents the highest vibration amplitudes encountered during a normal machinery train startup.
This startup was quite smooth with a clearly defined compressor first critical speed at 2,020 RPM. The observed critical response range extended from
Fig. 914 Bode Plot Of Refrigeration

Compressor Startup To Minimum Governor
Fig. 915 Polar Plot Of Refrigeration Compressor Startup To Minimum Governor
420
Chapter-9
approximately 1,600 to 2,500 RPM. The peak response was somewhat sharp, but
that was attributed to the fast ramp programmed into the electronic governor.
Hence, this transient response was not indicative of any mechanical abnormality.
A complete set of steady state data was acquired twenty-four hours after
startup. The orbit and time domain plots for the discharge end bearing are presented in Fig. 9-16 at an average speed of 3,680 RPM. The compressor orbits are
forward, with an elliptical pattern at the discharge, and a circular motion at the
suction (not shown). In addition, a subsynchronous instability was visible at the
discharge bearing. The dominant direction of this low frequency motion was horizontal, and the frequency oscillated between 1,980 and 2,100 CPM. Extended
observation of this subsynchronous component revealed that 2,060 CPM was the
major frequency, and peak amplitudes reached 1.5 Mils,p-p at the discharge horizontal. Time averaged behavior is displayed on the FFT plot shown in Fig. 9-17
with a defined peak at 2,060 CPM, and an average horizontal amplitude of 1.1
Mils,p-p. It should also be mentioned that subsynchronous vibration amplitudes
at the suction bearing generally remained below 0.25 Mils,p-p.
This 2,060 CPM frequency is recognized as the first critical speed of the
compressor rotor. As previously noted, the startup data displayed a translational
balance resonance (first critical) at 2,020 RPM. However, as stiffness characteristics change with speed, journal eccentricity, and temperature, the startup first
critical is generally different from the critical response observed at full speed.
Hence, the 2,060 CPM component is considered to be a re-excitation of the compressor first critical speed. This phenomena has occurred for over twenty years,
and previous studies have correlated the compressor instability with the extrac-
Fig. 916 Refrigeration Compressor

Inboard Bearing Orbit And Time Base Plots
Following Normal Train Startup

Inboard Bearing Spectrum Plots Following
Normal Train Startup
Resonant Response
421
tion sidestream. Historically, startup of the propylene refrigeration system produces flow and pressure fluctuations in the extraction line. These fluid based
excitations are transmitted to the discharge end of the compressor rotor via the
horizontal extraction nozzle. The observed shaft motion is primarily horizontal,
and the largest excitation occurs at the coupling (discharge) end of the rotor.
Fluid variations in the extraction stream provides a broadband excitation to the
compressor rotor, and the first critical speed is generally excited. In addition, the
turbine first critical is also driven due to the close proximity between the compressor and turbine balance resonant speeds. Although the dominant subsynchronous vibration occurs at the compressor discharge bearing, the other
machinery train bearings display low level excitations. However, as the refrigeration systems become lined out, the extraction stream flow instability diminishes, and excitation of the critical speeds of both rotors normally disappears.
In support of this explanation, the vibration response data acquired with an
average speed of 3,860 RPM, and a full process is shown in Figs. 9-18 and 9-19.
Although 1X running speed vectors have experienced minor changes, the subsynchronous motion no longer exists. Examination of the turbine data reveals a similar absence of motion at the first critical. Hence, the documented re-excitation of
the compressor first critical speed under unstable extraction flows was eliminated by establishing normal loading of the propylene refrigeration system.
Again, this behavior is totally consistent with the historical behavior of this
machinery train.

Inboard Bearing Orbit And Time Base Plots
Under Normal Process Load

Inboard Bearing Spectrum Plots Under
Normal Process Load
422
Chapter-9
MACHINERY STABILITY
Throughout this text, the attributes of well balanced and properly aligned
machines operating with concentric rotor elements have been repeatedly
endorsed. Reductions in shaft preloads are generally associated with reduced
forces and extended machinery life. In fact, many machinery problems that
appear to be extraordinarily complex are often beat into submission merely by
corrections to the basic mechanical parameters of balance, alignment, and element concentricity. There is an added dividend provided by smooth running
machines in the area of incrementally improved efficiency. In essence, more of
the input energy goes into productive work instead of being wasted on mechanical abnormalities.
However, the uninitiated may be surprised to find that in some situations,
these fundamental corrections may result in an inoperable machine. There are
many documented instances of properly executed balance or alignment corrections that have resulted in significantly higher vibration response amplitudes. In
many of these cases, an examination of the vibratory characteristics has
unveiled the presence of a new frequency component. Often this new vibration
component occurs at frequencies below rotating speed, and this subsynchronous
motion is often associated with machinery instability. Although this general definition of instability is not rigorously correct, it is still used throughout most
industrial locations.
The very nature of centrifugal machinery provides the fundamental mechanism for this type of behavior. In all cases, it must be recognized that centrifugal
machines consist of rotating cylinders or disks confined within stationary cylinders. If clearances between cylinders are large, there is no possibility for interaction between stationary and rotating parts. For example, a 6 inch diameter shaft
rotating within a 20 inch diameter annulus will function in the same manner as
it would in free space. However, as clearances decrease, there is increased opportunity for interaction between elements. For instance, if the 6 inch diameter
shaft now rotates inside a 6.008 inch diameter bearing; interaction between
cylindrical elements now exists across the contained fluid. The fluid might be
steam, a process gas, a process liquid, oil in a seal, or oil contained within a bearing. The general type of behavior for a cylinder rotating inside of a stationary cylinder is depicted in Fig. 9-20.
From this diagram, it is anticipated that the rotating element establishes a
minimum running clearance to the stationary cylinder. For an oil film bearing,
this clearance would normally be identified as the minimum oil film. The active
forces across the minimum oil film include the fluid radial force, plus a tangential component. In this simple example, these two forces should be vectorially
equal to the shaft load. Thus, the oil film forces are in equilibrium with the shaft
load. If Fig. 9-20 was representative of a journal and bearing in a horizontal
machine, the shaft load would primarily consist of the shaft weight. Furthermore, the described system would exhibit a minimum oil film in the lower right
hand quadrant of the bearing. Due to the counterclockwise rotation, it is intuitive that the shaft would climb the lower right hand side of the bearing. Addi-
Machinery Stability
423
Stationary
Rotating
Journal
Rotating
Fluid
Tangential
Force
Velocity
Profile at
Minimum
Oil Film
Stationary Bearing Housing
Shaft
Load
Fluid Radial Force
Fig. 920 Typical Radial Forces In A Fluid

Film Bearing At The Support Point
Fig. 921 Typical Bearing Oil Velocity Profile At The Point Of Minimum Oil Film
tional shaft loading would tend to drive the shaft further up the bearing wall.
Donald E. Bently referred to this action as the Newton Dogleg Law of Rotating
Machinery4 where he stated:
For every one of the Forward Circular Eigen (Self-Excited) Malfunction
Mechanisms of rotating machinery, from simple Oil Whirl and Oil Whip to the
complex Aerodynamic Mechanisms, the rotor system responds to radial input
force in dogleg styleThus, if you push a CCW rotating rotor system down
towards 6 oclock position, if that system has any Forward Circular Mechanism, it
responds by moving not simply to the 6 oclock position, but to some position
between 6 oclock and 3 oclock. This angle is called the Attitude Angle.
This action is similar to gyroscopic action, but obeys a different set of laws. It
is independent of the self-balancing laws, but gets interlocked with both, of course,
in the rotor system behavior
Clearly, a direct and intimate relationship exists between the rotating system, the fluid film, and the stationary system. It is also obvious that a force balance must exist at the bearings, and that the fluid tangential force strives to
move the shaft in a forward direction (i.e., with rotation). Any disruption of the
force balance within this pure circular system will allow the minimum oil film to
circumnavigate the bearing in a forward manner. The speed of progression is
dependent upon the oil film velocity as shown in Fig. 9-21. It is self-evident that
the journal rotational speed times the journal radius will provide the shaft surface velocity. This is the maximum oil film velocity. At the stationary bearing, the
oil film velocity must be zero due to the non-rotative nature of the bearing. In a
perfectly linear mechanical system, the average oil film velocity would be 50% of
4 Donald E. Bently, Attitude Angle And The Newton Dogleg Law Of Rotating Machinery
Bently Nevada Applications Note, (March 1977).
424
Chapter-9
the maximum velocity, or 50% of the rotational frequency. However, in a real

world mechanical system, the difference between journal and bearing surface
conditions, the bearing geometry, bearing clearances, and loads will result in an
average oil film velocity that is somewhat less than a simple 50% arithmetic
average. The velocity profile will not be a straight line relationship, and the average oil film velocity will be less than 50%. This manifests as a frequency that
appears at something less than 50% of running speed. In practice, ratios between
35% and 49% are commonly observed and documented.
This is the essence of a self-excited mechanism such as oil whirl. That is, the
physical geometry and loading of the mechanical system allows the establishment of a subsynchronous vibration component in oil lubricated bearing. This
motion is forward, circular, and typically appears between 35% and 49% of shaft
rotative frequency. The simplest case is known as oil whirl, and it is generally
detected in machines with plain sleeve bearings.
A visual dichotomy is always encountered when oil whirl orbits are
observed on an oscilloscope. Specifically, this is a forward circular mechanism,
but the Keyphasor dots seem to spin backwards against rotation. This gives the
appearance of a backwards motion, and a misinterpretation of the data is possible. In actuality, the backwards spinning Keyphasor dots are perfectly correct
for a forward subsynchronous vibratory component. In order to demonstrate this
behavior, consider Fig. 9-22 which describes several revolution of a shaft experiencing a forward subsynchronous whirl.
In this diagram of a counterclockwise rotating system, consider the first
Keyphasor dot at position A. Assume a negative going Keyphasor pulse, and
the expected blank-bright sequence indicates a CCW rotation. As the shaft
makes one complete revolution, the predominant subsynchronous component
has only completed about 45% of a full cycle (160). At this instant in time, the
second Key dot appears at position B. Note that the blank-bright sequence is
still consistent, and representative of forward CCW motion. As additional cycles
are completed, the Key dot continues to lag further and further behind. Thus,
K Dots
A
CCW
Shaft
Rotation
Fig. 922 Keyphasor Dot

Precession For Subsynchronous Excitation Occurring At
Less Than 50% Of Machine
Rotative Speed
F
D
K Dots Move CW
Machinery Stability
425
a visual observation on an oscilloscope will reveal the dots moving from A to C to

E. In the second group, the dots move from B to D to F. Therefore, the dots
appear to roll backwards in time, but the blank-bright sequence specifically identifies a forward orbit.
This characteristic can be put to good advantage when viewing live subsynchronous data in an orbital display on an oscilloscope. That is, when the dots
appear to move in a counter-rotation direction, the subsynchronous component
occurs at less than 50% of running speed. Conversely, when the dots appear to
move in the same direction as shaft rotation, the subsynchronous component
occurs at a frequency that is greater than 50% of rotative speed. The middle condition of a pair of fixed Keyphasor dots indicates a frequency that is locked onto
50% of running speed.
This is an extraordinarily important concept to remember and apply. It may
be extended to three fixed Keyphasor dots that identify a frequency component
at exactly one-third of running speed. Four fixed Key dots identify a frequency component at exactly one-fourth of running speed, etc. In many
instances, the orbital observation of fixed dots is faster, easier, and more accurate
than performing an FFT analysis of the data. Finally, the rate of Keyphasor dot
rotation is directly related to the frequency difference between the subsynchronous component and 50% of running speed. If the Key dots are moving very
slowly, the subsynchronous component is quite close to 50% of running speed.
Conversely, if the dots are rolling around the orbit at a rapid pace, the subsynchronous component is considerably removed from one half of running speed.
The vibration characteristics of a machine experiencing oil whirl are demonstrated in Fig. 9-23. Operating at 2,832 RPM, the orbit and time base data in
the top plot represents the unfiltered signals from a rotor kit with a diametrical
Fig. 923 Orbit & Time Base Of Oil Whirl
Fig. 924 Spectrum Analysis Of Oil Whirl
426
Chapter-9
bearing clearance of 18 Mils. The orbit-time-base data shown in the bottom of

Fig. 9-23 describes the low amplitude running speed (1X) vibratory behavior. For
precise frequency identification of the subsynchronous component, the same
information is viewed in the frequency domain in Fig. 9-24. From the FFT plots
the whirl frequency is 1,365 CPM which is equivalent to 48% of shaft rotative
speed. Again, the running speed motion is dwarfed by the violent oil whirl excitation at 48% of running speed.
Under the proper circumstances, oil whirl may turn into oil whip. If the
machine operates at a speed of twice the critical, the potential for whip exists. In
the previous example of a 48% whirl, the operating speed was about 2,832 RPM.
This speed is below the first critical. It should be noted that the bandwidth of the
critical speed range for this machine extends from 3,000 to 3,800 RPM. This
machine went into whirl at 1,600 RPM, and maintained a steady 48% of running
speed whirl during the initial speed ramp. This behavior is shown in Fig. 9-25
describing a cascade plot of vertical vibration spectra versus speed.
In this plot, the 48% Whirl tracked running speed until 6,200 RPM when
the Whirl frequency (6,200 RPM x 48% = 2,980 CPM) began to move into the critical speed range. As shown in the cascade plot, the whip frequency remained
locked into the critical speed range of 3,000 to 3,800 RPM as speed continued to
increase. Under this condition, the increased rotor speed had minimal influence
upon the whirl. That is, even at the top speed of 9,800 RPM, the oil whip was still
trapped within the natural rotor resonance range. Thus, a resonant response can
be obtained from a non-synchronous excitation.
Since this type of oil whip behavior involves the re-excitation of a major
rotor resonance, it is reasonable to expect that oil whip may be a potentially dan-

Revealing The Transition
From Oil Whirl To Oil Whip
Machinery Stability
427
gerous and destructive mechanism. In numerous field cases, machines have

operated successfully for many years with low levels of oil whirl. The whirl
existed, it was nondestructive, and it was tolerated. However, few machines have
successfully survived any type of extended operation with appreciable levels of
oil whip instability.
A closer examination of the orbital and time domain vibration of this oil
whip case is presented in the orbit and time base plots of Fig. 9-26. Note that the
unfiltered amplitudes are in the vicinity of 9.0 Mils,p-p, and the rotative speed
Fig. 926 Orbit Time Base Of Oil Whip
Fig. 927 Spectrum Analysis of Oil Whip
was 8,420 RPM. From the orbit plots it is clearly demonstrated that this is a forward and circular mechanism. The companion spectrum plot of the vertical and
horizontal probe signals is shown in Fig. 9-27. It reveals the precise oil whip frequency of 3,360 CPM. Under this combination of frequencies, the whip occurs at
40% of rotative speed. Also note that the whip frequency of 3,360 CPM falls
directly into the critical speed range of 3,000 to 3,800 RPM previously identified.
Extended operation under this oil whip condition would probably be hazardous to the equipment. Stated in another way, it is generally agreed that startup and shutdown ramps should specifically minimize the time required to pass
through rotor critical speeds. Under no conditions shall a machine be allowed to
dwell within the bandwidth of the rotor resonance. However, in an oil whip condition, the machine is continually running at operating speed, and the rotor is
violently shaking at its natural resonant frequency. It is no wonder that
machines with oil whip often experience significant mechanical failures.
Oil whirl and whip serve as an introduction into the broad topic of machinery instability. Process machinery is susceptible to a wide array of instability
mechanisms ranging from forced instability, to internal friction, to various types
428
Chapter-9
of rubs, and fluid induced aerodynamic instabilities. In some cases the simple
self-excited mechanisms evolve into resonant excitations, and in other instances
the minor instabilities are either tolerated or ignored. There are many excellent
technical papers on this topic. Occasionally, the high level of technology necessary to examine and explain some of these instability mechanisms renders any
potential solutions beyond practical approach. The machinery diagnostician is
encouraged to seek out articles by Donald E. Bently5,6, Edgar J. Gunter7, plus
Allaire and Flack.8 These papers provide practical and understandable explanations for most of the common instability mechanisms.
As mentioned at the beginning of this section, the absence of preloads such
as unbalance or misalignment may contribute to instabilities. Some types of
whirl or whip may be adequately suppressed by the application of preloads such
as intentional misalignment. A hot bearing pedestal may be cooled with water, or
a cool pedestal may be heated with a steam hose to provide a minor degree of stabilizing misalignment. Other preload mechanisms may also successfully restrain
these types of instability mechanisms.
Design characteristics that contribute to instability have been defined for
many years. In fact, machinery designers generally perform many optimization
studies during the development or design of a new bearing, seal, or rotor configuration. For instance, consider the stability diagram presented in Fig. 9-28. This
diagram displays the stability curves for a plain journal bearing (solid line), and
a particular pressure dam bearing (dashed line).
On this type of plot, stability occurs below each line, and instability is predicted for operation above each line. The vertical axis is the stability threshold
speed, and the Sommerfeld number is plotted on the horizontal axis. This is the
same non-dimensional number introduced in chapter 4 of this text. In Fig. 9-28
the various parameters that form the Sommerfeld number are also listed. Thus,
it is possible to examine the general form of the Sommerfeld number for potential clues to stability variations. For example, changes in oil viscosity will
directly influence the stability. Changes in viscosity are generally achieved by
varying the oil supply temperature. It is clear from the stability diagram that
increasing or decreasing the oil viscosity may prove to be beneficial or detrimental. It all depends on the particular operating location within this stability plot.
5 Donald E. Bently, Forced Subrotative Speed Dynamic Action of Rotating Machinery, American Society of Mechanical Engineers, ASME Paper No. 74-Pet-16 (1974).
6 Donald E. Bently, Forward Subrotative Speed Resonance Action of Rotating Machinery,
Proceedings of the Fourth Turbomachinery Symposium, Gas Turbine Laboratories, Texas A&M University, College Station, Texas (October 1975), pp. 103-113.
7 Edgar J. Gunter, Jr., Rotor Bearing Stability, Proceedings of the First Turbomachinery Symposium, Gas Turbine Laboratories, Texas A&M University, College Station, Texas, (1972), pp. 119141.
8 P.E. Allaire and R.D. Flack, Design Of Journal Bearings For Rotating Machinery, Proceedings of the Tenth Turbomachinery Symposium, Gas Turbine Laboratories, Texas A&M University,
College Station, Texas (December 1981), pp. 25-45.
Machinery Stability
429
Stability Threshold Speed (Radians/Sec)
100
Fig. 928 Typical Stability

Diagram For Pressure
Dam And Plain Journal
Bearing Configurations
Unstable
re Dam
10
Pressu
Plain Journal
1
0.01
Stable
Sommerfeld No. =
0.1
Viscosity x Speed x Length x Diameter

Bearing Load
10
2
Radius
Clearance
Fig. 9-28 also shows that changes in speed, load, geometry and bearing
clearance will all influence the bearing stability characteristics. Although the
machinery diagnostician may not be able to fully compute the specific stability
characteristics for a particular machine, the previous array of mechanical
parameters does provide some measure of guidance for potential mechanical
changes, or variations in operating conditions.
Case History 23: Warehouse Induced Steam Turbine Instability
The 4,700 pound steam turbine rotor shown in Fig. 9-29 produces approximately 15,000 HP at 5,500 RPM. This is a five stage turbine with a double flow
exhaust to a surface condenser. The turbine drives a low pressure air compressor,
a speed increasing gear box, and a high stage air compressor. This machinery
train had a generally successful operating history that was occasionally marred
by turbine alignment problems. The most recent difficulties were associated with
subsidence of the foundation sub-structure, and the detrimental effect upon
alignment within this machinery train.
Prior to a maintenance turnaround, the governor end shaft vibration
Fig. 929 Five Stage Steam Turbine Rotor Configuration
430
Chapter-9
response data displayed low amplitudes, combined with a flat or preloaded shaft
orbit, as displayed in Fig. 9-30. Concurrent with this condition, triaxial casing
vibration amplitudes were much higher than anticipated. It appeared that the
governor end bearing was heavily preloaded, and substantial energy was transferred through the bearings and into the governor housing. Based on this information, the governor end bearing was opened for inspection during the
shutdown. It was discovered that the bearing had been heavily loaded, and it
had sustained babbitt damage in the bottom half. A vertical alignment adjustment was made, and the governor end bearing liner was replaced. The exhaust
end bearing was also inspected. It was still in good condition, and was reassembled without any replacements.
Fig. 930 Initial Steady

State Data Of Preloaded
Steam Turbine Governor
End Bearing
The ensuing startup of this machine train was uneventful. Unfortunately,

after several hours of full speed operation the turbine started to misbehave. The
orbits and time base plots in Fig. 9-31 revealed maximum unfiltered amplitudes
in the vicinity of 2.5 Mils,p-p. A definite reduction in the orbit preload was apparent, along with the appearance of a subsynchronous component at nominally
2,640 CPM (49% of rotative speed). The bottom set of orbit and time domain plots
in Fig. 9-31 were filtered at this subsynchronous frequency. Clearly, a forward,
horizontally elliptical subsynchronous orbit was indicated. This was considered
to be unusual, and further examination of the motion was necessary.
The slow roll data, and the initial DC gap voltages of the probes were somehow lost during the startup. Hence, there was no opportunity to examine any
changes in relative shaft lateral position within the journal. There was also a
reduction in casing vibration, but the integrity of the entire casing data array
became highly questionable when a failed velocity coil spring was discovered.
Spectrum plots of the governor end shaft signals are presented in Fig. 9-32.
This FFT data revealed the major subsynchronous component at 49% of rotative
Machinery Stability
Fig. 931 Orbit And Time Base Plots Of

Subsynchronous Steam Turbine Instability
431
Fig. 932 Spectrum Plots Of Subsynchronous Steam Turbine Instability
speed, plus the smaller excitation at running speed, and a small upper sideband
at 1X+0.49X. As expected, plant operations personnel were reluctant to shutdown this machinery train for any additional mechanical inspections. Hence, the
opportunities for comprehending the influence of the turnaround work, plus
understanding the current abnormal behavior were rapidly dwindling.
At this point, the bearing stability characteristics were analyzed on a plot
similar to the diagram presented in Fig. 9-28. Based on the double axial groove
bearing installed in the turbine, it was determined that the stability threshold
was approximately 5,500 RPM. This really meant that at speeds approaching
5,500 RPM the bearing would become unstable. Hence, operation at 5,370 RPM
was uncomfortably close to this stability margin. Fortunately, operations personnel were agreeable to back off the machine load slightly, and at a speed of 5,200
RPM the subsynchronous response was substantially reduced.
Although this turbine was running in a more desirable manner, there were
still many questions to be resolved. A key point was uncovered by one of the
mechanics on the job when he pointed out the fact that the turbine used to run
with pressure dam bearings instead of axial groove bearings. Additional calculations revealed that pressure dam bearings had a stability threshold in the vicinity of 14,000 RPM. Hence, the use of pressure dam bearings on this turbine
would be a major stability improvement over the double axial groove liner
installed during the overhaul.
Further investigation revealed that the culprit resided in the plant warehouse. It seems that the turbine was originally supplied with axial groove bearings, but they proved to be unstable at full process rates. The OEM designed a
set of pressure dam bearings for this machine, and these modified bearings plus
432
Chapter-9
two sets of backup spares allowed many years of successful operation. Following
the foundation subsidence problems, the turbine consumed several sets of bearings. As expected, the warehouse reordered spare bearings, but they used the
part number for the original axial groove bearings instead of the newer part
number for the pressure dam bearings. Thus, the wrong bearings were retained
in stock, but no one noticed since the immediate problem was associated with the
settling of the foundation.
Eventually, the foundation subsidence was controlled, and the affected
baseplates were re-leveled, re-grouted, and the machinery train was properly
aligned. The correct pressure dam bearings were installed in the turbine, and
the warehouse corrected the erroneous part numbers. Shaft and casing vibration
characteristics returned to normal, and the governor end bearing failures ceased.
In the final overview, this turbine that required some type of maintenance on a
regular basis was transformed into a machine that is only opened for inspection
on 8 to 10 year increments.
Case History 24: Pinion Whirl During Coastdown
Most process machinery trains exhibit the highest vibration amplitudes
during startup. Occasionally, some machines will vibrate excessively during
coastdown. Such a machinery train is depicted in Fig. 9-33. In this train, the
steam turbine driver accepts superheated steam, and it exhausts to a low pressure 35 Psi header. The turbine is rated for 9,300 HP at the maximum continuous speed of 11,900 RPM. The turbine operating speed range varies between
8,000 and 11,000 RPM. As shown in Fig. 9-33, the turbine is directly coupled to a
two element speed reducing gear box. This is a down mesh box with 41 pinion
Inlet
700 psig
at 750F
CHD
Exhaust
DHD
45
45
45
45
CVD
DVD
Down Mesh
K hs
Steam Turbine
T1
Pinion
41 Teeth
220 Lbs.
9,300 HP @ 11,900 RPM

8,000 to 11,000 RPM
T2
CW
Single
Stage
Gear
Box
T3
AHD
45 45
BVD
BHD
45 45
Overhung
Blower
T5
Bull
Gear
126 Teeth
1,629 Lbs.
T4
AVD
K ls
T6
3,990 Lbs. Rotor
Ratio 3.0732:1
GVD
GHD
Side
Discharge
45 45
CW
EVD
EHD
45 45
FVD
FHD
45 45
HVD HHD
45 45
CCW
Fig. 933 Machinery Arrangement For Turbine Driven Overhung Blower
End
Suction
Machinery Stability
433
teeth, and 126 bull gear teeth. The gear set is a double helical arrangement, with
a double acting bull gear thrust bearing. The bull gear output coupling drives a
single stage overhung blower that contains a 3,990 pound rotor. This air machine
accumulates various foreign substances, and it must be cleaned periodically.
Under normal startup conditions, this machinery train seldom exhibits
shaft vibration amplitudes in excess of 3.0 Mils,p-p. Operating at full speed and
load, the runout compensated shaft vibration amplitudes are generally below 1.0
Mil,p-p at all measurement locations. From the train diagram it is clear that all
bearings are monitored with two proximity probes. Each journal bearing has X-Y
radial probes, and dual axial probes are mounted at each thrust bearing. Hence,
the train is well-monitored, and the possibility of undetected high vibration levels on this train is quite remote.
Across the down mesh double helical gear box it is significant to compute
the forces associated with this machinery. For example, Fig. 9-34 identifies the
bull gear weight as 1,630 pounds, and the pinion weight as 220 pounds. For a
reduced load operation of 4,000 HP, the transmitted forces are shown for a pinion
speed of 5,900 RPM, and 1,920 RPM on the bull gear.
Tangential
Force
10,880 Lbs.
Separation
Pinion
Force
41 Teeth
4,550 Lbs.
CW
Rotn.
Bull Gear
126 Teeth
Separation
Force
CCW
Rotation
4,550 Lbs.
Pinion
Weight
220 Lbs.
Fig. 934 Summary Of

Significant Radial Forces
Present In Mating Helical
Gear Set
Tangential
Force
10,880 Lbs.
Bull Gear
Weight
1,630 Lbs.
Under this reduced load operation, the tangential gear forces are equal to
10,880 pounds, and the companion gear separation forces are equal to 4,550
pounds. These gear contact forces are substantially larger than the weights of
the gear elements, and they dominate the force structure on this machine. This
relationship between forces is typical for many industrial gear boxes.
It is clear that significant forces are active across the gear teeth, and the
bearings respond with large stiffness values. In most startup and operating conditions, the gear elements exhibit minimal shaft relative vibration. However,
during coastdown, the normal behavioral pattern is reversed, and the pinion displays high shaft vibration as shown in the partial Bode plot of Fig. 9-35. From
434
Chapter-9
Fig. 935 Partial Bode Plot Of Pinion

Vibration During Normal Coastdown
Fig. 936 Partial Cascade Plot Of Pinion

Vibration During Normal Coastdown
full running speed to approximately 2,250 RPM the pinion motion remains below
0.5 Mils,p-p. At a speed of 2,250 RPM, the unfiltered pinion vibration levels suddenly begin to climb, and it reaches a maximum amplitude of 8.0 Mils,p-p at
approximately 800 RPM. As shown in Fig. 9-35, the synchronous 1X amplitudes
remain small, and the 1X phase angle remains essentially constant. The majority of the shaft vibration must occur at a frequency other than rotative speed.
The examination of the cascade plot in Fig. 9-36 reveals that the majority of the
shaft motion occurs at 49% of running speed. The shaft orbits and time domain
plots at 1,000 RPM are shown in Fig. 9-37 at a coastdown speed of 1,000 RPM.
Fig. 937 Orbit And Time

Base Plots Of Sustained
Oil Whirl On Pinion During
Coastdown
Mechanical Looseness
435
From this information it is clear that the pinion is experiencing a forward circular oil whirl mechanism at 49% of rotative speed.
This oil whirl instability is almost totally associated with the reduction in
gear contact forces as the machinery slows down. The most susceptible element
in the machinery train is the 220 pound pinion. It is concluded that, as the unit
slows down, the gear forces are gradually diminished, and the unloaded pinion
goes into an oil whirl that eventually occupies the full bearing clearance. Other
gear boxes also display this same type of coastdown whirl. Since this is an
unloaded condition, the transmitted forces between the pinion journals and the
bearings are negligible. The whirl will generally remain active until the machine
stops rotating. This type of whirl is generally non-destructive, but the diagnostician should always check gap voltages before startup, and at full speed and load
to verify that the bearings have not been damaged.
In this particular gear box, the 8.0 Mil,p-p excursions during coastdown
have occurred with clock-like regularity. As noted, this type of oil whirl is generally non-destructive, but it cannot be ignored. It could be reduced or eliminated
with a pinion bearing redesign. However, the expenditure for this modification
could not be justified by the operating company. The unfortunate part of this
story is that the unit is equipped with 0 to 5.0 Mil,p-p radial vibration monitors.
During coastdown of this train, the operators are accustomed to having the pinion vibration monitors fully pegged at speeds below 2,000 RPM. No one has any
idea if the maximum whirl amplitude approached 8.0 Mils,p-p (equivalent to the
pinion bearing clearance) or if the maximum whirl amplitude was 12.0 or 15.0
or 20.0 Mils,p-p (indicative of pinion bearing damage). There are many ways to
acquire and present this valuable coastdown information to the control room
operators. However, it is anticipated that no action will be taken until the pinion
bearings are wiped out during some future coastdown, and the operators attempt
to restart the machinery train with damaged pinion bearings. At that time, the
repair costs for the gear box, plus the associated production losses, will make the
condition monitoring expenditures appear to be insignificant.
MECHANICAL LOOSENESS
The last section addressed instabilities such as oil whirl and whip. The selfexcited whirl may be attributable to various sources of cylinders within cylinders
such as the bearings, seals, balance pistons, or wheels. As previously stated,
whirl generally appears at subsynchronous frequencies between 35% and 49% of
rotative speed. On the other hand, whip behavior occurs at resonant frequencies
that are consistent with the current effective stiffness. For rotors that operate
above one or more resonances, the potential whip frequencies can vary over an
appreciable frequency range. However, there are other subsynchronous mechanisms that appear on centrifugal machines. Some of these mechanisms produce
significant excitations at fractional frequencies of rotational speed. For instance,
vibration components at precisely 1/2, 1/3, and 1/4 of running speed may appear.
436
Chapter-9
If rotating speed is changed, the subsynchronous component will change to

maintain a constant fractional relationship to the running speed. Thus, the subsynchronous vibration component remains locked-in, or fixed at a fraction of
rotational speed. It should be clarified that these characteristics are applicable to
flexible rotors that operate above one or more shaft critical speeds. For stiff shaft
rotors that operate below any shaft critical, the appearance of fractional vibration components with an appreciable amplitude is much less common.
Many of these fractional frequency mechanisms were examined and discussed by Donald E. Bently9 in 1974. He showed that stiffness variations could
result in a change in the effective balance resonance frequency (critical speed) of
a machine running at full operating conditions. When this behavior is combined
with a force such as rotational speed unbalance in a lightly damped system, this
may produce a forced excitation of a system resonance.
Some individuals might argue that stiffness cannot change, and since the
mass is constant, the rotor natural frequency must remain at a fixed value. However, it must be recognized that stiffness is not a static quantity. Stiffness does
change with variations in parameters such as speed, bearing preload, physical
clearances, mechanical fits, applied loads, etc. In chapter 4 of this text it was
shown that oil film stiffness varies with machine speed (e.g., Fig. 4-1). In the
same chapter, the measurements of housing stiffness (e.g., Figs. 4-18 and 4-20)
also reveal variations in the housing stiffness as a function of frequency.
In all cases, the observed critical speed measured during a startup or a
coastdown is predominantly governed by the effective stiffness that is active at
that specific transient speed condition. To state it another way, the physical
parameters that contribute to the transient critical speed do not have the same
values at full operating speed and load. Although the effective values of transient versus steady state stiffness may be close to each other, the physical situation has changed, and the full speed critical may vary from the transient critical.
This behavior is clearly demonstrated by a comparison of startup versus coastdown vibration response characteristics. The resonance identified during a startup will generally be higher than the coastdown value. In essence, the
mechanical system displays an effective relaxation of stiffness during the shutdown. Thus, stiffness will vary, and the rotor resonance(s) will be directly associated with the effective stiffness at any point in time.
Back to the condition of a machine running at full speed. If it is assumed
that the rotor is operating above one or more rotor critical speeds, and if there is
an appreciable synchronous forcing function such as unbalance the stage is
set for a variety of subsynchronous mechanisms. For example, if a stationary
machine part such as a bearing housing becomes loose, the effective support
stiffness of the rotor is reduced in accordance with general equation (4-15), or
the simpler format of equation (4-16). This combination of a synchronous force
combined with a reduced support stiffness allows the rotor resonance to shift
into a frequency that is an even multiple, or a fixed fraction of rotating speed.
9 Donald E. Bently, Forced Subrotative Speed Dynamic Action of Rotating Machinery, American Society of Mechanical Engineers, ASME Paper No. 74-Pet-16 (1974).
Fig. 938 Orbit and Time Base Plots of

Loose Bearing with Locked-In Vibration
Component At One Half of Running Speed
437
Fig. 939 Spectrum Plots of Loose Bearing With Locked-In Vibration Component At
One Half of Running Speed
For demonstration purposes, consider the orbit and time base plots shown
in Fig. 9-38, plus the spectrum plots in Fig. 9-39. This data is representative of a
unit running at 11,520 RPM with a startup first critical speed of 6,200 RPM. For
a machine running at this speed, the nominal 1.4 Mils,p-p of 1X synchronous
motion is somewhat higher than desired. This running speed vibration could be
due to mass unbalance, excessive bearing clearance, or a combination of the two.
In this case, the machine displayed a locked-in 50% component at 5,760 CPM.
The FFT plots also displayed interactions between the 1X running speed and the
X/2 subsynchronous frequency. This included fractional frequency components at
3X/2, 5X/2, plus a minor response at 7X/2. Overall, the 50% fractional frequency
is considered to be the first critical that has been lowered by a reduced support
stiffness. Interaction between the 1X and the X/2 frequencies produces the string
of fractional speed components.
This data was obtained shortly after startup, and the overall vibration levels were considered to be unacceptable. The unit was shutdown, and it was discovered that the outboard bearing cap bolts were not properly torqued. This loose
bearing housing reduced the effective system stiffness, and combined with the
unbalance forcing function to generate the behavior displayed in Figs. 9-38 and
9-39. Tightening the bearing cap bolts to the correct torque level corrected this
problem, and the 50% of rotative speed component plus the string of fractional
components were eliminated.
In the overview, the generation of subsynchronous vibration components
that are integer fractions of rotative speed (e.g., X/2, X/3, or X/4) primarily
appear in underdamped systems that have a loose stationary mechanical ele-
438
Chapter-9
ment combined with a high amplitude forcing function such as unbalance. It is

possible to display fractions of excitations at other frequencies, but rotational
speed unbalance is the most common driving force. When this energy is transferred to the subsynchronous region, the resultant frequency is a re-excitation of
a fundamental rotor resonance (with a decreased stiffness). The fractional frequency will generally have a forward precession, and the locked-in subsynchronous component will interact with the other major excitations to produce a string
of fractional frequencies.
It should also be mentioned that pulsating torque in a motor or generator
may produce similar characteristics. Also, machines with various types of rubs or
asymmetrical shaft stiffness (non-circular cross section, or near a keyway) will
exhibit some of the same behavior. The diagnostician should also be aware that
some cases of mechanical looseness (e.g., loose hold down bolt) may only appear
as a high vertical vibration component at running speed on the casing or baseplate. This fundamental synchronous motion may become truncated due the relative movement or hammering of parts, and the truncated 1X sine wave would
then exhibit a series of running speed harmonics without the presence of any
subsynchronous frequency (e.g., Figs. 7-35 and 7-36).
Case History 25: Loose Steam Turbine Bearing
Another example of this behavior is shown in the steam turbine orbits presented in Fig. 9-40. This data was acquired on the exhaust bearing of an 8 stage
turbine that was rated at 5,400 HP at 7,490 RPM. This steam turbine was driving a six stage propane compressor. Due to an operational error, this refrigeration train had been subjected to a period of overspeed operation. This resulted in
significant mechanical damage, and it forced a major overhaul of both machines.
Following the re-build, the unit had a normal startup, and it displayed higher
than desired vibration levels at the turbine exhaust bearing. During the next
week, vibration amplitudes gradually increased, and the data shown in Fig. 9-40
represents the shaft vibration 9 days after startup.
Fig. 940 Shaft Orbits Of Loose Steam Turbine Bearing Housing with
Locked-In Vibration Component At One Third of Running Speed
The unfiltered orbit at the left side of Fig. 9-40 had vertical amplitudes in
excess of 9.0 Mil,p-p. Since the bearing diametrical clearance was only 6.0 Mils,
the observed vibration was substantial. By comparison, the 1X rotational speed
439
orbit shown in the middle of Fig. 9-40 seemed to be quite small, but the horizontal shaft motion at 7,245 RPM was 1.7 Mils,p-p. Hence, the running speed vibration was appreciable, but it was dwarfed by the subsynchronous component. The
orbit on the right side of Fig. 9-40 was filtered at X/3 or 2,415 CPM. Clearly, the 3
stationary Keyphasor dots are indicative of a locked component at one third of
running speed. Furthermore, the precession of this X/3 vibration is counterclockwise. which is forward, and in the direction of shaft rotation.
This shaft vibration data is shown in the frequency domain in Fig. 9-41.
Once again, the dominant excitation at one third of running speed appears at
2,415 CPM. In addition, the interaction components between 1X and X/3 are
identified at 2X/3, 4X/3, 5X/3, and 7X/3. The same type of behavior is transmitted
through to the bearing housing as illustrated in the casing plots of Fig. 9-42.
Fig. 941 Shaft Spectrums of Loose Turbine Bearing Housing with Locked-In Component At One Third of Running Speed
Fig. 942 Casing Spectrums of Loose Turbine Bearing Housing with Locked-In Component At One Third of Running Speed
To verify if this subsynchronous frequency tracked running speed, a 1,000

RPM speed change was performed. At 7,890 RPM the X/3 component had
increased to 2,630 CPM, and it was still locked into one third of rotative speed.
As the train speed was lowered to 6,870 RPM the subsynchronous frequency
dropped to 2,290 CPM which was still exactly one third of running speed. A further speed decrease to 6,840 RPM was sufficient to completely eliminate the subsynchronous vibration component. However, as soon as the X/3 component
decayed, the rotational speed vibration increased to 5.6 Mils,p-p horizontally, and
5.2 Mils,p-p vertically. As speed was increased, the rotational speed energy was
transformed back into the X/3 subsynchronous component, and it reappeared in
an identical manner to the previous Figs. 9-40 through 9-42.
Based on this information, it was concluded that mechanical looseness in
440
Chapter-9
the turbine exhaust end bearing housing was the most probable culprit. The unit
was operated for another two weeks before it could be shutdown for repair. At
that time it was discovered that the exhaust end bearing was improperly fitted
into the housing, and the bearing was actually supported by the anti-rotation
pins. As an interim measure, stainless steel shims were installed between the
bearing and the housing to fill in the clearance cavity. This temporary fix proved
to be quite effective, and the turbine operated with shaft vibration amplitudes of
less than 2.0 Mils,p-p for the duration of the two-year process run.
ROTOR RUBS
The physical contact between rotating elements and stationary machine
parts can generate a variety of rub conditions. For example, the following general
categories of rubs are encountered on process machinery.
Laby Rubs Shaft against close clearance aluminum labyrinths.

Intentional Rubs Rotating labyrinths cutting into abradable seals.
Light Rubs Short duration rubs due to process or external upsets.
Intermittent Rubs Due to tight clearances or pseudo bearings.
Heavy Rubs Due to foreign object ingestion, blade loss, or bearing failure.
Catastrophic Rubs Heavy radial or axial rub due to broken shaft or failed
coupling, also may occur during extensive blade failures.
The first category of laby rubs are quite common in machines such as centrifugal compressors with interstage aluminum labyrinths. During an overhaul,
these labys are often installed with undersized radial clearances, and the rotor
establishes the running clearances during startup and normal operation. These
are usually minor rubs that often escape detection due to their minimal severity.
The intentional rub category addresses the cases where abradable seals
are installed at impeller eyes, between impellers, or around balance drums.
These close clearance seals consist of a stationary abradable material combined
with rotating labyrinths. Typically, the machine is placed on slow roll, and the
rotating labyrinths cut their own running clearance into the abradable material.
In some cases the running clearances may be established in less than an hour,
and in other situations a full 8 hour shift might be required for the labys to cut
the proper clearances. As the machinery is started up, additional rubs may occur
as the labys cut further into the abradable material.
Light rubs during normal operation may be due to process upsets where
liquids are carried over into a compressor, or a minor surge develops due to
downstream process control problems. External influences such as earthquakes,
or heavy equipment rolling by the machinery deck may be sufficient to excite the
rotor and/or casing and produce a brief contact between the stationary and rotating machine elements. These types of events are also hard to detect unless the
diagnostician happens to be viewing vibration signals on an oscilloscope at the
time of the rub event.
Rotor Rubs
441
Intermittent rubs due to tight clearances are the most commonly

observed rub. This malfunction will either clear itself, or progress into a heavier
rub with associated mechanical damage. The tight clearances may be due to mispositioned stationary machine elements such as seals, or cocked rotating elements such as impellers or thrust collars. These close clearance elements often
act as pseudo bearing for the rotor system, and they provide additional lateral
restraint. The extra stiffness due to the rubbing element is often sufficient to
temporarily raise the critical speed, and allow the rotor to lock into this subsynchronous resonance. In the frequency domain, the intermittent rub looks like a
loose bearing housing with integer fractions of rotative speed (e.g., X/2, X/3, or X/
4) plus a string of fractional frequencies (e.g. 3X/2, 5X/2, 7X/2, etc.). This mechanism requires an initial driving force such as unbalance, and it is more prevalent
in systems with low damping. In many cases, the dominant subsynchronous component will have a backward precession, and the orbits may be unidirectional.
The appearance of heavy rubs due to problems such as the ingestion of foreign objects, the breakage or failure of turbine blades or compressor impellers,
plus the failure of journal or thrust bearings. These types of rubs are driven by a
large 1X rotational speed force, and they typically result in the automatic shutdown of the machinery. For units that are not equipped for auto trip on high
vibration, the excessive shaft and casing vibration levels usually convinces the
operators to shutdown the equipment. These types of rubs are generally not
investigated for very long because the machinery is shutting down.
The final category of catastrophic rubs are caused by failures on the
rotating assembly. These full radial or axial rubs may be due to broken shafts,
failed couplings, cracked gears, multiple blade failures, or any other malfunction
that compromises the structural integrity of the rotating assembly. Rubs of this
type are fully destructive, and when they occur, the machine is definitely coming
down with excessive internal damage. In some situations, various rotor elements
may even break through the casing and produce additional destruction.
From the above discussion it is reasonable to conclude that many types of
machinery rubs are possible, but very few rubs are of sufficient duration to be
properly investigated. The intermittent rubs due to tight clearances are probably
the only type of machinery rub that is commonly encountered and documented.
For instance, consider the spectrum plots of a shaft rub in Fig. 9-43. This data
was obtained on a 5,760 RPM rotor, and it clearly shows a major subsynchronous
excitation at 1,440 CPM. This frequency is exactly one fourth of running speed.
The two FFT plots also show the interaction between the rotational speed (1X)
vibration and the subsynchronous component (X/4) as a string of fractional frequencies (X/2, 3X/4, 5X/4, etc.). For reference purposes, the measured first critical speed of this rotor was approximately 1,300 RPM. It is logical to believe that
the subsynchronous component at 1,440 RPM is probably a re-excitation of the
first mode with additional lateral stiffness.
Obviously the data in Fig. 9-43 closely resembles the case of a loose stationary element that was discussed in the previous section. This is a difficult situation for the machinery diagnostician since both malfunctions (looseness and rub)
look the same in the frequency domain. If you shutdown the equipment and ini-
442
Chapter-9

of Radial Shaft Rub And
Locked-In Frequency
Component At One Fourth
of Running Speed
tiate an inspection for looseness in one of the stationary elements, and actually
discover a shaft rub, this can be quite embarrassing. One way to gather more
information on the actual mechanism is to do a complete job of examining the
available data. Specifically, if the shaft vibration FFT data shown in Fig. 9-43 is
presented in the orbital domain, the data shown in Fig. 9-44 may be examined.
Fig. 944 Orbits Of Radial Shaft Rub And Locked-In Frequency

Component at One Fourth of Running Speed
The unfiltered orbit on the left side of Fig. 9-44 shows the influence of the
one quarter running speed component, plus an overall counterclockwise precession. This is in the same direction as the synchronous 1X orbit showing a counterclockwise rotation in the middle diagram of Fig. 9-44. However, when a bandpass filter is applied to the overall signal, the data may be filtered precisely at
the X/4 subsynchronous frequency of 1,440 CPM, as shown on the right side of
Fig. 9-44. In this orbit, notice that the precession is clockwise, or in the direction
against rotation. This behavior is due to the fact that when the rotor hits the stationary rub point, the shaft is kicked back against rotation, which produces the
reverse precession orbit at the subsynchronous frequency.
Cracked Shaft Behavior
443
Depending on the type of rub, the subsynchronous frequency might drift

around slightly, and it may not stay locked into a fraction of running speed. This
is simply due to the fact that the rub conditions are continually changing, and
this produces variations in the rotor response. In addition, the subsynchronous
motion might drift back and forth between forward and reverse precession.
Again, this is due to the variations in the rub. In virtually all observed cases, a
heavy rub is characterized by a reverse precession at the subsynchronous frequency, plus a string of fractional components at appreciable amplitudes. Finally,
the multiple harmonics of running speed and the subsynchronous component
normally appear on both the shaft and the casing. However, the precession of the
casing subsynchronous motion will not be as reliable of a rub indicator as the
precession of the shaft orbit.
CRACKED SHAFT BEHAVIOR

One of the most dangerous problems in a mechanical equipment train is a
shaft failure. When this occurs, the transmitted horsepower across the broken
rotor is suddenly released. This energy may be consumed or dissipated by the
process stream(s), or it may sling parts such as couplings considerable distances
from the machinery. There have been reported cases of machine parts traveling
half a mile or more from the point of failure. Hence, cracked shafts are potentially hazardous to the physical plant facility, as well as the local personnel.
Machines that are subjected to frequent startups and shutdowns appear to
be more susceptible to shaft cracks due to the increased number of cycles
through the rotor resonance(s), plus the process heating and cooling. This is particularly true on power generation units that are started and stopped daily, and
rotors that are subjected to shaft bending modes. Overhung rotors with large
wheels such as vertical pumps also appear to be prime candidates for cracked
shafts. Certainly, reciprocating machines with their alternating forces and
stresses may likewise experience this type of failure.
Cracks may originate at high stress points such as the square corners of a
keyway, or they may occur underneath wheels, impellers, or inside hollow shafts.
In some instances, crevices or scratches may be subjected to chemical attack, and
these locations may grow into full shaft cracks. In other cases, the rotor may be
subjected to stress corrosion cracking. Obviously when lateral shaft bending
modes are involved, or stress reversals occur across torsional nodes, there is a
potential for this type of failure mechanism. It is virtually impossible to define
and categorize all of the plausible cracked shaft failure scenarios due to the
potential combinations of lateral and shear forces, bending moments, and torsional forces for every type of machine.
However, there is a common ground for detection of cracked shafts when it
is recognized that shaft stiffness must decrease in the presence of a crack. This
basic concept allows the detection of shaft growth by measuring the historical
changes in synchronous 1X shaft vibration. For instance, Figs. 9-45 and 9-46 document the vibration characteristics at the exhaust end of a steam turbine. This
444
Chapter-9

Steam Turbine With Solid Shaft
Fig. 946 Spectrum Plots Of Steam Turbine With Solid Shaft
data was taken under normal steady conditions with a solid rotor. The presence
of the 0.25 Mil,p-p component at twice running speed 2X is nothing more than the
influence of a surface imperfection below the proximity probe. This same steam
turbine with a cracked exhaust end shaft is presented in Figs. 9-47 and 9-48.
This 45 crack was due to torsional fatigue, and it was estimated that the crack

Steam Turbine With 10% Shaft Crack
Fig. 948 Spectrum Plots Of Steam Turbine With 10% Shaft Crack
445
extended through 10% of the shaft diameter. Note that the 1X rotational speed
vector changed, but there was minimal variation throughout the remainder of
the vibration signals. Some might argue that there must be measurable changes
in the higher order running speed harmonics such as the 2X, 3X, or 4X. Although
these types of harmonic changes occur is some situations, the experience of the
senior author of this text is that changes in the running speed 1X vectors represents the primary indication of cracked shafts. Investigators such as Bently and
Muszynska10 have found some useful information at higher order harmonics, but
the fundamental mechanism still appears at rotational speed for the vast majority of the documented cracked shafts.
Depending on the type and location of the crack, plus the magnitude of the
runout vectors, it is reasonable to expect that 1X vector changes due to shaft
cracks may appear as either increasing or decreasing vibration amplitudes. In
the example shown in Figs. 9-45 through 9-48 the 1X vectors changed on both of
the proximity probes. However, the cracked shaft data shown in 9-47 and 9-48
does not appear to be abnormal. If the initial plots (9-45 and 9-46) were not available for historical comparison, there would be minimal reason to suspect a
machinery problem based only on Figs. 9-47 and 9-48. Hence, the careful trending of the 1X vibration vectors is mandatory for proper condition monitoring.
Other symptoms of cracked shafts may appear on many types of machines.
For instance, unexplained changes in shaft slow roll vectors may be due to a
shaft crack. For any machine, the slow roll runout vectors measured with proximity probes under similar conditions must be repetitive. In addition, any unexpected changes in full speed 1X vectors due to minor process variations, small
speed changes, or minor load changes should be viewed with suspicion. Furthermore, startup and shutdown data on major machinery trains should always be
acquired and examined. Appreciable changes to balance resonance frequencies
(critical speeds) or variations in amplification factors should be questioned.
Some machines are equipped with sufficient radial vibration transducers to
be able to detect the general shaft mode shapes of the entire machinery train.
This is particularly feasible for large turbine generator sets with hard couplings
and multiple locations for radial vibration measurements. For these types of
machines it makes sense to plot the runout compensated 1X running speed vectors into a machinery train mode shape using one of the techniques discussed in
chapter 3. These train mode shapes should be documented in a new or re-built
condition, and checked periodically with current measurements. Any significant
variations in these mode shapes may be indicative of a developing shaft crack.
Finally, abnormal response to balance shots should be carefully examined.
It makes sense to compute the full array of balance sensitivity vectors using
equation (11-17), and tabulating the results (e.g., Table 11-1). Unusual changes
in these presumably constant balance sensitivity vectors may be an early warning of a potential crack in a machine shaft.
10 Donald E. Bently and Agnes Muszynska, Detection of Rotor Cracks, Proceedings of the Fifteenth Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University, College
Station, Texas (November 1986), pp. 129-139.
446
Chapter-9
Case History 26: Syngas Compressor with Cracked Shaft

Over the years various alliances have been formed between operating companies to share technology and spare parts for similar operating plants. These
spare parts pools have achieved some good measures of financial and technical
success. However, there are occasional problems that develop in this type of environment that are not solved, but are inadvertently passed along from one partner corporation to another. The following case history describes the events
associated with one such machinery problem.
The rotor in question was a common spare for a high stage syngas compressor. It was assembled with a completely refurbished set of diaphragms into a
complete bundle assembly. This spare bundle was installed in a barrel compressor during a routine maintenance overhaul. The spare rotor dimensions, impeller
fits, axial spacing of elements, and all bearing clearance were well within normal
specifications. All dimensions had been checked by two independent inspectors,
and there was no reason to anticipate any problems with this machine.
Following the conclusion of this maintenance turnaround, the operating
personnel experienced the usual array of startup problems. When the syngas
train was ready for slow roll, the majority of the problems in the remainder of
the plant had been corrected. It certainly appeared that a long month of intensive work was rapidly drawing to an end. The syngas train startup was initially
330
ot
at
io
>>
0
30
9,200
9,400
60
300
8,200
6,800
5,600
90
9,000
270
7,200
7,700
9,600
8,500
8,800
120
Fig. 949 Coastdown

Polar Plot Of Syngas Compressor With Subsurface
Circumferential Crack In
The Shaft
240
9,800
150
210
Normal First
Critical Response
180
Dots ( ) Indicate Speed in RPM

Plot Radius = 6.0 Mils,p-p
uneventful, and all 4 rotors passed through their respective critical speed ranges
with no problem. In fact, the maximum vibration level of the high stage compressor through the critical speed was less than 1.0 Mil,p-p. At this stage, everyone
was starting to smile, and beginning to think about a full nights sleep, plus a
Sunday afternoon barbecue. Unfortunately, the festive atmosphere in the control
447
room grew progressively quieter as the syngas train speed increased, and the
high stage compressor vibration continued to grow. Normal operating speed for
this machinery train was in the vicinity of 10,200 to 10,500 RPM. At an operating speed of 9,800 RPM the shaft vibration was all at 1X running speed, and the
runout compensated 1X amplitudes were nominally 5.3 Mils,p-p. Since the bearing diametrical clearance was slightly less than 8.0 Mils, it was considered to be
unwise to allow any further increases in shaft vibration.
The syngas train was then slowly unloaded, speed was gradually decreased,
and the polar plot data presented in Fig. 9-49 was recorded. This data was post
processed from the field tape recording. During the actual coastdown, the oscilloscope revealed that the 1X amplitude decreased, and the 1X phase rolled around
the orbit as speed and load were reduced. This field observation was fully
reflected in the spiraling polar plot shown in Fig. 9-49. That is, the 1X amplitudes decreased with speed, and the phase angle rolled continuously. The phase
actually changed by about 360 for every 1,000 RPM of speed drop. This was certainly abnormal behavior, and there was no direct explanation. As the compressor reached the critical speed range, it passed through this balance resonance
with low amplitudes, and normal 1X vector response.
After shutdown, the bearings were pulled and inspected, but no significant
damage was found. The coupling to the low stage syngas compressor was thoroughly checked, and it was found to be in excellent condition. All four of the
radial, and both of the axial proximity probes on the high stage were checked for
proper calibration on the actual shaft material. Again, no abnormalities were
discovered. All of the compressor hold down bolts were checked, the foundation
was checked, all of the large bore piping was examined for correct location and
proper spring hanger settings. Again, there was no obvious culprit to blame for
the observed behavior of the high stage compressor.
Some of the plant personnel wanted to put it back together, cut the probe
wires and just go and run the machine. Fortunately, the plant management
elected to strip down the compressor and find the root cause of the problem.
Although this was a difficult decision, it was certainly the correct thing to do.
Shop disassembly of the unit provided no additional clues at to the origin of
the abnormal 1X vector changes with speed and load. All of the impellers, the
balance piston, and the thrust collar dimensions were proper, and well within
normal assembly tolerances. All of the diaphragm fits and clearances were
proper, and axial spacing was correct throughout the entire length of the bundle.
All internal passages were clear, and the seals were in good condition. Initially,
the only good news from the shop disassembly and inspection was that the midspan impeller labyrinths had contacted the shaft. This physical observation reinforced the high vibration measured by the shaft proximity probes.
As potential failure mechanisms were systematically eliminated, the probability of a cracked shaft became more and more plausible. To check for cracks in
the shaft and impellers a dye penetrant inspection was used. The dye check produced no indication of shaft cracks, and a further test of the shaft was conducted
with the additional sensitivity of a Zyglo inspection. Once again, there was no
indication of any shaft crack or discontinuity. At this stage, there was no physical
448
Chapter-9
evidence of any problem with this high stage barrel compressor rotor.
During a re-examination of all of the shop test data, it became apparent
that the visual inspections, dye penetrant checks, and Zyglo tests are all predicated upon the fact that the crack must extent to the surface of the machine element. If the crack was somehow retained below the surface, these inspection
techniques would not identify the presence of a crack. On the basis of this conclusion, a final check of the compressor shaft was performed with ultrasonic inspection. Amazingly, the ultrasonic test revealed a circumferential crack below one
smooth section of the shaft as illustrated in Fig. 9-50. This crack was approximately 1/4 to 3/8 of an inch below the shaft surface, and it extended over an arc
of approximately 60. The length of the crack could not be accurately determined
due to various steps and keyways in the shaft. However, it was clear that this
crack was present, and it was significant in size.
Compressor Shaft Section
Crack
Fig. 950 Circumferential

Shaft Crack In High Pressure Syngas Compressor
Another spare rotor was pulled from the parts pool. The bundle was then
reassembled, and reinstalled in the machine. Although the balance on this spare
rotor was not as refined as the previous unit, it did come up and run at full speed
without any apparent sensitivity to load. Above 8,000 RPM the 1X vectors were
rock solid, and the plant was back in full production within the next 72 hours.
The question that remained was what caused the subsurface crack in the
pool rotor? In an effort to answer this question, the serial number of this rotor
was used to trace the operating history of this shaft. Interestingly enough, it was
determined that this shaft had never successfully operated under load. Although
it had been installed in several machines over the years, they all experienced
high vibration levels during startup or loading. In each case, this rotor was
replaced with a functional unit, and the cause of the problem was never determined. The rotor was always stripped and inspected after each failed run, but no
problems were detected. Hence, the various members of the spare parts pool that
worked with this specific rotor could find no difficulties, and they had it restacked, and returned back to the available pool of spare rotors.
Clearly this subsurface crack had existed for many years. It is speculated
that it was an inclusion in the original shaft forging. Unfortunately this unusual
shaft crack was not discovered until a considerable amount of time and money
had been wasted by several different operating companies.
Foundation Considerations
449
FOUNDATION CONSIDERATIONS
All machinery trains are supported by some type of foundation. Clearly, the
foundation must be rigid enough to maintain alignment of the equipment, and it
must be strong enough to accommodate the dynamic or vibratory loads emitted
by the machinery. These oscillating loads cover the full range of dynamic forces
from simple mass unbalance to complex impact or shock loads. In addition, the
supports must be able to handle thermal distortion from the machinery, plus the
loads and moments imposed by the piping systems.
In some cases, foundations may be difficult to design, and they may be difficult to analyze. Historically, foundation designs have been based upon empirical
solutions, plus general rules of thumb. Concepts such as keep the center of gravity
of the entire structure well below grade, or have the foundation weight five times
the weight of the machinery are quite nebulous. In recent years, these trial and
error designs have been supplemented by computerized structural design programs that can address the static as well as the dynamic loads. Although this
approach adds increased sophistication to the analysis, the results of these computer solutions should always be examined by an experienced field engineer to
validate the physical appropriateness of the foundation design.
Machinery on airplanes, ships, drilling rigs, trucks, and locomotives are
supported by structural steel frames and baseplates. In these applications the
foundation is flexible, and considerable computational technology must be
employed to provide an adequate support for the equipment. Flexible foundations are also used for machinery in large residential or office buildings. In these
installations, the equipment is often isolated from the surrounding structure
with spring mounted baseplates, isolation pads, plus shock absorbers. The transmissibility of machine vibration to the surrounding structure, and the eventual
impact upon the human occupants are key issues in these types of installations.
Again, the dynamic machinery forces must be considered, but the machinery
vibration must not be transmitted to the people.
Within the process industries, most of the machinery is installed on rigid
steel and concrete foundations. Various configurations are used that vary from
foundations built on pilings in swampy regions, to solid monolithic structures
built at grade level. There are also a variety of flexible foundations mounted on
slender columns. Although the majority of the machinery foundations are solid
reinforced concrete structures, the diagnostician must recognize that flexible
concrete structures are possible. In many cases, these flexible foundations are
elevated units (20 to 100 feet tall) that locate the machinery close to other process equipment such as chillers or upper levels of fractionating columns. These
installations should be carefully examined for the potential of structural resonances. It is common to discover one or more structural natural frequencies
below machine operating speeds. In these installations, the machinery train is
subjected to the structural resonance(s) during every startup and shutdown.
Even on large foundations mounted at grade, structural resonances of the
foundation may be a problem. Forced and induced draft fans, plus some turbine
generator sets actually have foundation resonances that encroach upon the
450
Chapter-9
machinery operating speed domain. It is not uncommon to have a 1,200 CPM

natural resonance on an apparently rigid foundation. Intuition can sometimes be
deceptive, and a large reinforced concrete structure may have a vertical stiffness
of 5,000,000 to 20,000,000 pounds per inch. Combined with the weight of the
foundation, this could yield a natural frequency k m that is very low, and presumably separated from operating speed. However, when the mechanical system
is tested, the intuitive observations may not be supported by the test data.
The field tests performed on a foundation may be quite simple. For example, a deflection profile may be obtained by acquiring vibration measurements at
various elevations on the structure. This type of test is normally performed with
the machinery train operating, and the vibration data filtered at a specific frequency. Normally, this is the machine running speed, and a synchronous
Keyphasor pulse is typically included to establish rotational speed vectors at
each elevation. The use of vectors allows the diagnostician to identify the direction of the vibration component at each elevation. This yields the development of
a definitive mode shape plot of the structure. In most conditions, the maximum
horizontal motion will exist at the top of the structure, and the foundation vibration at grade will be negligible. However, if the vectors change direction from one
elevation to another, this could be indicative of a loose bolted surface or joint
(e.g., loose hold down bolt). In another possibility, if the top end motion is excessive, this could be symptomatic of a support structure with insufficient rigidity
(i.e., low stiffness). On the other hand, if the support structure is still vibrating
at the bottom grade level, the diagnostician should suspect a greater system
problem such as discussed in case history 27.
As another approach, a static impact or a controlled shaker test may be performed on the foundation and associated structure. This test may consist of a
simple frequency response function (FRF) using an accelerometer plus a force
hammer or sledge. Alternatively, it may be a complex modal examination where
the entire structure is covered with an accelerometer array. In this type of test
the input force may be provided by a battering ram equipped with a force transducer or an electro-mechanical shaker. The expected results from this field test
include the natural frequencies of the structure, plus the associated mode
shapes. In some instances, it may be necessary to use the modal parameters from
these modal tests to fine tune a finite element model of the structure to match the
actual behavior. Once this step is completed, the FEA model may then be modified with suitable physical changes to correct the machinery problem in question.
In all cases, the diagnostician must remember that any machinery installation must be considered as a complete mechanical system. The behavior of the
rotating shaft or reciprocating plunger, the bearing oil film characteristics, the
bearing housing support characteristics, plus the baseplate, foundation, and the
load bearing soil characteristics all influence the vibratory behavior. These factors must be considered during the design phase of the machinery, and they must
be re-examined during troubleshooting of the mechanical equipment.
451
Case History 27: Floating Induced Draft Fan

The induced draft fan on a large cracking furnace consisted of a dual inlet,
radial flow fan rotor that weighed 27,000 pounds. This centrifugal fan rotor was
supported between a 226 inch bearing span. The fan was driven by a steam turbine through a speed reducing gear box, and the normal fan operating speed varied between 600 and 780 RPM.
This unit had an uneventful fourteen year operating history that ended one
October evening with a fan inboard journal bearing failure. It was noted that the
high fan vibration before the failure was reduced by installing a new inboard
bearing. Approximately three weeks later, the same bearing failed again, and the
fan rotor plus both bearings were replaced. At this time, the fan was in-place
trim balanced at full operating speed. In December of the same year, the inboard
fan bearing failed again, and the coupling to the bull gear was destroyed. To
repair this damage, the bearing was replaced, a new coupling was installed, and
the fan rotor was subjected to another field trim balance. Unfortunately, the fan
vibration did not remain constant, and two more balance attempts were performed in January of the new year. In essence, a machine that had a good operating history for fourteen years had now experienced three serious mechanical
failures in four months. In addition, the replacement rotor apparently would not
hold a constant balance state.
Although some individuals viewed this behavior as strictly a balancing
problem, it was quite clear that some other malfunction was active. Hence, the
real failure mechanism must be identified and corrected before worrying about
the balancing aspects of this rotor. In an effort to understand the behavior of this
machinery, a complete set of casing vibration measurements were made on the
fan, gear box, and steam turbine driver. At all measurement locations, the predominant frequency component was the fan rotational speed, and the entire
train was preferentially shaking in the horizontal direction. Next a series of
three consecutive startup and shutdown runs were performed, and synchronous
BB
Fig. 951 Horizontal 1X

Running Speed Casing
Vibration Measurements
During Three Separate
Startups On Induced Draft
Fan Bearing Housing
1X Casing Vibration Amplitude (Mils,p-p)
16
J
B
H
14
12
10
Run #1
Run #2
Run #3
J
B
J
4
B H J
B
J
2
BJ BJ HJ HJ
H
H H J BJ H H
J BHJ BHJ BJ BJ BH
0H
6
100
200
300
400
500
600
Shaft Speed (Revolutions/Minute)
700
800
452
Chapter-9
1X fan vibration amplitudes were logged at 50 RPM increments. The horizontal

results of this test are presented in Fig. 9-51.
These measurements were acquired with low frequency velocity pickups,
and the data was manually corrected for the transducer roll-off characteristics.
There were no mechanical changes performed between runs, and a slow roll
period of 40 to 50 minutes was included at the beginning of each run to insure a
minimal fan rotor bow. From this data it was self-evident that the fan vibration
response was quite different during each run. The data from Fig. 9-51 might suggest the possibility of a resonance at 750 RPM, but examination of the corrected
phase angles revealed no evidence of a resonance. At this point, there was no
direct explanation for the variable behavior of this fan rotor.
The blades were checked for soundness and proper attachment to the back
plate and the shrouds. The entire fan impeller assembly was securely attached to
the shaft, and there was no angular shifting of the impeller with respect to the
shaft. The stationary inlet cones were properly positioned, and there was no
visual evidence of any rubs between the inlet cones and the impeller. The pillow
block bearings were inspected for damage plus abnormal fits or clearances. There
were no problems at either fan bearing, and proper clearances were measured.
Furthermore, additional operating tests such as changes in the fan inlet dampers produced minimal changes in the horizontal fan vibration. Overall, the
unusual changes depicted in Fig. 9-51 could not be explained.
Although the fan pedestals appeared to be sufficiently rigid for this type of
machinery, a vibration profile was obtained on the inboard and outboard structures. These horizontal vibration measurements were made at ten different elevations, and the results are presented on Fig. 9-52. The top measurement
elevation (96) was at the centerline of the bearing and shaft. The bottom location (0) was at grade (ground level). The intermediate eight measurements were
distributed between the steel support pedestals, and the concrete foundation.
These horizontal vibration measurements were filtered at the fan running
100
Bearing Centerline Elevation
B
JB J
Fig. 952 Horizontal 1X

Running Speed Casing
Vibration Profile On Both
Induced Draft Fan Bearing
Housings And Support
Pedestal Structures
Height Above Grade (Inches)
90
80
70
B J
J B
J
B
J B
JB
Outboard Pedestal
JB
J B
Grade Elevation
JB
Inboard
Pedestal
60
50
40
30
20
10
0
0
Horizontal 1X Casing Vibration Amplitude (Mils,p-p)
453
speed of 750 RPM, and all readings were directly inphase. Certainly the amplitudes at the bearing housings were higher than desired, but the main peculiarity
of this data set was that the concrete pedestals were vibrating at grade level
(1.5 Mils,p-p). At this location the pedestal motion should have been very close
to zero. It then seemed appropriate to begin digging next to the pedestals and
determine the elevation at which the concrete pedestals were not moving.
At approximately two feet below grade, the pedestals were still moving, and
warm water was encountered. This was certainly an unusual discovery for a
chemical plant in January. Further excavation revealed that there was a small
river flowing underneath the entire machinery train. It was eventually determined that this water flow originated with a broken cooling water return line.
Operations personnel knew that cooling water makeup rates for the past few
months had been much higher than normal, but they had been unable to determine the source of the cooling water loss.
It was also interesting to go back in the records and correlate the fact that
the periods of low fan vibration occurred on nights when the ambient temperature was substantially below freezing. Conversely, the highest vibration amplitudes appeared during the warmest days. It was speculated that on the cold
nights, the subsurface water would freeze to some extent, and provide support
for the fan foundation. During the warmth of the day, the water flow would
increase, which essentially reduced the support stiffness, and allowed the foundation to float and vibrate on the subsurface river.
To correct this problem, the broken section of the cooling water return line
was located and repaired. However, considerable sub-surface water still
remained below the equipment foundation. To remedy this situation, a well-point
pumping network was installed around the entire machinery foundation. Following approximately 36 hours of continuous pumping, the area was pumped reasonably dry. At this stage, the machinery foundation mat was excavated, and it
was tied into the massive reformer exhaust stack foundation with reinforced concrete. The entire area was then back-filled, the soil compacted, and the unit prepared for operation.
Three runs were made to check the repeatability of the fan vibration. Two
of the runs were performed during warm daylight, and the third run was executed shortly after midnight with freezing ambient temperatures. At this time,
virtually identical behavior was noted on both the shaft and the casing vibration
vectors during each run. Finally, a trim balance was performed by welding a 220
gram weight to the fan center plate. This correction reduced the runout compensated shaft vibration to 0.3 Mils,p-p at both bearings, and the horizontal housing
vibration was less than 0.8 Mils,p-p at both bearings. The fan continued to perform in this manner regardless of ambient temperature, and a new era of reliable service was initiated on this critical piece of process machinery.
454
Chapter-9
Case History 28: Structural Influence of Insufficient Grout

In some situations, knowledge of the foundation natural frequency and/or
mode shape may be sufficient to identify the problem, and point to an obvious
solution. For instance, consider the case of an overhung booster compressor that
had a history of high casing vibration. This machine operated at a full load speed
of 6,500 RPM, and it was driven by a two pole induction motor through a double
helical gear box. Interestingly enough, the motor and gear box vibration amplitudes were always quite acceptable, but compressor vibration was much higher
than desirable. During the first twenty months of operation, the compressor had
suffered two seal failures, and one damaged bearing. Certainly, this was unacceptable behavior for a new machinery installation.
In an effort to determine the true cause of the repeated failures, an extensive overhaul and inspection of the compressor was initiated. The compressor
was uncoupled, the case was split, and the rotor, bearings, seals, and coupling
were taken to the shop for detailed examination. During this time period, the
large bore piping, compressor casing, supports, and foundation were checked for
any obvious deficiencies. Externally, the foundation was a solid monolith, and
epoxy grout was used between the baseplate and the concrete foundation. As
part of the inspection, a simple hammer test on the compressor pedestals and the
baseplate below the compressor revealed a distinctly hollow sound. This was
unusual, and it indicated an incomplete grout pour during the initial machinery
installation. The solution to this deficiency resided with filling both pedestals
with grout, plus injecting grout between the baseplate and the foundation.
Before implementing this physical correction, it was considered desirable to
measure the major mode shapes and natural resonances of this compressor support structure. In previous years, this would have been a complicated project
requiring two or more days for setup and testing. However, the use of modern
instrumentation plus suitable software reduced this work to a manageable exercise. In all fairness, it should also be mentioned that approximately three hours
was devoted to establishing the three-dimensional matrix structure, and proper
polarity directions for the tri-axial accelerometer. This setup work was performed offsite, and the actual field test of this structure was performed in less
than two hours using an HP-35670 four channel Dynamic Signal Analyzer, plus
an impact hammer, and a tri-axial accelerometer.
In this application, the DSA was controlled with Hammer-3D software by
David Forrest11. This HP I-Basic program runs directly on the HP-35670A, and
it eliminates the need for external devices such as a separate computer system.
Within Hammer-3D, the test structure geometry and transducer array are
defined. The field work consists of acquiring frequency response functions (FRF)
on the structure. For this type of test, the impact hammer location remained constant, and the triaxial accelerometer was moved between measurement points. A
total of 24 points were identified on this compressor support assembly, and 72 (=
3 x 24) FRFs were acquired and stored on floppy disk. From a physical dimension
11 David Forrest, Hammer-3D Version 2.01, Computer Program in Hewlett Packard Instrument Basic by Seattle Sound and Vibration, inc., Seattle, Washington, 1997.
455
standpoint, the pedestals were nominally 36 inches tall, 36 inches in length, and
8 inches wide at the top. The associated baseplate below the pedestals was 36
inches wide, and 58 inches across (pedestal to pedestal). These physical dimensions resulted in a nominal 18 to 20 inch spacing between the 24 measurement
points.
After acquisition of the field data, curve fitting was applied to each user
defined resonant frequency range on each FRF. The individual modes were then
assembled, scaled, and presented as animated mode shapes on the DSA. These
are quite realistic modes due to the fact that tri-axial measurements were
obtained, and cross-coupling between orthogonal directions is simultaneously
displayed. It should be mentioned that this is a complicated procedure, and a
manual solution is generally unattainable. Hence, the computational power
within the DSA is mandatory for proper and rapid presentation of results.
The animated display on the HP-35670A is quite descriptive, and physically understandable. Unfortunately, reproduction of this dynamic display to a
static diagram is often difficult, and proper interpretation may be lost by the
examination of a single diagram. Hence, it is generally desirable to examine the
minimum and maximum deflections with respect to the stationary structure.
tal
es
d
Pe
rth 16
o
N
sta
18
h P16
rt
No
e
ed
14
Dotted Line
Stationary
Shape
17
15
14
18
Solid Line
Deflected
Mode Shape
17
15
24
13
9
5
10
11
8
4
uth
23
20
12
10
tal
es
d
Pe
So
Fig. 953 Baseplate Deflection Mode At

3,530 CPM - Maximum Negative Deflection
3
Dotted Line
Stationary
Shape
11
21
19
7
Solid Line
Deflected
Mode Shape
22
21
6
9
5
23
20
24
13
22
12
19
l
sta
de
e
P
uth
So
8
Fig. 954 Baseplate Deflection Mode At

3,530 CPM - Maximum Positive Deflection
For instance, consider the measured deflection mode shapes presented in

Figs. 9-53 and 9-54. Both of these diagrams are associated with a resonant mode
at 3,530 CPM. The dotted lines in both drawings depict the stationary structure
of the baseplate, plus the north and south compressor support pedestals. The
solid lines represent the deflected mode shape of the structure. In Fig. 9-53, the
maximum negative deflection is shown, and the companion diagram in Fig. 9-54
displays the maximum positive deflection at 3,530 CPM. Based on these two diagrams, it is clear that a drum mode exists on the compressor baseplate. This is
456
Chapter-9
most evident by looking at the relative positions of the center points 6 and 7 in
both figures. Surprisingly, the measured frequency for this drum mode occurs at
3,530 CPM, which is uncomfortably close to the normal motor operating speed
range of 3,580 to 3,595 RPM. Although both compressors pedestals display minimal motion at this frequency, there still remains an undesirable coincidence
between this baseplate natural frequency and the motor running speed.
The second dominant mode measured on this support structure occurs at a
frequency of 6,600 CPM. The maximum negative and positive mode shapes at
this frequency are presented on Figs. 9-55 and 9-56 respectively. Within this
tal
es
rth
d
Pe
d
Pe
No
No
Dotted Line
Stationary
Shape
17
15
14
22
13
5
1
12
19
8
uth
23
20
11
21
22
10
11
7
2
24
9
23
20
15
13
10
Dotted Line
Stationary
Shape
17
24
9
Solid Line
Deflected
Mode Shape
18
rth 16
16
14
tal
es
18
tal
e
ed
So
Fig. 955 Pedestal Deflection Mode At

6,600 CPM - Maximum Negative Deflection
21
7
2
Solid Line
Deflected
Mode Shape
12
19
3
8
4
uth
sta
e
ed
So
Fig. 956 Pedestal Deflection Mode At

6,600 CPM - Maximum Positive Deflection
mode, the baseplate motion is reduced, but the north pedestal motion is excessive. This is demonstrated by comparing the differential position of point 15 in
Figs. 9-55 and 9-56. The motion at the south pedestal is substantially less that
the measured deflection of the north pedestal. This is interpreted as a better
grout injection in the south versus the north pedestal. Amazingly enough, this
mode at 6,600 CPM is close to the compressor operating speed range of 6,500 to
6,530 RPM. Again, this is an undesirable condition for this compressor support.
In the overview, it is clear that these support structure mode shapes are
detrimental to the typical requirement for a solid compressor support. The close
proximity between these natural frequencies and the excitations due to the
machine operating speed range is likewise objectionable. It is quite unusual to
encounter a situation where two distinct structural resonances are close to two
fundamental machine excitations. However, due to the inconsistencies of the
original grout pour, any combination of baseplate and pedestal natural resonant
frequencies are physically possible. In this case, the resultant natural frequencies coincided with the machinery operating speeds.
It is concluded that the flexibility of the compressor pedestals and the base-
457
plate are excessive and unacceptable. From a correction standpoint, the previously mentioned solution of filling these voids with epoxy grout remains as the
most prudent course of action.
For comparative purposes, after the grout injection was completed, and the
compressor reinstalled, a substantial reduction in casing vibration amplitudes
was apparent. Historically, unfiltered casing velocity levels varied between 0.18
and 0.22 IPS,o-p on the compressor bearing housing. After the compressor
rebuild, and the grout repairs, the maximum casing vibration was 0.03 IPS,o-p.
The majority of this reduction was due to attenuation of the motor and the compressor rotational speed vibration components.
Prior to this repair, the radial shaft vibration amplitudes on the compressor
ranged from 2.2 to 3.0 Mils,p-p. After the rebuild and grout repair, all of the centrifugal compressor shaft vibration levels dropped below 1.0 Mil,p-p. The success
of this simple repair was effectively demonstrated by successful and continuous
operation. There were no additional mechanical failures on this machinery train,
and vibration amplitudes remained low and constant.
In retrospect, this type of problem is generally encountered on new construction projects. If the field inspectors do not perform a surface hammer test
after the grout is cured, this type of flaw may go undetected. This could easily
result in machinery problems that persist for an extended period of time. Certainly, the initial diagnosis and the physical solution to this insufficient grout
problem makes sense to most people. However, some individuals will question
the need for performing the structural mode shape measurements discussed in
this case history. If these structural tests had not been performed, the solution
(additional grout), and the beneficial results (extended run time without failures) would have been the same. However, there would still be an uncertainty as
to the root cause of the previous failures. In addition, the potential benefits associated with spending money for injecting epoxy grout to fill in the voids might be
hard to sell to management. With the rapid availability of this structural mode
shape and frequency data, the cause and effect relationship is understandable,
and the repair costs are much easier to justify. Stated in another way, the availability of this information represents the difference between a shotgun approach,
and a properly engineered explanation.
458
Chapter-9
BIBLIOGRAPHY
1. Allaire, P.E., and R.D. Flack, Design Of Journal Bearings For Rotating Machinery,
Proceedings of the Tenth Turbomachinery Symposium, Gas Turbine Laboratories,
Texas A&M University, College Station, Texas (December 1981), pp. 25-45.
2. Bently, Donald E., Forced Subrotative Speed Dynamic Action of Rotating Machinery, American Society of Mechanical Engineers, ASME Paper No. 74-Pet-16 (1974).
3. Bently, Donald E., Forward Subrotative Speed Resonance Action of Rotating
Machinery, Proceedings of the Fourth Turbomachinery Symposium, Gas Turbine
Laboratories, Texas A&M University, College Station, Texas (October 1975), pp.
103-113.
4. Bently, Donald E., Attitude Angle And The Newton Dogleg Law Of Rotating
Machinery Bently Nevada Applications Note (March 1977).
5. Bently, Donald E., Vibration levels of machinery, Orbit, Vol. 13, No. 3 (September
1992), p. 4.
6. Bently, Donald E., and Agnes Muszynska, Detection of Rotor Cracks, Proceedings
of the Fifteenth Turbomachinery Symposium, Turbomachinery Laboratory, Texas
A&M University, College Station, Texas (November 1986), pp. 129-139.
7. East, John R., Turbomachinery Balancing Considerations, Proceedings of the
Twentieth Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M
University, College Station, Texas (September 1991), pp. 209-214.
8. Forrest, David Hammer-3D Version 2.01, Computer Program in Hewlett Packard
Instrument Basic by Seattle Sound and Vibration, inc., Seattle, Washington, 1997.
9. Gunter, Jr., Edgar J., Rotor Bearing Stability, Proceedings of the First Turbomachinery Symposium, Gas Turbine Laboratories, Texas A&M University, College Station, Texas (1972), pp. 119-141.
Unique Behavior
10
10
he common machinery malfunctions discussed in chapter 9 occur on a wide variety of machines. The typical frequencies
observed with those common malfunctions generally occur between one quarter
of rotative speed and twice running speed. Many process machines are subjected
to additional excitations that impose significant dynamic loads upon the machinery at other frequencies. In chapter 10, the excitations produced within two element and epicyclic gear boxes will be discussed. Common fluid excitations and
electrical phenomena will also be examined. Finally, the application of rotating
machinery technology to reciprocating compressors will be reviewed. As usual,
each of these topics will be highlighted with numeric examples, and actual
machinery case histories.
PARALLEL SHAFT - TWO ELEMENT GEAR BOXES

Speed increasing, or speed reducing gear boxes are devices that emit a distinctive set of excitations. Gear box elements move with definable static position
changes, and they generate specific frequencies that may be used for mechanical
diagnosis. Due to the vast array of gear box configurations, the current discussion will concentrate on the common two element, parallel shaft, single or double
helical gears used within the process industries. A review of the complex excitations generated by epicyclic gears is included in following section of this chapter.
Gear boxes are complicated machines that have evolved from slow speed water
wheels to a vast array of industrial machines. In many respects, gear design,
configuration, fabrication, and application is a science unto itself. Due to the
complexity of this subject, the reader is encouraged to examine books by authors
such as Lester Alban1, Darle Dudley2, and M.F. Spotts3 that go into specific
details regarding the mechanics of various types of gear boxes. There are also
numerous standards, handbooks, and design guides available from the American
1 Lester E. Alban, Systematic Analysis of Gear Failures, (Metals Park, Ohio: American Society
for Metals, 1985).
2 Darle W. Dudley, Gear Handbook, (New York: McGraw-Hill Book Company, 1962).
3 M.F. Spotts, Design of Machine Elements, 6th Edition, (Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1985).
459
460
Chapter-10
Gear Manufacturers Association (AGMA) in Arlington, Virginia. The gear OEMs

also produce some excellent technical references on all aspects of gearing.
The discussion contained in this text is divided into a review of static element shifts, the computation of the major gear contact forces, and the dynamic
vibratory characteristics. The position changes or static shifts of gear elements
are dependent upon rotation, the driver element, and the applied forces. It is
often difficult to maintain a clear perspective of the gear force directions under
normal conditions. This issue is clouded by the various potential variations in
gear set arrangements. Hence, it is reasonable to examine the expected types of
forces and their directions in both the radial and the axial planes.
CCW
CW
Output
CCW
CW
Input
Output
Input
Fig. 101 Expected Radial Loading Of Down Mesh Gear Sets - View Towards Input
In order to describe the radial position characteristics of two element gear

sets Figs. 10-1 and 10-2 have been constructed. The sketches in Fig. 10-1
describe the expected radial load directions for a down mesh set. The left hand
diagram depicts a speed decreasing box where the pinion is the input element,
and the bull gear is the reduced speed output. The right hand sketch in Fig. 10-1
describes a speed increasing box where the bull gear is the input, and the pinion
is the high speed output. The heavy arrows on each sketch describe the general
load direction of the overall forces acting at each respective bearing.
The two sketches presented in Fig. 10-2 depict the expected radial forces for
an up mesh gear box. The left hand diagram describes a speed decreasing unit
where the bull gear is the input element, and the pinion is the increased speed
output. The right hand sketch in Fig. 10-2 shows a speed decreasing gear box
where the pinion is the input, and the bull gear provides the slow speed output.
These simple diagrams define the general load directions for each gear eleCW
CCW
Input
CW
CCW
Output
Output
Input
Fig. 102 Expected Radial Loading Of Up Mesh Gear Sets - View Towards Input
Parallel Shaft - Two Element Gear Boxes
461
ment. This is important for the proper location of bearing thermocouples, pressure dams, and the evaluation of radial shaft position shifts as measured by
proximity probe DC gap voltages. Many OEMs now provide analytical calculations that predict the vertical and horizontal journal centerline position at full
load. These radial positions are unique to each gear box design and should be
compared with actual shaft centerline position shifts. Note, this measurement is
often difficult to execute due to the tooth engagement between gears at the rest
position. Hence, an accurate zero speed starting point (particularly for the pinion) may be difficult to obtain.
Any gear box evaluation should always include a detailed examination of
the operating shaft positions at each of the four radial bearings. When available,
the measured position should be compared with the radial location calculated by
the OEM. An incorrectly positioned bearing will cause significant distress within
the gear box. Unless the journal locations are checked for proper running position, the diagnostician may end up chasing a variety of abnormal dynamic characteristics when the real problem is easily identified by the radial journal
position data. This also reinforces the argument for installing X-Y radial proximity probes at all gear box bearings. Many facilities tend to install proximity
probes only at the input and output bearings, and they often ignore the blind or
outboard end bearings. This practice can result in the unavailability of some critical journal position and vibration information.
On a helical gear, the gear tooth contact force is typically resolved into three
mutually perpendicular forces. The two radial forces consist of a tangential and a
separation force. The tangential force is based upon the transmitted torque and
the pitch radius. Calculation of the torque is determined in equation (10-1), followed by the tangential force in equation (10-2):
where: Torque
HP
RPM
ForceTan
Rpitch
33, 000 HP
5, 252 HP
Torque = --------------------------------- = -----------------------------2 RPM
RPM
(10-1)
63, 024 HP
12 Torque
Force Tan = --------------------------------- = --------------------------------------R pitch RPM
R pitch
(10-2)
=
=
=
=
=
Transmitted Torque Across Gear Teeth (Foot-Pounds)

Transmitted Power Across Gear Teeth (Horsepower)
Gear Element Rotational Speed (Revolutions / Minute)
Tangential Force Across Gear Teeth (Pounds)
Pitch Radius of Gear Element (Inches)
In these expressions the speed and pitch radius must be for the same gear
element. That is, if the bull gear speed is used to compute the torque, then the
bull gear pitch radius must be used to determine the tangential force. Similarly,
if the pinion speed is used to calculate the transmitted torque, then the pinion
pitch radius must be used to compute the correct tangential force. Note that the
transmitted torque is different for the pinion and the bull gear, but the tangential force for both elements must be the same. As another check, the pitch line
velocity for both gear elements must also be identical.
462
Chapter-10
The tangential force is the vertical force acting between the gears. Obviously, one gear element is subjected to an upward tangential force, and the mating gear element is subjected to a downward tangential force (necessary to be
equal and opposite). Based upon the gear pressure angle, and the helix angle, the
gear separation factor may be computed as in equation (10-3). Multiplying the
previously calculated tangential force by this non-dimensional gear separation
factor provides the gear separation force as shown in equation (10-4):
where:
tan
SF = -------------cos
(10-3)
tan
Force sep = Force Tan SF = Force Tan -------------cos
(10-4)
SF =
=
=
Forcesep =
Separation Factor (Non-Dimensional)

Pressure Angle Measured Perpendicular to the Gear Tooth (Degrees)
Helix Angle Measured from the Gear Axis (Degrees)
Separation Force Between Gears (Pounds)
This separation force acts to the right on one gear, and to the left on the
mating gear element. Again, a force balance must be achieved in the horizontal
plane, and the separation force must be less than the tangential force. For standard gears, the typical pressure angle is either 14.5, 20, or 25. The most
common value encountered for the pressure angle is 20. The helix angle typically varies between 15 and 35. Although these angles are similar, it is mandatory for the diagnostician to keep the numbers straight. Finally, the third
segment of the overall gear contact force is the axial component. The magnitude
of this thrust load is obtained from the following expression:
Force Thr = Force Tan tan ( )
(10-5)
where: ForceThr = Axial (Thrust) Force Between Gears (Pounds)
As a side note, if the helix angle is 0, the helical gear equations simplify
into spur gear equations. That is, the cosine of 0 is equal to 1, and the separation force is equal to the tangential force times the tangent of the pressure angle.
Also, the tangent of 0 is equal to zero, and the thrust load is zero. Obviously,
spur gears cannot transmit an axial force.
The axial or thrust loads on a double helical (herringbone) gear are theoretically balanced by the two sides of the gear. If the gear is machined incorrectly,
an axial force will occur on a double helical gear, and this may generate significant axial loads. However, on a single helical gear box, the thrust loads are
always present. These axial forces must be accommodated by thrust bearings for
each element of a single helical gear set. It is meaningful to understand the normal versus the counter thrust directions for a single helical gear. This helps in
setting up the thrust monitors properly (i.e., normal versus counter), and it
allows a proper evaluation of measured thrust behavior.
463
Thrust
Thrust
Left
Hand
Right
Hand
Thrust
Output
Right
Hand
Left
Hand
Thrust
Input
Output
Input
CW
CCW
CW
CCW
Fig. 103 Normal Thrust Direction For Single Helical Down Mesh Gears
The two diagrams presented in Fig. 10-3 describe the normal thrust directions for a single helical gear box equipped with down mesh gears. The drawings
in Fig. 10-4 depict the thrust directions for up mesh gears. In each case, the
gears are identified as either right-hand or left-hand. This is a common designation of how the teeth curve away from the mesh line. If the teeth lean or are
inclined to the right or the clockwise direction, the element is referred to as a
right-hand gear. Conversely, if the teeth lean or are inclined to the left, or in a
counterclockwise direction, the element is identified as a left-hand gear. In any
pair of mating helical gears, one element must be right-handed and the other
gear element must always be left-handed.
Thrust
Left
Hand
Right
Hand
Thrust
Right
Hand
Left
Hand
Thrust
Output
Thrust
Input
Output
Input
CCW
CW
CCW
CW
Fig. 104 Normal Thrust Direction For Single Helical Up Mesh Gears
As mentioned earlier in this section, gear boxes emit many unique excitations. The actual excitations vary from low to high frequencies. For example, a
typical parallel shaft, two element (bull gear and pinion) gear box, will normally
produce the following group of discrete frequencies.
464
Chapter-10
Bull Gear Rotational Speed Fbull

Pinion Rotational Speed Fpin
Gear Mesh Frequency Fgm
Assembly Phase Passage Frequency Fapp
Tooth Repeat Frequency Ftr
Gear Element Resonant Frequencies
Casing Resonant Frequencies
The bull gear and pinion rotational speeds are the actual speeds of each element. These rotative speeds maintain a fixed ratio that is completely dependent
on the number of bull gear and pinion teeth. If at all possible, the diagnostician
should obtain an exact tooth count on both elements. This must be an exact number (0 allowable error). The gear mesh frequency is equal to the speed times the
number of teeth as shown in equation (10-6):
F gm = F bull T bull = F pin T pin
where:
Fgm
Fbull
Fpin
Tbull
Tpin
=
=
=
=
=
(10-6)
Gear Mesh Frequency (Cycles / Minute)

Rotational Speed of Bull Gear (Revolutions / Minute)
Rotational Speed of Pinion (Revolutions / Minute)
Number of Teeth on Bull Gear
Number of Teeth on Pinion
The gear mesh frequency must be the same for both the bull gear and the
pinion. This commonality also provides a good means to verify the validity of a
presumed gear mesh frequency in an FFT plot. In other situations, if the bull
gear speed is known, the pinion speed may be determined from (10-7) when the
actual number of bull gear and pinion teeth are known. Obviously, the inverse
relationship is also applicable.
T bull
F pin = F bull -------------T pin
(10-7)
The gear mesh frequency provides general information concerning the gear
contact activity and forces. This type of measurement is usually obtained with a
high frequency casing mounted accelerometer. Typically, the best data is
acquired at the gear box bearing housings, since this is the location where the
meshing forces are transmitted to ground. As an example of this type of information, the data presented in case history 19 should be of interest.
The next two excitations of assembly phase passage frequency and tooth
repeat frequency require an understanding of the concept of phase of assembly.
This is clearly explained by John Winterton4 as follows:
Mathematically, the number of unique assembly phases (Na) in a given
tooth combination is equal to the product of the prime factors common to the num4 John G. Winterton, Component identification of gear-generated spectra, Orbit, Vol. 12, No. 2
(June 1991), pp. 11-14.
465
ber of teeth in the gear and the pinion. The numbers 15 and 9 have the common
prime factor of 3. Therefore, three assembly phases exist. The number of assembly
phases determines the distribution of wear between the teeth of the gear and pinion
Winterton goes on to define the assembly phase passage frequency as
shown in equation (10-8):
F gm
F app = -----------Na
where:
(10-8)
Fapp = Assembly Phase Passage Frequency (Cycles / Minute)

Na = Number of Assembly Phases (Non-Dimensional Prime Number)
This is followed by the determination of the tooth repeat frequency. This is

generally the lowest level excitation within the gear box. Typically it falls below
500 CPM, and sometimes it appears as an amplitude modulation. The tooth
repeat frequency may be computed in the following manner:
F gm N a
F bull N a
F pin N a
F tr = --------------------------------- = ----------------------------- = --------------------------T bull T pin
T pin
T bull
where:
(10-9)
Ftr = Tooth Repeat Frequency (Cycles / Minute)
Other investigators have a tendency to consider only a true tooth hunting

combination where the number of assembly phases Na is equal to one. This is
proper and normal for precision high speed gears. However, equation (10-9) represents the correct computation for all cases.
The natural frequencies of the bull gear and pinion are often located above
the normal operating speed range. Thus, examination of the transient Bode plots
will generally reveal a stiff shaft response of both gear elements. However, gears
do exhibit natural frequencies that generally appear at frequencies above rotational speed. Typically, a bull gear and a pinion will each display a stiff shaft
translational followed by a pivotal mode that are both governed by bearing stiffness. At higher frequencies, the shaft stiffness controls the resultant natural frequencies. In a simplistic model, these are free-free modes that are dependent on
mass and stiffness distribution across each respective gear element. These
higher order modes are independent of bearing stiffness, and they are often
excited during failure conditions. Typically, discrete frequencies between 60,000
and 180,000 CPM (1,000 and 3,000 Hz) are detectable. These gear element resonant frequencies are often frequency modulated by the running speed of the
problem element. For instance, if the bull gear is under distress, a higher level
bull gear natural frequency will be modulated by bull gear speed.
It should also be mentioned that most gear boxes display a variety of casing
resonant frequencies. The distribution of these frequencies will depend on the
gear casing construction. A fabricated box will be lighter than an older cast box.
In general, the thinner gear box casings will exhibit higher natural frequencies
466
Chapter-10
than heavier and thicker wall casings. Typically, an industrial helical gear box
may exhibit multiple casing natural frequencies, and they may appear anywhere
between 30,000 to 300,000 CPM (500 to 5,000 Hz). Various attachments to the
gear box may also appear as narrow band structural resonances. Items such as
unsupported conduit, thermowells, small bore piping, and proximity probe holders may be detectable on the gear box. In one case, long unsupported stingers
were used on proximity probes in a large gear box installation. Unfortunately,
the natural resonance of the probe stingers was 3,580 CPM, which was excited
by the synchronous machine speed of 3,600 RPM.
Parallel shaft gear boxes are also built with multiple gear elements. For
example, an intermediate idler gear may be installed between a bull gear and a
pinion to obtain a specific speed ratio, or maintain a particular direction of rotation. Some gear boxes contain multiple gears, such as the seven element box discussed in case history 21 in chapter 9. These additional gear elements provide
additional rotational speed excitations. If the unit contains direct mesh to mesh
contact across the box, the gear mesh frequency will remain constant. However,
if the gear box contains any variety of stacked gear arrangements, the unit will
emit multiple gear mesh frequencies. These multiple rotational speeds and gear
mesh frequencies will often interact in a variety of signal summations, amplitude modulations, and frequency modulations.
Interactions of the multiple frequencies will depend on load, which influences the journal radial positions and the tooth contact between gears. In all
cases, the documentation of vibratory data with the box in good condition will be
beneficial towards analysis of a variety of potential future malfunctions.
Case History 29: Herringbone Gear Box Tooth Failure
The speed increasing gear box shown is Fig. 10-5 is situated between a low
pressure and a high pressure compressor. The normal operating speed for the LP
compressor and the bull gear is 4,950 RPM, and the pinion output to the HP
compressor runs at 11,585 RPM. A flexible diaphragm coupling is installed
between the LP compressor and the bull gear, and another diaphragm coupling
is used between the pinion and the HP compressor. The axial stiffness for both
couplings are approximately equal, and the gear box has successfully operated in
this configuration for many years.
The gear configuration consists of a double helical, or herringbone, arrangement. This type of gear provides a generally balanced axial load between the two
sets of gear teeth. Some axial load is inevitable, and a thrust bearing is mounted
on the outboard end of the bull gear. Due to the dual mesh interaction, the pinion
must follow the bull gear axial position, and a separate pinion thrust bearing is
not required. If this unit was a single helical gear, a separate pinion thrust bearing would have been incorporated.
The tooth failure problem on this unit was initiated during a routine topping off of the oil reservoir. For whatever reason, excess oil was pumped into the
reservoir, and oil backed up the gear box drain line. This reverse flow filled the
gear box with lubricant, and oil began spewing from the gear box atmospheric
467
Pinion Output at 11,585 RPM

47 Pinion
Teeth
110 Bull
Gear Teeth
Thrust
Collar
Fig. 105 General Arrangement Of Two Element Gear Box With Herringbone Gears
Bull Gear Input at 4,950 RPM
vents. This external oil flow was ignited by a hot steam line, and a fire ensued.
The machinery train was tripped and the deluge system activated. These combined actions extinguished the fire with minimal external damage to the machinery. Unfortunately, the train was restarted to a fast slow roll, and allowed to run
at 1,500 RPM for approximately 90 minutes. Evidently the gear box was still
filled with oil during this abbreviated test run.
Following evaluation of the vibration, mechanical, and process data, the
train was shutdown for gear box disassembly and inspection. Upon removal of
the top half, it was visibly noted that the gears were in good condition. Following
removal of both gear elements it was clear that all four journal bearings were
damaged. The inside sections of the four journal bearings that are exposed to the
interior of the gear box had melted babbitt, whereas the outside sections of all
four bearings retained babbitt. This indicates that the internal gear box temperature was in excess of 500F to melt part of the bearing babbitt.
Shop examination of the bull gear and pinion revealed that both elements
were coated with varnish. This was indicative of burnt or oxidized oil on the surface of the gears. After the varnish was removed, the gears appeared to be in
good physical condition with minimal surface wear on the teeth. A dye penetrant
inspection did not reveal any cracks or discontinuities, and the shaft journals
were considered to be in good condition.
Since the shop inspections revealed no evidence of physical damage to the
gear set, the gear elements were reinstalled in the box. Although axial spacing
between the bull gear shaft and the LP compressor shaft was maintained, the
bull gear coupling hub was mounted 0.25 further on to the shaft than previous
installations. The effect of an axially mis-positioned hub on a diaphragm coupling would be the generation of an axial preload on the bull gear. This axial load
could force one side of the herringbone gear to carry the majority of the load.
The train was successfully restarted, and machinery behavior appeared to
468
Chapter-10
be normal, and consistent with previous vibration data. After six days, a high
frequency vibration component around 75,600 CPM (1,260 Hz) was noticed on
the gear box. This frequency was approximately 15 times the bull gear speed,
and it was intermittently transmitted to both compressors. The unit operated in
this manner for approximately one month when a leaking thermowell forced the
train down to slow roll speeds for 45 minutes for thermowell replacement. During the subsequent restart, two compressor surges occurred, and this event may
have overloaded the gear teeth.
Two days after the restart, shaft vibration amplitudes experienced a series
of minor step changes in a gradually increasing trend. The bull gear radial shaft
vibration data revealed minor 1X vector changes. However, the largest change
occurred on the bull gear axial probes, where the synchronous 1X motion
increased from 0.48 to 1.29 Mils,p-p. This is certainly abnormal behavior for a
double helical gear with normally balanced axial forces.
Simultaneously, the casing vibration amplitudes began to grow in the vicinity of 75,600 CPM (1,260 Hz) with peak levels reaching 12.0 Gs,o-p. It should also
be noted that the previously dormant gear mesh frequency at 544,500 CPM
(9,075 Hz) had blossomed into existence, and it was modulated by bull gear rotational speed. Furthermore, audible noise around the gear box had significantly
increased. FFT analysis of microphone data recorded on the compressor deck
revealed a dominant component at 1,238 Hz (15th harmonic), with sideband
modulation at bull gear rotational speed of 4,950 RPM (82.5 Hz).
Considering the available information, the bull gear distress was self-evident, and a controlled shutdown was the only reasonable course of action. Following an orderly shutdown, a visual inspection revealed 12 broken teeth on the
coupling side of the bull gear. Additional shop inspection revealed multiple
cracked teeth on the bull gear combined with an erratic and accelerated wear
pattern on both gear elements.
In retrospect, the gears were probably solution annealed during the period
of high internal gear box temperatures. A micro-hardness survey revealed that
the broken gear teeth had a Rockwell C surface hardness of 20 for the first 2 Mils
(0.002 inches) of tooth surface. The hardness then increased with depth to levels
consistent with the original gear tooth heat treating. Normal surface hardness
for these gears should be 37 on the Rockwell C scale. This softening of the gear
teeth surfaces represents the root cause of this failure. However, the tooth failure
was logically due to a combination of the following events:
1.
2.
3.
4.
Initial fire, and the probable surface annealing of the gear teeth.
Potential axial load imposed by a mis-positioned bull gear coupling hub.
Potential impact loads suffered during compressor surges.
Probable high cycle fatigue of the heavily loaded and soft gear teeth.
In all likelihood, the primary damage of softening the gear teeth occurred
during the fire. Based upon the metallurgical findings, it is clear that this gear
was destined for premature failure. The actual influence of items 2 and 3 in the
above list are difficult to quantify. In all probability, these contributors acceler-
469
ated the failure, but the life span of the bull gear teeth was greatly reduced by
the loss of tooth surface hardness.
The symptoms of this failure included minor changes in the shaft radial
vibration, significant changes in the bull gear axial vibration, plus substantially
increased activity at the gear mesh frequency. These conditions are fully explainable based upon the physical evidence of broken gear teeth. However, the casing
excitation at 75,600 CPM (1,260 Hz), and the dominant sound emitted by the
gear box at the same general frequency are not immediately obvious.
In an effort to understand the significance of this frequency component, a
simple impact test was performed on the failed bull gear resting in the journal
bearings (with the pinion removed). The main component encountered during
this test occurred at a frequency of 76,800 CPM (1,280 Hz). Clearly, this is quite
close to the frequency identified during the failure, and it could be a resonant frequency of the bull gear assembly.
Since additional testing on the bull gear was not a viable option, a 28 station undamped critical speed model for the gear element was developed. The calculated first two modes include a stiff shaft translational response at 8,500 RPM,
followed by a stiff shaft pivotal mode at 10,250 RPM. Both resonances have
greater than 93% of the strain energy in the bearings, with less than 7% of the
strain energy in the shaft. Hence, these first two modes would be influenced by
changes in bearing stiffness and damping characteristics. The calculated higher
order modes are bending modes of the bull gear rotor. These resonances are completely dependent on shaft stiffness (i.e., bearing stiffness is inconsequential).
These higher order resonances were computed for a planar condition of zero
speed. This is equivalent to the bull gear sitting at rest in the gear box bearings
without a mating pinion. This simplified analysis considers the case of a stationary element without external forces or rotational inertia.
Of particular interest in this simplified analysis was the appearance of a
free-free mode at a frequency of 73,200 CPM (1,220 Hz). This frequency is close to
the 74,280 CPM (1,238 Hz) detected during operation, and the 76,800 CPM
(1,280 Hz) measured the stationary impact test on the failed gear. Additional
examination of this relationship, and further refinement of the analytical model
would be academically interesting. However, this is not a cost-effective exercise,
and it is necessary to draw a logical conclusion based upon the available information. In this case, it is reasonable to conclude that the frequency in the vicinity of
75,000 CPM (1,250 Hz) is a natural resonance of the bull gear assembly. This frequency appears during the static impact tests, and it is also excited during operation with failed gear teeth. In this condition, the rotating bull gear is
periodically subjected to multiple impacts due to the absence of various gear
tooth. It is postulated that these impacts during operation excite this bull gear
resonance.
470
Chapter-10
EPICYCLIC GEAR BOXES

Other gear boxes are even more complex due to the internal configuration
of the gear elements. One of the most complicated industrial gear boxes is the
epicyclic gear train. In these units, a moving axis allows one or more gears to
orbit about the central axis of the train. Simple epicyclic gear boxes contain a
central sun gear that meshes with several planet gears that are evenly spaced
around the sun gear. Both the sun and planet gears are externally toothed spur
gears. The planet gears also mesh with an internally toothed ring gear. The input
and output shafts are coaxial. These shafts may rotate in the same direction, and
they may rotate in opposite directions. This is dependent on the actual mechanical configuration of the individual gear box.
Epicyclic gear boxes derive their name from the fact that the planet gears
produce epicycloidal curves during rotation. In actuality, there are three general
types of epicyclic gear boxes. Perhaps the most common type is the planetary
arrangement that consists of a stationary ring gear combined with a rotating sun
gear, and moving planet carrier. The star configuration of a epicyclic gear box
consists of a stationary planet carrier coupled with a rotating sun gear, and a
rotating outer ring gear. The third type, and probably the least common type of
arrangement, is the solar gear. This low ratio epicyclic box has a fixed sun gear
combined with a moving ring gear, and planet carrier.
Before addressing any specific details on these three epicyclic gear arrangements, it would be beneficial to mention the common characteristics between the
three configurations. For instance, on a sun gear input, the tangential tooth load
at each planet is derived by an expansion of equation (10-2) into the following:
63, 024 HP
Force tan plt = -----------------------------------------------------------------R pitch sun F sun N p
where:
HP
Rpitch-sun
Fsun
Np
Forcetan-plt
=
=
=
=
=
(10-10)
Transmitted Power Across Sun Gear Teeth (Horsepower)

Pitch Radius of Sun Gear (Inches)
Sun Gear Rotational Speed (Revolutions / Minute)
Number of Planet Gears (Dimensionless)
Tangential Force Across Planet Gear Teeth (Pounds)
The number of external teeth on the sun and planet gears, and the number
of internal teeth on the stationary ring gear must maintain a particular tooth
ratio to allow assembly. Specifically, the following tooth assembly equations were
extracted from Dudleys Gear Handbook5:
5 Darle
T ring = T sun + 2 T plt
(10-11)
T ring + T sun
---------------------------------- = Integer
Np
(10-12)
W. Dudley, Gear Handbook, (New York: McGraw-Hill Book Company, 1962), pp. 3-15.
Epicyclic Gear Boxes

where:
471
Tsun = Number of Sun Gear Teeth (Dimensionless)

Tplt = Number of Planet Gear Teeth (Dimensionless)
Tring = Number of Ring Gear Teeth (Dimensionless)
With three different gear configurations, epicyclic boxes may emit a variety
of excitations that vary from low to high frequencies. For example, a generic epicyclic box has the potential to produce the following array of frequencies.
Sun Gear Rotational Speed Fsun

Planet Gear Rotational Speed Fplt
Planet Carrier Rotational Speed Fcar
Ring Gear Rotational Speed Fring
Planet Pass Frequency Fplt-pass
Planet Absolute Frequency Fplt-abs
One or More Gear Mesh Frequencies Fgm
Gear Element Resonant Frequencies
Casing Resonant Frequencies
For a planetary gear box the ring gear speed Fring is equal to zero. On a star
configuration the planet carrier is fixed, and frequency Fcar is zero. Similarly, the
sun gear speed Fsun is zero on a solar gear. The individual gear mesh frequencies
are a bit more complicated, and they will be reviewed in conjunction with each
gear box discussion. The gear element natural resonant frequencies, and the casing resonant frequencies listed in the previous summary, originate from the
same sources discussed under two element gear boxes.
At this point, it is meaningful to examine the specific frequencies associated
with a planetary gear box. For instance, consider the typical planetary gear
train shown in Fig. 10-6. This is a basic arrangement that may be used as either
a speed increasing or a speed decreasing device. For discussion purposes, assume
that this gear box is used as a speed increaser. The input shaft is coupled to the
planet carrier, and it rotates counterclockwise at a frequency indicated by Fcar.
In this example, three planet gears are attached to the carrier, and each planet
mates with the stationary ring gear (Fring=0). As the planet carrier rotates in a
counterclockwise direction, the planet gears must turn clockwise at a planet
rotational speed of Fplt. The center output gear is the sun gear, and it mates with
the three planets. The sun gear has a rotational speed of Fsun in a counterclockwise direction. In this case, the collinear input and output shafts rotate in the
same direction when viewed from one end of the gear box.
The planet spin or rotational frequency Fplt is calculated based upon a gear
tooth ratio as follows:
T ring
Planet Spin Frequenc y planetary = F plt = F car -------------T plt
(10-13)
The planet passing frequency is determined by multiplying the actual num-
472
Chapter-10
Planet
Carrier
Fcar
Sun
Gear
Fsun
Fplt
Planet
Gear
Fplt
Planet
Gear
Fplt
Fplt
Fplt
Fcar
Stationary
Ring Gear
Fring =0
Fig. 106 Typical Planetary Configuration Of Epicyclic Gear Box - Stationary Ring Gear
ber of planets Np by the planet carrier speed Fcar as in the next equation:
Planet Pass Frequenc y planetary = F plt pass = N p F car
(10-14)
Since the planets are rotating or spinning on one axis, and that axis is
rotating in a circle, the planet absolute frequency is the sum of the planet carrier
and the planet spin speed as shown in equation (10-15). This frequency is seldom
observed, but it is identified as part of the kinematics of the machine.
Planet Absolute Frequenc y planetary = F plt abs = F car + F plt
(10-15)
The most difficult calculation is associated with determination of the output

sun gear rotational speed Fsun. This is not a direct gear ratio due to the element
arrangement within the gear box. The calculation may be performed by examining relative speeds and ratios of the various components. In direct terms, the
overall gear box ratio may be determined by equation (10-16):
T ring
Sun Gear Frequenc y planetary = F sun = F car 1 + --------------

T sun
(10-16)
The planet gear mesh frequency is another common excitation for this type
of machine. This frequency is easily envisioned as the planet rotational frequency Fplt times the number of planet teeth Tplt. It may also be computed based
upon the number of ring gear teeth Tring and the planet carrier input speed Fcar
as shown in equation (10-17):
Planet Gear Mesh planetary = F gm plt = F plt T plt = F car T ring (10-17)
473
Finally, the high speed sun gear mesh frequency is determined by the product of the sun gear rotational speed Fsun and the number of sun gear teeth Tsun
as shown in equation (10-18).
Sun Gear Mesh planetary = F gm sun = ( F sun T sun )
(10-18)
It should be restated that equations (10-13) through (10-18) are directly

applicable only to a planetary gear box configuration as shown in Fig. 10-6. For
demonstration purposes, these equations will be applied on a planetary gear set
containing 214 ring gear teeth, 94 planet gear teeth, and 26 sun gear teeth.
Assume that this is a speed increasing box with three planets, and that the input
speed is 1,780 RPM. Before starting the gear frequency calculations, the validity
of this assembly may be checked with equations (10-11) and (10-12) as follows:
T ring = T sun + 2 T plt = 26 + 2 94 = 26 + 188 = 214
This calculation agrees with the actual ring gear tooth count of 214 teeth.
Next, the second assembly equation should be checked as follows:
T ring + T sun
214 + 26
240
---------------------------------- = --------------------- = --------- = 80 = Integer
Np
3
3
The final value of 80 is an integer number, and the equation is satisfied.
Hence, the basic epicyclic gear assembly equations are satisfied, and it is appropriate to commence the computation of the fundamental excitation frequencies.
The planet rotational frequency Fplt, also known as the planet spin speed, is
determined from equation (10-13) as follows:
T ring
214
F plt = F car -------------- = 1, 780 RPM --------- = 4, 052 RPM
T plt
94
The planet passing frequency Fplt-pass is computed from equation (10-14):
F plt pass = N p F car = 3 1, 780 RPM = 5, 340 RPM
Next, the planet absolute frequency Fplt-abs may be determined from equation (10-15) in the following manner:
F plt abs = F car + F plt = 1, 780 + 4, 052 = 5, 832 RPM
The sun gear rotational frequency Fsun may be calculated with (10-16):
T ring
214
F sun = F car 1 + -------------- = 1, 780 1 + --------- = 16, 431 RPM
26
T
sun
This final ratio of 9.2308:1 (=1+214/26) may seem excessive, but for a planetary gear box of this general arrangement it is quite common. In fact, overall
speed ratios of 12:1 are a common and acceptable practice for planetary boxes.
474
Chapter-10
The planet gear mesh frequency Fgm-plt may be determined from both portions of equation (10-17) to yield the following identical results:
F gm plt = ( F plt T plt ) = ( 4, 052 94 ) = 380, 900 CPM
F gm plt
or
= ( F car T ring ) = ( 1, 780 214 ) = 380, 900 CPM
Finally, the higher frequency sun gear mesh frequency Fgm-sun is easily calculated from expression (10-18) as follows:
F gm sun = ( F sun T sun ) = ( 16, 431 26 ) = 427, 200 CPM
From this basic planetary gear train, a total of seven fundamental excitations have been identified. Field vibration measurements on this gear box will
generally reveal various interactions and modulations between these frequencies. Due to the potential narrow pulse width of some interactions, the resultant
vibration data should always be viewed in both the time and the frequency
domain to make sure that all of the significant vibratory motion is detected.
The second type of epicyclic gear box commonly encountered is the star
configuration as illustrated in Fig. 10-7. For discussion purposes, assume that
the input occurs at the center sun gear that rotates at a speed of Fsun in a counterclockwise direction. In this type of box, the planets are fixed in stationary
bearings, and the planet carrier frequency Fcar is zero. Although the planets continue to rotate clockwise at a frequency of Fplt, there is no translation of the
planet bearing centerlines. Also note that the sun gear is not restrained by a
bearing, and it essentially floats within the mesh of the planets.
The planet gears in Fig. 10-7 engage an outer ring gear that turns in a
clockwise direction at a rotational speed of Fring. The ring gear then connects to
the output shaft directly, or it may mate with an outer coupling assembly
Stationary
Planet
Carrier
Fcar =0
Internal
Ring Gear
Fring
Fplt
Fsun
Fring
Fplt
Fplt
Planet
Gear
Fplt
Sun
Gear
Fsun
Fig. 107 Typical Star Configuration Of Epicyclic Gear Box - Stationary Planet Carrier
475
through a spline arrangement. Obviously, the speed of the outer coupling assembly must be equal to the rotational speed of the internal ring gear Fring. Normally, the circumferential outer coupling assembly is connected to a common end
plate, and this plate is secured to the output shaft, as illustrated in Fig. 10-10 (in
case history 30). As shown in this sketch, the input and output shafts rotate in
opposite directions for this star configuration.
The planet spin or rotational frequency Fplt for this star arrangement is calculated from the sun gear as follows:
T sun
Planet Spin Frequenc y star = F plt = F sun ------------T plt
(10-19)
The planet passing frequency is equal to zero, since the planet carrier does
not rotate (i.e., Fcar =0). By the same logic, the planet absolute frequency Fplt-abs
is identical to the planet rotational speed Fplt. The output ring gear rotational
speed Fring may be determined by:
T sun
Ring Gear Frequenc y star = F ring = F sun -------------T ring
(10-20)
The planet gear mesh frequency is constant across all three gears, and it
may be computed with equation (10-21):
Gear Mesh star = F gm = F plt T plt = F ring T ring = F sun T sun (10-21)
Equations (10-19) through (10-21) only apply to a star arrangement. A set
of example calculations for a star configuration are presented in case history 30.
Prior to this machinery story, the third type of epicyclic gear box should be
reviewed. As previously stated, this is commonly known to as a solar configuration. This name stems from the fact that the sun gear remains fixed (i.e., Fsun=0),
and all of the other gears are in motion, as shown in Fig. 10-8. For discussion
purposes, assume that the input to this gear train is the clockwise rotating ring
gear at a frequency of Fring. The engaged planets are driven in a clockwise direction at a planet spin speed of Fplt. The planet gears translate around the fixed
sun gear, and they drive the planet carrier in a clockwise direction at a speed of
Fcar. In this case, the input and output shafts rotate in the same direction when
viewed from one end of the gear box.
The planet spin or rotational frequency Fplt is calculated as follows:
T ring
Planet Spin Frequenc y solar = F plt = F ring -------------T plt
(10-22)
The planet passing frequency is determined by multiplying the actual num-
476
Chapter-10
Planet
Carrier
Fcar
Internal
Ring Gear
Fring
Fplt
Fcar
Fring
Fplt
Fplt
Planet
Gear
Fplt
Stationary
Sun Gear
Fsun =0
Fig. 108 Typical Solar Configuration Of Epicyclic Gear Box - Stationary Sun Gear
ber of planets Np by the planet carrier speed Fcar as in the next equation:
Planet Pass Frequenc y solar = F plt pass = N p F car
(10-23)
Since the planets are rotating or spinning on one axis, and that axis is
rotating in a circle, the planet absolute frequency is shown in equation (10-24).
Planet Absolute Frequenc y solar = F plt abs = F car + F plt
(10-24)
The determination of the output planet carrier rotational speed Fcar is presented in the following equation (10-25):
T sun
Carrier Frequenc y solar = F car = F ring 1 + --------------

T
ring
(10-25)
Note that the value resulting from the ratio of Tsun/Tring will be considerably less than one. When this value is summed with one, it is clear that the overall gear ratio will be quite low. Hence, a solar configuration of an epicyclic gear
box is strictly a low speed ratio device. This type of machine exhibits only a single gear mesh frequency that is the planet rotational speed Fplt times the number of planet teeth Tplt. It may also be computed based upon the number of ring
gear teeth Tring and the ring gear speed Fring as shown in the next equation:
Gear Mesh solar = F gm = F plt T plt = F ring T ring
(10-26)
Once again, it should be noted that equations (10-22) through (10-26) only
apply to a solar arrangement of an epicyclic gear box. Furthermore, multiple
beats and signal modulations are possible on any epicyclic gear box due to the
477
interaction of a large variety of fundamental excitations. There is also the potential for considerable cross-coupling between the lateral and torsional characteristics in these gear boxes. This becomes even more complicated when compound
epicyclic gear boxes are examined that contain two planets on the same shaft. In
other cases, two epicyclic gear boxes may be used in tandem (i.e., coupled
together) to achieve some very high speed ratios. In either case, the array of
mechanical excitations becomes quite large, and the potential for interaction
with one or more system resonances becomes significant. Specifically, consider
the situation described in the following case history.
Case History 30: Star Gear Box Subsynchronous Motion
The machinery train discussed in this case history consists of a 16,500 HP
gas turbine driving a synchronous generator through an epicyclic gear box. The
arrangement of the gear box and generator, plus the installed proximity probes
are shown in Fig. 10-9. The epicyclic gear box was configured in a star arrangement similar to the previously discussed Fig. 10-7. This box contained 3 planet
gears with 47 teeth Tplt on each gear. The power turbine input speed to the sun
gear Fsun was 8,568 RPM, and the sun gear contained 25 teeth Tsun. The outer
ring gear must rotate at 1,800 RPM to drive the generator. This internally
toothed ring gear contained 119 teeth. As shown in Fig. 10-10, the ring gear was
directly mated to an outer coupling assembly that drove the output gear box
shaft. The gear box output shaft was restrained by an outer and an inner journal
bearing as shown in Fig. 10-10. The sun gear floated on the planet mesh, and
each planet was supported by a fixed journal bearing.
The coupling between the epicyclic gear box and the generator was a close
coupled gear coupling. Although technical specifications were not available for
this assembly, it was clearly a hard coupling with high torsional and lateral stiffness. Hence, any lateral or torsional excitations on the gear box could be easily
Driven CCW
by 16,500 HP
Gas Turbine
at 8,568 RPM
Synchronous
Generator
Epicyclic
Gear
9,500 KW
12,470 Volts, 550 Amps
1,800 Rpm, 3, 60 Hz
4.760:1
K
5V
5H
6V
45 45
6H
45 45
15
2V
75
CW
2H
15
CW
1V
75
15
4V
3V
1H
4H
45 45
75
3H
Fig. 109 Machinery And Vibration Transducer Arrangement For Star Gear and Generator
478
Chapter-10
transmitted to the generator, and vice versa.

Before addressing the specific problems on this machinery, it would be
desirable to check the epicyclic gear box configuration, and identify the anticipated excitation frequencies. As before, the validity of this assembly may be
checked with equations (10-11) and (10-12), as follows:
T ring = T sun + 2 T plt = 25 + 2 47 = 25 + 94 = 119
This calculation agrees with the actual ring gear tooth count of 119 teeth.
Next, check the second assembly equation as follows:
T ring + T sun
119 + 25
144
---------------------------------- = --------------------- = --------- = 48 = Integer
Np
3
3
The value of 48 is an integer number, and the equation is satisfied. Hence,
the basic epicyclic gear assembly equations are satisfied, and it is appropriate to
compute the fundamental excitation frequencies. The planet rotational or spin
frequency Fplt is determined from equation (10-19) as follows:
T sun
25
F plt = F sun ------------- = 8, 568 RPM ------ = 4, 557 RPM
47
T plt
The ring gear rotational frequency Fring may be verified with (10-20):
T sun
25
F ring = F sun -------------- = 8, 568 RPM --------- = 1, 800 RPM
T ring
119
The gear mesh frequency Fgm may be determined from the first portion of
equation (10-21) to yield the following:
F gm = F plt T plt = 8, 568 RPM 25 = 214, 200 CPM
This star epicyclic gear configuration provides the proper output speed of
1,800 RPM, and the above calculations also define the planet rotational speed
Fplt, plus the gear mesh frequency Fgm. Under normal machinery behavior the
dominant shaft vibration frequency on both the gear box output and the generator should be 1,800 CPM. On the power turbine shaft (sun gear input), the major
shaft vibration frequency should be 8,568 CPM. The planet rotational frequency
of 4,557 CPM probably would not appear unless one or more planets were in a
state of distress. Finally, the major high frequency casing vibration component
should logically occur at the gear mesh frequency of 214,200 CPM. Although this
is a complex mechanical system, the number of fundamental excitations are limited and definable.
Initial operational tests on this unit revealed normal and acceptable behavior at full load. Vibration levels were low, journal positions were proper, bearing
temperatures were normal, and overall skid vibration was acceptable. The transient startup and coastdown Bode plots were likewise normal, and all static plus
dynamic measurements pointed towards a normal machinery train. However,
479
12.2"
4.0"
Ring Gear
Probes
1V & 1H
12.6"
Outer Coupling
Forward Probes
2V & 2H
Outer Coupling
Aft Probes
3V & 3H
Outer Coupling
With Spline Fit
To Ring Gear
CW Rotation at
1,800 RPM
Internal Ring Gear
119 Teeth
CW Rotation at
1,800 RPM
Inner
Bearing
Output
Bearing
Output
Shaft
Probes
4V & 4H
Output Shaft
Fig. 1010 Low Speed Probes Installed In Star Configuration Epicyclic Gear Box
during a reduced load test, an unusual subsynchronous vibration component at

1,240 CPM appeared on the generator. This frequency component at nominally
69% of rotational speed was forward and elliptical. The highest amplitudes
occurred at the inboard coupling end bearing. Further investigation revealed
that the same frequency appeared on the epicyclic gear box output bearing.
As the examination of the data progressed, it was clear that additional
measurement points along the low speed rotor elements would be beneficial. The
machinery was originally equipped with X-Y proximity probes mounted at 45
from vertical center. This included the gear box output (probes 4V & 4H), the
generator coupling end bearing (probes 5V & 5H), and the generator outboard or
exciter end bearing (probes 6V & 6H). All three sets of original transducers are
depicted on the machinery arrangement diagram Fig. 10-9.
In order to obtain more information about this low frequency phenomena,
three more sets of X-Y probes were installed through the top cover of the gear
box. These supplemental probes were mounted with the vertical or Y-axis probe
at 15 to the left of the vertical centerline, and the horizontal or X-axis transducer was positioned at 75 to the right of vertical. The physical location of these
additional probes are shown in Figs. 10-9 and 10-10. The first set of probes (1V &
1H) were positioned on the outer diameter of the ring gear. The next two sets of
pickups (probes 2V & 2H) were located on the forward part of the outer coupling,
and probes 3V & 3H were located on the aft portion of the outer coupling. In
essence, probes 3V & 3H were in line with the inner bearing. The location and
axial distance between gear box transducer planes is described in Fig. 10-10.
This diagram also depicts the outer coupling spline fit to the ring gear, and the
general attachment of the outer coupling to the gear box output shaft.
Armed with this full array of transducers observing the low speed rotors, a
series of detailed operational tests were conducted in an effort to quantify the
480
Chapter-10
characteristics of this subsynchronous vibration component.

A graphical description of this subsynchronous vibration is presented in
Fig. 10-11. This steady state data was obtained with the low speed shaft operating at 1,800 RPM, and a 1,600 KW load on the generator. The shaft vibration
from each set of X-Y proximity probes was band-pass filtered at the maximum
subsynchronous frequency of 1,246 CPM, and the resultant orbits plotted at each
measurement station. Since the subsynchronous orbits are all elliptical, a comparison of severity was achieved by identifying the vibration amplitude associated with the major axis of each ellipse. The subsynchronous orbits presented in
Fig. 10-11 also include a simulated timing signal at the tuned frequency of 1,246
CPM. This simulated timing signal is not related to a stationary reference point
like the shaft once-per-rev Keyphasor signal. However, the relative phase relationship between all six measurement locations is accurate and consistent.
From Fig. 10-11, it is noted that the subsynchronous motion on the ring
gear and the forward end of the outer coupling assembly are in unison. Similar
orbit patterns are evident, and the respective amplitudes and phase markers are
almost identical. At the aft end of the outer coupling assembly the subsynchronous motion was attenuated. The orbit at the output shaft of the gear box reveals
a phase reversal plus a lower amplitude at the subsynchronous frequency. The
magnitude of the subsynchronous vibration continues to diminish across the coupling to the generator, with the lowest amplitudes appearing on the outboard
end of the generator. In essence, the subsynchronous motion has a conical mode
shape with maximum amplitudes at the ring gear and outer coupling; with a
nodal point between the outer coupling aft, and the gear box output.
or
erat
Gen
Exciter End
Major Axis=
0.5 Mils,p-p
Cpl
x
r Bo
Gea
1.0 Mil / Div.

Drive End
Major Axis=
1.0 Mils,p-p
Output Shaft
Major Axis=
1.2 Mils,p-p
Outer
Coupling Aft
Major Axis=
0.8 Mils,p-p
Ring Gear
Major Axis=
3.6 Mils,p-p
Outer
Coupling Fwd
Major Axis=
3.4 Mils,p-p
Fig. 1011 Initial Mechanical Configuration Subsynchronous Shaft Orbits Filtered At

1,246 CPM With Rotational Speed Of 1,800 RPM And 1,600 KW Load on Generator
481
Closer examination of the variable speed startup data with cascade plots
revealed an independent lateral vibration component that migrates from 1,000
to 1,300 CPM. This behavior was dominant on the gear box, with much lower
amplitudes on the generator. In addition, the unfiltered shaft vibration data
revealed an amplitude modulation between running speed and the subsynchronous component. This behavior was most visible on the ring gear, and the outer
coupling assembly. Finally, when the subsynchronous component was active, casing accelerometers reveal a 1,240 CPM modulation of the 214,200 CPM gear
mesh frequency.
From this data array it is clear that the subsynchronous shaft vibration
encountered on the low speed end of this machinery train originates within the
epicyclic gear box. It was also determined that the subsynchronous excitation
occurs at a frequency that only varies between 1,220 and 1,260 CPM. It was also
documented that the largest subsynchronous vibration amplitudes appear during a limited torque range of 5,700 to 6,300 foot-pounds.
This information was substantiated during variable speed generator tests
at 1,500 and 1,800 RPM. Specifically, test data at 1,800 RPM and 1,600 KW
(6,260 foot pounds) revealed a subsynchronous frequency of 1,240 CPM. By comparison, at a reduced generator speed of 1,500 RPM and a load of 1,212 KW
(5,690 foot pounds), the subsynchronous component appeared at essentially the
same frequency of 1,220 CPM. Any changes in generator load (up or down) would
attenuate, or completely eliminate the subsynchronous vibration.
Load changes, or more specifically torque changes, are directly associated
with the repeatable appearance of this subsynchronous component. Since the
subsynchronous excitation appears at essentially a constant frequency, and the
maximum amplitude occurs within a limited torque range serious consideration should be directed towards a resonant response in a twisting direction.
That is, the excitation of a lower order torsional resonance should be considered
as a realistic possibility. This concept was reinforced when the undamped torsional response calculations revealed a first mode at 1,320 RPM. Although some
of the mass elastic data was questionable, the coincidence of the calculated torsional critical speed and the measured behavior could not be ignored.
However, the correlation of the measured subsynchronous vibration component with the calculated torsional resonance frequency was an unpopular conclusion. This was complicated by the fact that the subsynchronous component was
sensitive to changes in the oil supply temperature, which could be related to
changes in damping. In addition, the machinery adversely responded to minor
alignment changes across the low speed coupling between the gear box and the
generator. It appeared that raising the generator by 10 Mils unloaded the gear
output bearing, and allowed it to go unstable. Hence, the popular corporate theory was that the observed behavior was nothing more than a bearing stability
problem. Since the major activity occurred within the epicyclic gear box, the parties in charge of the machinery elected to suppress the subsynchronous instability by unbalancing the outer coupling assembly. This change was implemented
by adding a 79 gram unbalance weight to the forward axial face of the outer coupling assembly. The influence of this 79 gram unbalance weight on the subsyn-
482
Chapter-10
or
erat
Gen
Exciter End
Major Axis=
0.6 Mils,p-p
Cpl
x
r Bo
Gea
0.5 Mils / Div.

Drive End
Major Axis=
1.0 Mils,p-p
Output Shaft
Major Axis=
0.4 Mils,p-p
Outer
Coupling Aft
Major Axis=
0.8 Mils,p-p
Outer
Ring Gear Coupling Fwd
Major Axis= Major Axis=
1.8 Mils,p-p 1.8 Mils,p-p
Fig. 1012 Unbalance Of 79 Grams On Outer Coupling - Subsynchronous Orbits Filtered

At 1,248 CPM With Rotational Speed Of 1,800 RPM And 1,600 KW Generator Load
rato
e
Gen
Exciter End
Major Axis=
0.7 Mils,p-p
Cpl
x
r Bo
Gea
1.25 Mils / Div.

Drive End
Major Axis=
0.6 Mils,p-p
Ring Gear
Major Axis=
4.5 Mils,p-p
Outer
Coupling Fwd
Major Axis=
4.3 Mils,p-p
Outer
Coupling Aft
Major Axis=
1.6 Mils,p-p
Output Shaft
Major Axis=
0.6 Mils,p-p
Fig. 1013 Unbalance Of 79 Grams On Outer Coupling - Synchronous Shaft Orbits Filtered At Rotational Speed Of 1,800 RPM With 1,600 KW Generator Load
Process Fluid Excitations
483
chronous motion is shown on Fig. 10-12. Note that the gear box vibration
amplitudes at 1,248 CPM on the ring gear and the outer coupling have been
attenuated from nominally 3.6 to 1.8 Mils,p-p. However, the vibration response
over on the generator at this frequency was virtually unaffected. Since the
amplitude of this subsynchronous vibration on the generator was the primary
concern of the OEM, the addition of the 79 gram unbalance to the gear box outer
coupling assembly was not an acceptable solution.
The inappropriateness of this weight addition was further demonstrated by
examining the synchronous 1X motion at each of the measurement planes as
presented in Fig. 10-13. The previous running speed vibration at 1,800 RPM
before installation of the 79 gram unbalance varied between 0.3 and 0.7 Mils,p-p
at each in the measurement location. However, after the 79 grams was attached,
the 1X amplitudes increased to 4.3 Mils,p-p at the outer coupling, and 4.5 Mils,p-p
on the ring gear. Clearly, these increased synchronous amplitudes would be detrimental to the long-term reliability of this gear box.
In the final assessment, the addition of unbalance weights to the gear box
does not represent a viable solution to the subsynchronous vibration problem. In
fact, it does impose additional dynamic forces upon the gear elements. The
proper engineering solution included a modification of the gear box output bearing to cope with the occasional instability due to bearing unloading. In addition,
it was necessary to recognize that the subsynchronous motion at nominally 1,240
CPM was logically a torsional natural frequency. This resonance appeared as a
lateral vibration due to cross-coupling between torsional and lateral motion
across the gear teeth. This torsional resonance was directly related to the stiff
gear coupling between the epicyclic gear box and the generator. In all cases,
units that contained the stiff gear coupling exhibited this subsynchronous component at reduced load, and identical units that had a torsionally softer flexible
disk coupling did not experience this subsynchronous lateral response.
PROCESS FLUID EXCITATIONS

Fluid handling machines invariably contain some arrangement of stationary and rotating blades or vanes. This applies to machines handling incompressible fluids such as pumps, hydraulic turbines, and extruders, plus machines that
handle compressible fluids such as steam or gas turbines, centrifugal compressors, expanders, and blowers. In many cases, the number of blades or vanes
times the shaft rotative speed provides a simple expression for computation of a
potential blade passing frequency Fb as shown in the next equation.
Blade Pas sin g Frequency = F b = N b RPM
where:
Fb = Blade Passing Frequency (Cycles / Minute)

Nb = Number of Blades or Vanes (Dimensionless Integer)
RPM = Rotative Speed (Revolutions / Minute)
(10-27)
484
Chapter-10
An application of this concept is presented in Fig. 10-14 that displays a pair

of spectrum plots obtained from a large single shaft gas turbine. The top FFT
plot in Fig. 10-14 was acquired at the inlet end bearing housing. It displays a frequency array that coincides with virtually all of the axial flow air compressor
stages. The sixth stage was not evident and there is no guarantee that some of
the other components are attributable to only one stage. In addition, the components that represent several stages (same number of blades on more than one
stage) might be due to an excitation from only one stage, or the interaction of
several stages to produce one component at a common frequency.
Fig. 1014 Blade Passing Excitations On A Large Single Shaft Industrial Gas Turbine
The simpler plot at the bottom of Fig. 10-14 was obtained from the exhaust
end bearing housing. This diagram is dominated by the hot gas power turbine
first and second stage blade passing frequencies. Due to the clear demarcation of
blade counts, this data carries more credibility than the complex spectra
extracted from the inlet end bearing. However, the diagnostician should still
review this data with caution. In most cases, the impedance path between the
rotor excitation and a bearing cap acceleration measurement is unknown. It is
therefore difficult to correlate these frequency components and amplitudes to
specific levels of severity within the machine. At best, the various components
may be identified in terms of harmonic order, and potentially associated with
specific mechanical elements (e.g., number of turbine blades on a particular
stage). From there on, the machinery diagnostician is faced with routine examination of the high frequency spectra, and a trending of the results. Naturally, for
this type of program to be effective, the same accelerometer must be mounted in
the same location, and the high frequency data acquired and processed in the
same manner. Variations of any of these steps would invalidate the accumulated
database.
485
It must also be recognized that high frequency blade passing excitations are
often influenced by stationary objects. This interaction between the rotating and
the stationary mechanical systems is a nuisance when dealing with compressible
fluids, and it forms a mandatory part of the analysis when examining machines
that handle incompressible liquids.
For instance, consider Table 10-1 of calculated pump vane passing frequencies. This data is for a centrifugal pump with a vaned diffuser. Typically, the
number of diffuser vanes exceed the number of impeller vanes. Due to the close
coupled configuration of this type of machinery, there is a definite interrelationship and resultant excitation between stator and rotor parts. Thus, a six vane
impeller running inside of a nine vane diffuser will produce a blade passing frequency at four times rotative speed. If the machinery diagnostician is expecting
to see a 6X blade passing frequency on the pump, the appearance of a strong 4X
component can be most disconcerting.
Table 101 Vane Pass Frequency For Various Combinations Of Impeller And Diffuser Vanes
Pump
Diffuser
Vanes
Number of Pump Impeller Vanes

3
14
16
15
27
15
10
28
10
21
16
11
12
12
10
12
21
32
45
12
25
35
13
12
12
25
12
14
40
27
14
15
15
27
15
16
14
16
16
15
15
49
63
17
18
16
35
18
35
16
18
18
10
35
35
28
19
18
20
20
18
56
56
18
20
21
21
21
81
486
Chapter-10
Table 10-1 was generated with a computer program for computation of vane
passing frequencies for diffuser pumps by James E. Corley6. This program was
implemented over a wide range of impeller and diffuser vane configurations, and
the results summarized in Table 10-1. For purposes of clarity, it should be
restated that this table and the associated discussion is limited to vaned diffuser
type centrifugal pumps. For the more common configuration of volute pumps, the
vane pass frequency reverts back to the original definition of speed times the
number of impeller vanes stated in equation (10-27).
The previous discussion should not imply that pump excitations are limited
to vane passing activity. In reality, a variety of generic excitations as discussed in
chapter 9, plus the hydraulic behavior previously mentioned, are possible. Other
problems such as cavitation, internal recirculation, flow distribution, plus difficulties associated with mechanical seals and couplings occur. For more detailed
information on pump behavior, the reader is encouraged to examine technical
papers such as the excellent documents by Nelson and Dufour7, and Schiavello.8
Although the vast majority of process fluid excitations are directly associated with rotating machinery blade or vane passing frequencies, other types of
fluid excitations do exist, and they are responsible for some significant mechanical failures. In all cases, it must be recognized that the fluid flow stream (compressible or incompressible) carries a substantial amount of energy, and it may
be a significant excitation source. The appearance of pressure fluctuations at the
boundary layer of a fluid stream, the problems associated with turbulent flow,
the destructive forces associated with vortex shedding frequencies, or any of the
acoustic mechanisms are generally formidable engineering problems. In addition, the cross-coupling of fluid excitations into piping systems or support structures may add a new dimension of complexity to an already difficult problem.
The vortex induced vibration problem is particularly interesting, since it
encompasses the flow of fluids over stationary objects. This behavior is clearly
described by Robert D. Blevins9 as follows:
Structures shed vortices in a subsonic flow. The vortex street wakes tend to
be very similar regardless of the geometry of the structure. As the vortices are shed
from first one side and then the other, surface pressures are imposed upon the
structureThe oscillating pressures cause elastic structures to vibrate and generate aeroacoustic soundsThe vibration induced in elastic structures by vortex
shedding is of practical importance because of its potentially destructive effect on
bridges, stacks, towers, offshore pipelines, and heat exchangers
6 James E. Corley, Tutorial Session on Diagnostics of Pump Vibration Problems, Proceedings
of the Fourth International Pump Symposium, Turbomachinery Laboratory, Texas A&M University,
College Station, Texas (May 1987).
7 W.E. (Ed) Nelson and J.W. Dufour, Pump Vibrations, Proceedings of the Ninth International
Pump Users Symposium, Turbomachinery Laboratory, Texas A&M University, College Station,
Texas (March 1992), pp. 137-147.
8 Bruno Schiavello, Cavitation and Recirculation Troubleshooting Methodology, Proceedings
of the Tenth International Pump Users Symposium, Turbomachinery Laboratory, Texas A&M University, College Station, Texas (March 1993), pp. 133-156.
9 Robert D. Blevins, Flow-Induced Vibration, Second Edition, (New York: Van Nostrand Reinhold, 1990), p. 43.
487
Blevins goes on to explain that vortex shedding from a smooth circular cylinder in a subsonic flow is a function of the Reynolds number. In this context, the
Reynolds number NRe is defined in the following manner:
DV
N Re = --------------
where:
(10-28)
NRe = Reynolds Number (Dimensionless)

V = Free Stream Velocity Approaching the Cylinder (Inches / Second)
D = Cylinder Diameter (Inches)
= Kinematic Viscosity (Inches2 / Second)
In chapter 4 of this text, equation (4-5) identified the variables used to compute the Reynolds through the minimum oil film of a bearing. An initial comparison between equations (4-5) and (10-28) reveals some differences. However, a
closer examination of (4-5) shows that the term x R is surface velocity of a
rotating shaft with units of inches per second. This is equivalent to the free
stream velocity approaching the cylinder V shown in equation (10-28). The oil
film height in inches designated by H in equation (4-5) is equivalent to the cylinder diameter D used in (10-28). Finally, the remaining terms are all associated
with the moving fluid viscosity. In equation (10-28) the kinematic viscosity was
used, whereas the absolute or dynamic viscosity was applied in (4-5). The two
viscosity formats are directly related, as shown in equation (10-29).
G
= -------------
where:
(10-29)
= Absolute or Dynamic Viscosity (Pounds-Seconds / Inch2),

G = Acceleration of Gravity (386.1 Inches / Second2)
= Fluid Density (Pounds / Inches3)
A dimensional analysis of equation (10-29) reveals that the units are correct. Furthermore, both Reynolds number equations (4-5) and (10-28) are equivalent. Whereas (4-5) was used to define the ratio of inertia to viscous forces in a
fluid film bearing expression (10-28) is applied to a fluid stream flowing across
a smooth circular cylinder. In this application, the pattern generated by vortices
down stream of the cylinder may be predicted based upon the value of the Reynolds number. Specific flow regimes have been identified by various investigators, and the reader is again referenced to the text by Robert Blevins for detailed
information. Some of the information on this topic is also available in the Shock
and Vibration Handbook10 in the section authored by Blevins. It should also be
mentioned that numerous studies have been conducted on vortex shedding, and
the associated vortex induced vibration. Hence, this is a well-documented technical field that incorporates many empirical studies and analytical solutions. The
references provided within the Blevins text reveal the true breadth of this physi10 Cyril M. Harris, Shock and Vibration Handbook, Fourth edition, (New York: McGraw-Hill,
1996), pp. 29.1 to 29.19.
488
Chapter-10
cal behavior that stretches across many technical fields.

Of particular importance within the machinery business is the relationship
defined by the Strouhal number NStr. This is a non-dimensional number that
allows computation of the fundamental or predominant vortex shedding frequency Fs as defined in the following expression:
Fs D
N Str = ----------------V
where:
(10-30)
NStr = Strouhal Number (Dimensionless)

Fs = Vortex Shedding Frequency (Cycles / Second)
The cylinder diameter D, and the constant velocity of the fluid stream V is
the same value used in equation (10-28). Thus, for a given cylinder diameter D,
and flow velocity V, the vortex shedding frequency Fs may be computed if the
Strouhal number NStr is known. Fortunately, there are multiple empirical tests
that display a consistent relationship between the parameters specified in equation (10-30). For instance, on pages 48 through 51 of Bevins text, a variety of
charts describe the Strouhal number for a circular cylinder, an array of inline
and staggered cylinders, plus various other geometric cross sections. From this
database, it is clear that the Strouhal number for the vast majority of cases will
vary between values of 0.1 and 0.8.
For the simple case of a circular cylinder with a Reynolds number between
500 and 1,000,000, this data suggests that the Strouhal number has a value of
nominally 0.2. Many technical references identify this value as 0.22. However,
across the range of Reynolds numbers specified, 0.2 represents a more realistic
average for the Strouhal number. If value this is substituted into (10-30), the vortex shedding frequency Fs may be computed directly from:
0.2 V
F s = -----------------D
(10-31)
As a practical example of the application of these vortex shedding concepts,

consider the situation of a gas turbine exhaust stack. If the top cylindrical portion of the stack has a diameter of 20 inches, and the environment consisted of
standard temperature (60F) and pressure (14.7 Psia), with wind gusts of 50
miles per hour it would be desirable to compute the anticipated vortex shedding frequency. From various sources, the absolute viscosity of air under these
conditions is 0.018 centipoise. This is equal to 0.00018 poise. From the conversion factors presented in Appendix C of this text, 1 poise is equivalent to 1 dynesecond/centimeter2. Thus, the absolute or dynamic viscosity of air at standard
temperature and pressure would be equal to 0.00018 dyne-second/centimeter2.
Converting the metric viscosity units to English units may be accomplished in
the following manner:
489
2
= 0.00018 Dyne-Sec/Cm 2.248 10
Pound/Dyne ( 2.54 Cm/Inch )
= 0.00018 2.248 10 6.4516 = 2.611 10
Pound-Sec/Inch
From Table B-2 in the appendix of this text, the density of air at standard
temperature and pressure is equal to 0.07632 pounds per foot3. This density
value may be converted to consistent units as follows:
3
= 0.07632 Pound/Foot ( 1 Foot 12 Inches ) = 4.417 10
Pound/Inch
Based on these physical properties, the kinematic viscosity of air may now
be computed from equation (10-29):
9
G
2.611 10 Pound-Sec/Inch 386.1 Inch/Sec
2
= -------------- = -------------------------------------------------------------------------------------------------------------- = 0.0228 Inches /Sec.
5
3
4.417 10 Pound/Inch
The peak wind velocity of 50 miles per hour may now be converted into
compatible engineering units of inches per second in the following manner:
V = 50 Miles/Hour 5, 280 Feet/Mile 12 Inches/Foot 1 Hour 3, 600 Sec = 880 Inch/Sec.
Sufficient information is now available to compute the Reynolds number of
the air flow over the cylindrical stack with equation (10-28):
DV
20 Inches 880 Inch/Sec.
N Re = --------------- = ----------------------------------------------------------- = 772, 000
2
0.0228 Inches /Sec.

The Reynolds number of 772,000 falls within the previously specified range
of 500 and 1,000,000. This provides confidence in using a Strouhal number of 0.2
for this case of a circular stack. More specifically, these conclusions allow the
direct application of equation (10-31) as follows:
0.2 V
0.2 880 Inch/Sec
F s = ------------------ = ------------------------------------------- = 8.80 Cycles/Sec 60 Sec/Min = 528 Cycles/Min.
D
20 Inches
Hence, with a 50 mile per hour wind, the anticipated vortex shedding frequency would be 528 cycles per minute (8.80 Hz). This is an appreciable frequency that could influence the gas turbine, or any of the associated mechanical
equipment. Furthermore, this frequency will change with wind speed. If any
combination of wind speed and associated vortex shedding frequency coincided
with a natural frequency of the stack, the results could be devastating.
The traditional solution to this type of problem resides in the modification
of the stack outer diameter to disrupt the vortex shedding, and therefore eliminate (or substantially minimize) the excitation source on the stationary cylinder.
Common modifications include helical strakes wrapped around the stack, or a
series of external slats or shrouds that are designed to break up the vortices. In
some cases, an analytical model using Computational Fluid Dynamic (CFD) software might be sufficient to properly examine the system. In other situations, the
development and testing of a scale model in a wind tunnel might be appropriate.
490
Chapter-10
In all cases, the diagnostician must be aware of this vortex shedding phenomena
and the potential for induced vibration into the structure.
Another mechanism that periodically appears within fluid handling systems is the acoustic resonance problem. This is the classic organ pipe behavior
that appears in virtually every physics textbook. The traditional discussion of
standing wave theory relates the velocity of sound (i.e., sonic velocity) in the fluid
media Vs with the occurring acoustic frequency Fa and the wavelength as presented in the following expression:
V s = Fa
where:
(10-32)
Vs = Sonic or Acoustic Velocity (Feet / Second)

Fa = Acoustic Frequency (Cycles / Second)
= Standing Wave Length (Feet)
The velocity of sound Vs will vary according to the media. Solids will generally display the highest values, sonic velocity in liquids will generally be lower,
and gases will display even low speeds. For example, Table 10-2 summarizes
some common values for the velocity of sound in assorted solids and liquids.
Table 102 Typical Values For Sonic Velocity In Various Solids And Liquids
Velocity of Sound
Material
Solids
Liquids
At 60F and 14.7 Psia
(Feet/Second)
Lead
4,030
Brass
11,480
Copper
11,670
Iron & Steel
16,410
Aluminum Alloys
16,740
Graphite
19,700
Alcohol
3,810
Oil - Sp. Gr.=0.9
4,240
Mercury
4,770
Fresh Water
4,860
Glycerin
6,510
The velocity of sound in solids will remain constant over a wide range of
conditions. This is due to the fact that the material density sol and modulus of
elasticity E remain constant over a wide range of conditions. The sonic velocity in
any solid may be computed with the following common expression:
491
Vs
where:
sol
GE
------------------------144 sol
(10-33)
Vs-sol = Velocity of Sound in a Solid (Feet / Second)

G = Acceleration of Gravity (= 386.1 Inches / Second2)
E = Modulus of Elasticity (Pound / Inch2)
sol = Solid Material Density (Pounds / Inch3)
To check the validity of (10-33), the properties from Table B-1 in appendix B
may be extracted and inserted into this expression. For example, pure copper has
a modulus of elasticity equal to 15,800,000 pounds/inch2, and a density of 0.323
pounds/inches3. Equation (10-33) may now be evaluated as follows:
6
Vs
sol
386.1 Inches/Sec 15.8 10 Pounds/Inch

------------------------------------------------------------------------------------------------------ = 11, 450 Feet/Minute
2
2
3
144 Inches /Foot 0.323 Pounds/Inch
The resultant value of 11,450 feet per minute is comparable to the velocity
of sound in copper listed in Table 10-2 of 11,670 feet per minute. The 2% variation between velocities is due to the fact that average values to three significant
figures are used for the physical properties in Table B-1. By comparison, experimental results provide the sonic velocity listed in Table 10-2.
The velocity of sound in liquids is determined with an expression equivalent to equation (10-33). The difference between calculating the sonic velocity in
solids versus liquids is that Youngs modulus of elasticity E is used for solids, and
the bulk modulus B is used for liquids. Variations in pressure and temperature of
the liquid will be compensated by using the bulk modulus and density at the
actual fluid operating conditions in the following expression.
Vs
where:
liq
GB
------------------------144 liq
(10-34)
Vs-liq = Velocity of Sound in a Liquid (Feet / Second)

B = Bulk Modulus (Pound / Inch2)
liq = Liquid Material Density (Pounds / Inch3)
If a lubricating oil at atmospheric pressure and 60F has a bulk modulus of

219,000 pounds per inch2, and a density of 0.0327 pounds per inches3, the sonic
velocity is computed with equation (10-34) in the following manner:
2
Vs
liq
386.1 Inches/Sec 219, 000 Pounds/Inch

--------------------------------------------------------------------------------------------------- = 4, 240 Feet/Minute
2
2
3
144 Inches /Foot 0.0327 Pounds/Inch
This sonic velocity in oil agrees directly with the value in Table 10-2. Next,
the velocity of sound in gases must be addressed. It is well understood that gases
are even more sensitive to variations in pressure and temperature. For perfect
gases, the sonic velocity may be computed with the following expression:
492
Chapter-10
Vs
where: Vs-gas
g
k
R
T
z
mw
=
=
=
=
=
=
=
gas
gkRTz
-------------------------------------------mw
(10-35)
Velocity of Sound in a Gas (Feet / Second)

Acceleration of Gravity (= 32.17 Feet / Second2)
Specific Heat Ratio of Cp/Cv (Dimensionless)
Universal Gas Constant (= 1,546 Foot-Pound force/ Pound mole -R)
Absolute Gas Temperature (R = F + 460)
Gas Compressibility Based on Temperature and Pressure (Dimensionless)
Gas Molecular Weight (Pound / Pound mole)
The compressibility z is typically determined from charts based upon the

pseudo-reduced temperature and the pseudo-reduced pressure of the gas. If the
constants g and R are included into equation (10-35), the expression may be
somewhat simplified as follows:
Vs
gas
kTz
= 223 ----------------------mw
(10-36)
For situations where the specific heat ratio k is unknown, the following
equation (10-37) may be used to compute k based upon the molecular weight mw
and the specific heat of the gas at constant pressure cp.
1
k = ------------------------------------- 1.986
1 ----------------------
mw c p
where:
(10-37)
cp = Specific Heat at Constant Pressure (BTU/Pound-F)

1.986 = Universal Constant R = 1,546 Foot-Pound force/ Pound mole -R divided
by 778.3 BTU/Foot-Pounds (BTU/ Pound mole -R)
For example, considering a normal composition of air at 60F (520R) and

14.7 Psia pressure, the molecular weight mw, and the specific heat at constant
pressure cp, may be obtained directly from Table B-2 located in appendix B of
this text. Based on these values, the specific heat ratio k may be calculated with
equation (10-37) in the following manner:
1
1
1
1
k = ----------------------------------------------------------------------------------- = -------------------------------- = ---------------------- = ------------- = 1.40
1
0.286
0.714
1.986 BTU
1.986
--------------------------------1 ----------------
Pound mole-R
1 -------------------------------------------------------------------
6.9451
Pound 0.2398 BTU
28.962
-------------------------------- -----------------------------
Pound-F
Pound mole
This answer of k=1.40 for air at standard temperature and pressure is a
normal and expected value. Under these conditions, the pseudo-reduced temperature is 2.18, and the pseudo-reduced pressure is 0.027 the compressibility
factor z is very close to 1.0. The velocity of sound may now be easily determined
by substitution of the known parameters into equation (10-36) as follows:
Vs
air
493
kTz
1.40 520 1.0
= 223 ----------------------- = 223 --------------------------------------- = 223 25.136 = 1, 118 Ft/Sec.
mw
28.962
This solution is consistent with published values of the velocity of sound in

air. For gas streams, the composition and effective mole weight mw, plus the
effective specific heat ratio k, the compressibility z, and the absolute temperature T all contribute to the final value of the sonic velocity Vs. Clearly, the general wave equation (10-32) may be subjected to variations due to predictable
changes in the velocity of sound in the gas stream. Furthermore, the reader
should be aware that sonic velocity in a liquid may be a parabolic shaped curve
when velocity is plotted against temperature. In this situation, two entirely different temperatures share the same value for the velocity of sound. For all critical calculations, the machinery diagnostician should obtain the best possible
reference for the specific fluid properties, plus the velocity of sound, within the
bounds of the actual process conditions.
Returning back to equation (10-32), the concept of a unique frequency Fa
relating the acoustic velocity Va to the wavelength is straightforward relationship for standing waves. The frequency Fa is generally identified as the fundamental frequency. Integer multiples of this fundamental frequency are often
referred to as overtones or harmonics. A frequency of 2Fa would be the first overtone, or the second harmonic. Similarly, a frequency of 3Fa would be identified as
the second overtone or the third harmonic. The concept of harmonic orders of an
acoustical frequency are identical to the multiples or harmonics of any other
vibration component or fundamental frequency.
Within a contained system, such as an organ pipe or a length of schedule 40
steel pipe in an oil refinery, it is possible to establish an acoustic resonance. This
occurs when an excitation frequency coincides with the physical characteristics
of system length and acoustic velocity as described by (10-32). In terms of wave
behavior, a good explanation was presented by Schwartz and Nelson11 as follows:
A standing wave is propagated and reflected when a pressure or velocity
perturbation encounters a discontinuity in an acoustic system. Resonant conditions will result when the total propagation time up and down the system is such
that the reflected wave reaches the excitation source in phase with a subsequent
perturbation. Under these conditions the excitation frequency is said to be at one
of the natural frequencies of the acoustic system
This is exactly the same type of physical behavior exhibited by any other
resonant response. That is, the geometry and physical characteristics of the system define a variety of potential resonances. These resonances remain dormant
until an excitation frequency coincides with the natural resonant frequency.
When that condition occurs, the influenced mechanical element will experience a
condition of high vibration.
11 Randal E. Schwartz and Richard M. Nelson, Acoustic Resonance Phenomena In High
Energy Variable Speed Centrifugal Pumps, Proceedings of the First International Pump Symposium, Turbomachinery Laboratories, Texas A&M University, College Station, Texas (May 1984), pp.
23-28.
494
Chapter-10
Extension of the concept described in equation (10-32) to integer multiples

of the wavelength is generally understandable. However, this same approach
applies to particular fractions of the fundamental wavelength. Specifically, passage lengths equal to one half and one quarter of the fundamental wavelength
may also support acoustic resonances. For instance, consider the condition of a
pipe open at both ends, as shown in Sketches A and B of Fig. 10-15. In this type
Half Wave - Open Both Ends
Quarter Wave - Open One End
Pressure
Velocity
Sketch A
Sketch C
Pressure
Velocity
Sketch B
Sketch D
Fig. 1015 Standing Pressure And Velocity Waves In Pipes
of representation, the end of the pipe is considered to be open if the cross-sectional area changes by a factor of two or more. This physically occurs in many
types of branched piping systems, pulsation bottles, or internal crossover passages on multistage centrifugal pumps.
Sketch A of Fig. 10-15 depicts a standing half wave condition. As shown in
this diagram, the velocity is a maximum at each end of the open pipe (anti-node),
and pressure pulsation is at a minimum (node) at both ends. The peak pressure
(anti-node) is at the middle of the pipe, and this is coincident with the minimum
velocity (node). The first overtone, or second harmonic, of a pipe open at both
ends is presented in Sketch B. Again, the velocity is a maximum at the pipe ends,
and the pressure is at a minimum. Since this is a half-wave condition, the wave
length is equal to equal to twice the passage length Lhalf as follows:
= 2 L half
(10-38)
Substituting equation (10-38) into (10-32), and solving for the acoustic natural frequency Fa, the following expression is easily generated:
Vs
Vs
F a = ------- = -----------------------
2 L half
495
However, this equation only covers the condition depicted in Sketch A of

Fig. 10-15. The general solution for this half-wave behavior may be extrapolated
from the previous expression by including an integer multiplier Nhalf as follows:
Vs
F a = N half -----------------------2 L half
where:
(10-39)
Nhalf = Harmonic Integer of 1, 2, 3, 4 (Dimensionless)

Lhalf = Physical Passage Length for Half-Wave Resonance (Inches)
It should be mentioned that the equations for a half-wave resonance also

apply to a pipe with both ends closed. Although the nodal locations of the pressure and velocity waves become transposed, the physical behavior and the computational equations are the same as a half-wave pipe open at both ends.
The same general approach is applicable to the quarter-wave condition.
This is described by a pipe open at one end, and closed at the opposite end as
shown in Sketches C and D of Fig. 10-15. Sketch C depicts a standing quarter
wave with a maximum velocity (anti-node), and minimum pressure pulsation at
the open end. The peak pressure (anti-node) is at the closed end of the pipe, and
this is coincident with the minimum velocity (node). The first overtone of a pipe
open at one end is presented in Sketch D. Again, the velocity is a maximum at
the end, and the pressure is at a minimum. Since this is a quarter-wave condition, the wave length is equal to equal to four times the passage length Lqtr:
= 4 L qtr
(10-40)
Substituting equation (10-40) into (10-32), and solving for the acoustic natural frequency Fa, the following expression is generated:
Vs
Vs
F a = ------- = -------------------
4 L qtr
Again, this expression only covers the condition shown in Sketch C of Fig.
10-15. The general solution for this quarter-wave behavior may be extrapolated
from the last equation by including an integer multiplier Nqtr as follows:
Vs
F a = N qtr -------------------4 L qtr
where:
(10-41)
Nqtr = Harmonic Integer of 1, 3, 5, 7 (Dimensionless)

Lqtr = Physical Passage Length for Quarter-Wave Resonance (Inches)
A quarter-wave resonance is sometimes referred to as a stub resonance.

This is an accurate description of a piping element that is connected to a main
line, and has a closed end. Various types of piping dead legs, clean-out nozzles,
and nipples fall into this category. Although these are benign looking piping elements, they can harbor quarter-wave resonances that excite the associated pip-
496
Chapter-10
ing system. The quarter-wave stub resonance may also adversely influence
pressure gauge, and dynamic pressure pulsation measurements. Since the passage length Lqtr is often quite short, an acoustic resonant frequency may appear
at an undesirable frequency, and corrupt the pressure pulsation data under
investigation. One solution to this problem is to effectively extend the passage
length with a measurement fixture similar to the diagram previously discussed
in Fig. 6-33. This would substantially reduce the acoustic resonant frequency to a
range that could be removed via electronic filtration, or just totally ignored.
In most cases, the fluid excitation problem analysis will require a combination of vibration measurements, dynamic pressure pulsation measurements,
some level of analytical simulation, plus a recommended fix. Generally, a final
round of testing is also necessary to verify the validity of the implemented solution. An example of this type of fluid problem is presented in the following case
history of a high pressure boiler feed water pump.
Case History 31: Boiler Feed Water Pump Splitter Vane Failures
A group of three large boiler feed water pumps experienced a series of
mechanical failures to their impeller suction splitter vanes. Although the
machinery was installed with some questionable procedures, the destructive failures could not be attributed to the marginal installation. The BFW pumps were
all steam turbine driven, and all three pumps were connected to common suction
and discharge headers. Speed control for each turbine was based upon discharge
header pressure.
The internal configuration of these pumps is depicted in Fig. 10-16. A double suction first stage wheel was located at the outboard end of each pump. A
short internal crossover connected the discharge of the first stage to the suction
of the second stage wheel. A long internal crossover of approximately 70 inches
in overall length extended from the discharge of the second stage to the third
stage suction. Another short crossover connected the third to the fourth stage,
and the final discharge exited the middle of the pump casing at an average pres70 Long 2nd to 3rd Stage Crossover
Failed
Inlet
Splitters
3rd
4th
2nd
1st
Inboard
Coupling End
Discharge
1,050 Psig
Suction
90 Psig
Fig. 1016 Boiler Feed Water Pump Internal Configuration
497
sure of nominally 1,050 Psig.

Each pump stage was equipped with a four vane inlet splitter, and each
impeller contained seven vanes. The documented failures generally originated at
the suction of the third stage as indicated in Fig. 10-16. The splitter vanes at this
location were the most susceptible to breakage, and they caused considerable
internal pump damage during each failure. In most instances, the fourth stage
wheel would be damaged. In some cases, the upstream second stage impeller
would also be subjected to varying degrees of damage.
The machinery was subjected to a series of detailed vibration tests, and it
was discovered that the major frequency component occurred at seven times
rotative speed. This frequency was the impeller vane passing, and it was strongest at the inboard, coupling end of the pump. This location was also coincident
with the majority of the splitter vane failures.
Based on this initial data, an additional test was performed, and dynamic
pressure pulsation probes were added to the transducer suite. The significant
results of this test are presented in the summary plot shown in Fig. 10-17.
Within this plot, the amplitude at impeller vane passing frequency (7X) was plotted for a series of controlled operating speeds. The presented information
includes the vertical casing velocity at the outboard bearing housing, vertical
casing velocity acquired at the coupling or inboard bearing housing, and dynamic
pressure pulsation obtained in the 2nd to 3rd stage internal crossover.
It is clear that the maximum 7X vane passing amplitudes occurred at a
running speed of 3,410 RPM. It is also evident that velocity levels approaching
0.6 IPS,o-p, and pressure pulsation amplitudes in excess of 70 Psi,p-p, were intolerable for this machinery. Furthermore, the peaked response on this plot was
indicative of a possible resonant behavior. The actual frequency of the peak vane
passing response occurred at 23,870 CPM (3,410 CPM times 7 vanes). Based
upon the geometry of this system, the excitation frequency, and the high pressure pulsation levels, an acoustic resonance was suspected.
0.8
80
70
0.6
60
0.5
50
0.4
0.3
Inboard Velocity
0.2
Outboard Velocity
0.1
0
3,300
P P
P
3,400
40
3,500
30
B B
B
P P P P
3,600
Speed (Revolutions / Minute)
3,700
20
10
0
3,800
Pressure Pulsation (PSI,p-p)
Fig. 1017 Initial Casing

Velocity And 2nd To 3rd
Stage Internal Crossover
Pressure Pulsation Filtered
At 7X Impeller Vane Passing Frequency
Casing Velocity (IPS,o-p)
0.7
Interstage Pressure Pulsation
498
Chapter-10
The circulating boiler feed water pump operated at a temperature in the

vicinity of 310F. At this temperature, the velocity of sound in water (i.e., sonic
velocity) is approximately 4,750 feet/second. Since the vane passing amplitudes
increased with speed, it is highly probable that the observed behavior is related
to flow. If this is an acoustic related phenomena, the offending member would
probably be open at both ends rather than closed at one end. Hence, it makes
sense to use the half-wave equation (10-39) for examination of a possible acoustic
interaction. The fundamental half-wave frequency would occur when the integer
Nhalf was set equal to 1. Based upon the velocity of sound Va equal to 4,750 feet
per second, and the 7X vane passing excitation frequency Vs of 23,870 cycles per
minute, equation (10-37) may now be solved for the associated physical passage
length as follows:
Vs
4, 750 Feet/Sec 12 Inches/Foot
57, 000
L half = ---------------- = --------------------------------------------------------------------------------------------- = ------------------ = 71.6 Inches
2 Fa
2 23, 870 Cycles/Min 1 Min 60 Sec
795.7
0.8
80
0.7
70
0.6
60
Inboard Velocity
0.5
50
0.4
40
Interstage Pressure Pulsation

0.3
0.2
30
B B B
B B B B
P P P P P P P
Outboard Velocity
0.1
0
3,300
3,400
3,500
3,600
3,700
20
10
0
3,800
Pressure Pulsation (PSI,p-p)
Fig. 1018 Final Casing

Velocity And 2nd To 3rd
Stage Internal Crossover
Pressure Pulsation Filtered At 6X Impeller Vane
Passing Frequency
Casing Velocity (IPS,o-p)
Recall that the physical passage length of the 2nd to 3rd stage crossover
was previously identified as approximately 70 inches. Due to the uncertainties
involved, the correlation between the calculated half wave resonance passage
length is considered to be in good agreement with the approximate physical
length of the 2nd to 3rd stage crossover.
The final assessment concluded that an acoustic resonance in the 2nd to
3rd stage crossover was directly excited by the 7X pump vane passing frequency.
Since there was no possibility of changing the internal pump passage length, or
changing the pump operating temperature to vary the acoustic velocity the
only remaining viable option was to change the exciting frequency. Hence, the
seven vane impellers were replaced with new six vane wheels.
To test the validity of this conclusion, the same field test was repeated with
a pump equipped with six vane wheels. The results of this test are presented in
Fig. 10-18. Note that the pressure pulsation amplitudes are now in the vicinity of
499
20 Psi,p-p, and the inboard bearing housing velocity amplitudes are slightly
above 0.3 IPS,o-p at the highest pump speeds. However, back within the normal
operating range of 3,400 to 3,500 RPM, the 6X vane passing amplitudes are in
the vicinity of 0.1 IPS,o-p, which is quite tolerable for these particular machines.
Eventually, all three of the BFW pumps were converted to six vane impellers,
and the string of splitter vane failures was completely stopped.
Case History 32: Hydro Turbine Draft Tube Vortex
An entirely different class of machine is examined in the following case history of a vertical hydro turbine driving an 18 megawatt generator at 277 RPM. A
general arrangement drawing of this train is presented in Fig. 10-19. The overall
height of the rotating assembly is approximately 42 feet, and the overall rotor
weight approaches 200,000 pounds. The entire rotor assembly is supported on a
thrust bearing mounted at the top of the 26 pole generator. Lateral rotor support
is provided by radial bearings at three different elevations. As shown in Fig. 1019, an upper guide bearing is located at the top of the generator, above the thrust
assembly. Directly below the generator, a lower guide bearing is installed. Both
the upper and lower generator bearings are tilt pad assemblies that have individual clearance adjustments for each pad. The elevation difference between the
upper and lower generator bearings is 127 inches. A turbine guide bearing is
located approximately 193 inches below the lower generator bearing. This turbine bearing was a segmented journal bearing that consisted of six fixed pads.
An unusually long 4,200 foot penstock supplied water to this unit. The final
water inlet to the spiral case was a straight run of pipe, and the water passed circumferentially through a row of twenty vertical wicket gates as shown in Fig. 1019. Energy was extracted from the water stream with a Kaplan turbine equipped
with six variable pitch blades. A traditional elbow draft tube was installed that
consisted of a vertical drop, followed by the elbow, and a final return to an open
channel that directed the water downstream.
During original commissioning of this unit, it was determined that steady
state operation under load was acceptable, but high vibration amplitudes were
encountered during load rejection. In this condition, the load is suddenly
removed from the generator (tripped breaker), but the energy of the water flow is
still applied to the turbine. This situation results in a significant speed increase
of the rotating assembly. For instance, during load rejection from 12 megawatts,
the speed will increase from 277 RPM to a runaway speed of well over 400 RPM.
The earliest field test data revealed a peak vibration response at a frequency of
660 CPM. This frequency did coincide with a calculated rotor critical speed at the
same speed. It was reasoned that although the rotor speed (420 RPM maximum)
never reached the resonant frequency (660 RPM), there was sufficient energy
imparted to the rotating assembly to excite this major system resonance.
The initial commissioning tests demonstrated that the highest runaway
speed was directly related to the generator load (expected result). Similarly, the
peak vibration amplitudes steadily increased with load, and maximum runaway
500
Chapter-10

Viewed From Top Of
Generator Looking Down
1Y
90
Elevation 1
Upper
Guide
Bearing
1X & K
CW
Rotation
Thrust
Bearing
139"
2Y
90
2X
Generator
3Y
Elevation 2
Lower
Guide
Bearing
90
22"
Elevation 3
3X
75"
Coupling
Assembly
4Y
90
Elevation 4
4X
5Y
Turbine
Guide
Bearing
90
92"
5X
Head Cover
Elevation 5
20
Wicket
Gates
Water
Flow
Spiral
Case
Kaplan Turbine
6 Blades
Draft
Tube
Water
Flow
Fig. 1019 Mechanical Configuration Of Hydro Electric Turbine Generator Set
501
speed. At the highest loads, the massive structure of the power house was
uncomfortably shaken during load rejection. Some relief was obtained by air
injection into the head cover, but high vibration amplitudes still occurred after
the conclusion of the air injection. Additional tests were conducted with changes
in system control variables in an effort to attenuate the response at 660 CPM
during load rejection. However, the entire sequence between load rejection, overspeed with high vibration, concluding with coastdown at acceptable vibration
levels occurred within 30 seconds. Clearly, the dynamics of the mechanical system overwhelmed the capabilities of the turbine control system.
At the completion of the original field testing, it was generally agreed that
operation of the hydro turbine generator should be limited to a maximum load of
6 megawatts (1/3 of rated capacity). Under this steady state load, the vibration
levels encountered during load rejection were deemed to be tolerable. It was also
suggested that stiffening the lower generator guide bearing support would raise
the 660 CPM critical, and reduce the vibration encountered during load rejection. Finally, it was suggested that a fourth lateral bearing be installed above the
coupling in order to raise the critical speed. Most parties agreed that the vertical
span between the generator lower guide bearing and the turbine guide bearing
was abnormally long.
During the next three years more tests were conducted by various individuals, and the unit continued to be limited to 6 megawatts. The external support
for the lower guide bearing was stiffened, and alignment plus clearance adjustments were made to the unit. It was acknowledged that the lateral motion of the
Kaplan turbine blades was excessive during load rejection from high loads. This
was evident by the noise immediately following load rejection, plus the visible
radial rubs between the tips of the turbine blades and the casing.
Up to this point, the vibration data was primarily based upon a pair of X-Y
proximity probes mounted below the generator lower guide bearing, plus casing
readings with various hand held transducers. In an effort to better understand
the dynamic behavior of this unit, a series of additional transducers were temporarily installed at five different elevations. These locations are identified in Fig.
10-19 as elevations 1 through 5. Initially, low frequency accelerometers were
only installed adjacent to the probes mounted above the lower guide bearing (elevation 2). Following the first test, it was clear that casing motion was appreciable, and all subsequent tests were conducted with low frequency, high sensitivity
accelerometers mounted next to each proximity probe. When necessary, the
acceleration data was double integrated to casing displacement in a custom analog integration box and this data was electronically added to the shaft vibration signals. Signal addition of the shaft relative with the casing absolute
displacement was achieved with a standard voltage summing amplifier. The
resultant output was shaft absolute vibration. Although this data manipulation
was not necessary during steady state operation, it did provide an enhanced perspective during some of the transient speed tests.
All of the X-axis probes were aligned in the direction of the downstream
water flow, and the Y-axis probes were located 90 away, or perpendicular to the
flow. An optical Keyphasor was installed at the top of the generator in-line with
502
Chapter-10
the X-axis vibration transducers. A series of pressure pulsation transducers were

mounted in the penstock, head cover, and draft tube for most of the testing. In
addition, strain gages were mounted on the lower portion of the exposed shaft.
These sensors were powered by batteries, and the dynamic signals transmitted
to a receiving antenna for demodulation and final processing.
The generator lower guide bearing pads were replaced due to concerns over
an incorrect pad curvature, and the first series of constant load followed by load
rejection tests were performed. There was some concern over the proper support
of the transducers mounted above the coupling assembly. In order to obtain the
best possible data, a rigid uni-strut framework was constructed to support the
vibration transducers. The preliminary results of these tests were totally consistent with historical behavior. Vibration amplitudes above the coupling (elevation
4) were the most severe, with typical levels between 40 and 55 Mils,p-p during
load rejection from 8 megawatts. An example of this vibratory behavior is presented in the transient capture plot shown in Fig. 10-20.
Fig. 1020 High Speed

Transient Capture Of Shaft
Vibration At Elevation 4
Above Coupling Assembly
Normal Load Rejection
From 8 Megawatts
This time domain sample is similar to the type of data that can be obtained
with a long sweep rate on an oscilloscope. However this data was digitized and
rapidly sampled with a HP-35665A. The DSA allowed accurate expansion of the
time domain sample, and it was easily determined that the initial vibration
cycles occurred at 60 CPM. This frequency only appeared for a few cycles, and
then the signal blossomed into a brief response at 660 CPM that also appeared
for only a few seconds, and then decayed away. The maximum shaft vibration
response at 660 CPM was 52.7 Mils,p-p. This large shaft vibration amplitude was
visibly distinguishable, and physically threatening. After observing this shaft
vibration and the associated structural motion, most individuals elected to stay
out of the turbine pit during the load rejection tests.
It should also be mentioned that the maximum vibration amplitudes were
not coincident with rotational speed. Hence, traditional synchronous tracking of
the running speed motion would not allow examination of the high amplitude
vibration components. In this case, frequency analysis of the data was manda-
503
Fig. 1021 Waterfall Plot

Extracted From Transient
Capture Of Shaft Vibration
At Elevation 4 Above Coupling Assembly - Normal
Load Rejection From 8
Megawatts
tory. Unfortunately, FFT processing of this transient capture data is very difficult due to the short duration of the specific events, plus the low frequencies
involved. In actuality, if the transient FFT data is not properly handled, the significant information may be lost, distorted, or otherwise corrupted. In all cases,
the diagnostician should verify that the final processed data in whatever format
agrees with the overall time record.
In this specific situation, the transient capture data presented in Fig. 10-20
was post processed in a zero to 3,000 CPM (0 to 50 Hz) span to examine the frequencies of interest. The time record length for a 400 line FFT would be 8 seconds based upon the measurement speed information presented in Table 7-1.
However, if the resolution was decreased from 400 to 100 lines, the frequency
resolution would suffer, but the time record would decrease from 8 to 2 seconds
(=8x100/400). Thus, the FFT data shown in Fig. 10-21 is incremented at 2 second
intervals. Furthermore, this data was processed using a flat top filter to enhance
the amplitude accuracy. The large component at 660 CPM on the waterfall plot
has an amplitude of 47.7 Mils,p-p. This value is consistent with the 52.7 Mils,p-p
overall level displayed on the time domain plot of Fig. 10-20.
It is clear from the overall database that the largest deflections occurred
above the coupling assembly (elevation 4). However, it is meaningful to examine
the behavior at the other measurement locations in a consistent manner. It was
determined that the motion at all elevations was basically forward and circular.
That is, the shaft precession was in the direction of rotation. The amplitudes
504
Chapter-10
Upper
Guide
Bearing
0 Mils,p-p
Upper
Guide
Bearing
(-1.7 Mils,p-p)
139
139
Lower
Guide
Bearing
8.2 Mils,p-p
22
11.9 Mils,p-p
Lower
Guide
Bearing
15.1 Mils,p-p
22
20.9 Mils,p-p
75
75
46.5
Mils,p-p
20.0 Mils,p-p
92
92
Turbine
Guide
Bearing
10.0 Mils,p-p
Fig. 1022 Measured Shaft Mode Shape

At 60 CPM Immediately Following Normal
Load Rejection From 8 Megawatts
Turbine
Guide
Bearing
14.8 Mils,p-p
Fig. 1023 Measured Shaft Mode Shape

At 660 CPM At 12 Seconds After Normal
Load Rejection From 8 Megawatts
were similar from the X and Y probes at each elevation, and the signals were
nominally separated by 90 (i.e., circular). Unfortunately, there were no shaft
vibration measurements down in the flooded portion of the turbine shaft, so the
analysis had to be based upon data obtained from the dry elevations 1 through 5.
Since the shaft motion was forward and circular at 60 and 660 CPM, it
makes sense to average the X and Y data into shaft mode shapes for the exposed
portion of the shafts. The resultant mode shape at 60 CPM was obtained immediately following load rejection, and this information is shown in Fig. 10-22. For
comparative purposes, the mode shape at 660 CPM obtained approximately 12
seconds after load rejection from 8 megawatts is presented in Fig. 10-23.
The higher frequency mode shape at 660 CPM is associated with the fundamental rotor resonance at 660 RPM that was predicted early in the game. This
mode is fundamentally driven by the generator mass, and it is active during all
load rejections. As turbine runaway speed increases with higher megawatt loads,
the excitation for this 660 CPM resonance becomes greater. It is no wonder that
the concrete structure shakes, and windows start breaking when a load rejection
occurs from anywhere near design load.
The lower frequency behavior at 60 CPM was initially somewhat of a mystery. It was determined that this frequency really appeared in the general
domain of 60 to 90 CPM (1.0 to 1.5 Hz). The dichotomy of this response was that
505
the shaft vibration values shown on the mode shape plot of Fig. 10-22 are not
excessive, yet this is the same condition when metal to metal rubbing occurs on
the turbine runner. In essence, it was concluded that a shaft mode exists at nominally 60 to 90 CPM that is driven by the heavy Kaplan turbine overhang. This
mode has large displacements down at the turbine, and much lower displacement levels up at the dry measurement planes. This conclusion was further substantiated by refinement of the analytical model to reveal a turbine conical mode
in the vicinity of 100 RPM.
Throughout the accumulated database it was perfectly clear that high
vibration amplitudes at 660 CPM were always preceded by the activity at 60 to
90 CPM. Although this excitation progression was not immediately understood,
a variety of mechanical changes were implemented. A stiffer generator shaft was
fabricated and installed, and vertical alignment of the entire unit was significantly improved. In addition, the lower guide bearing was set to the correct
clearances, and the turbine guide bearing was replaced with a preloaded bearing.
All of these changes provided a much smoother running machine at steady state
conditions of 277 RPM (at any load up to 18 megawatts).
However, the load rejection behavior from 8 megawatts, and the associated
50+ Mils,p-p of shaft vibration above the coupling remained virtually unchanged.
At this stage, folks were getting distressed. A lot of money had been spent on
mechanical improvements, and considerably more money had been lost due to
the reduced power generation. Fortunately, one of the senior engineers for the
operating company theorized that the problem originated with a strong vortex in
the draft tube during load rejection. If the vortex swirl frequency was in the
vicinity of 60 to 90 CPM, it could easily excite the overhung turbine mode at the
same frequency, and that motion could in turn couple to the second mode at 660
CPM. In support of this hypothesis, the draft tube pressure pulsation during
load rejection is shown in Fig. 10-24. This time domain transient capture
Fig. 1024 High Speed

Transient Capture Of Draft
Tube Dynamic Pressure
Pulsation Following Normal Load Rejection From
8 Megawatts
revealed the presence of a nominal 60 CPM component throughout the majority

of the coastdown. By comparison, the shaft vibration on Fig. 10-20 displays only
a few cycles of 60 CPM activity. Hence, if the draft tube pressure pulsation is due
506
Chapter-10
to a swirling vortex, the energy transfer to the rotor might be sufficient to excite
both rotor criticals including the first mode at nominally 60 CPM, and the second
mode at 660 CPM.
In order to test this hypothesis, a high pressure air injection system was
devised. This system consisted of two pressure vessels that were tied into the
plant air system, and both vessels were pumped up to approximately 100 Psig. A
six inch line was then run from the vessels through a flow meter, and into the
turbine head cover vacuum breakers. This system was automated to dump a regulated flow of air into the turbine, and potentially break up any vortex in the
draft tube. A similar system was tried during initial commissioning of this unit,
but it was limited in capacity and associated flow time. The current system was
sized for the delivery of high pressure air for approximately one minute.
In order to provide a meaningful test, the hydro generator was subjected to
a series of load rejections performed both with and without the air injection. For
safety considerations, the maximum allowable shaft vibration above the coupling
(elevation 4) was limited to 50 Mils,p-p. The summarized vibration data of this
final test is presented in Fig. 10-25. At the first test point of 4 megawatts, the air
injection had little influence. However, as load was increased, the presence of the
air injection became much more significant. In fact, as shown in Fig. 10-25, the
load rejection test from 10 megawatts without air displayed a peak amplitude of
50 Mils,p-p. By comparison, the automated air injection limited the vibration
severity to 32 Mils,p-p. At this point, the tests without air injection were terminated due to the previously mentioned vibration limit. However, the tests with
air injection continued up to full rated capacity of 18 megawatts.
Although 55 Mils,p-p was reached during this final test, the unit was immediately re-rated for full capacity. In retrospect, the injected air successfully eliminated the draft tube vortex, which minimized the rotor resonant excitation at 60
and 660 CPM. The only side effect was a huge downstream air bubble that
relieved itself into the open water channel.
60
Fig. 1025 Shaft Vibration

Measured Above The Coupling During Load Rejections
With And Without Supplemental Air Injection Into The
Turbine Head Cover
Average Shaft Vibration (Mils,p-p)
Without Air Injection

50
40
30
20
10
BJ
B
J
J
With Air Injection
BJ
NOTE: Vibration Amplitudes Represent Average
Between "X" and "Y" Maximum Amplitudes.
0
4
10
12
14
Load Rejection Level (Megawatts)
16
18
507
ELECTRICAL EXCITATIONS
The next category of machine specific excitations considers some of the common characteristics encountered on electrical machinery. This type of mechanical
equipment includes both motors and generators operating in either a synchronous or an induction configuration. This type of machinery consists of a rotor
confined within a stationary stator, and supported by a pair of radial journal
bearings. Due to the self-centering effect of the magnetic fields between rotor and
stator, a thrust bearing is not necessary on horizontal units. However, on vertical
motors and generators, a thrust runner will be installed at the top of the unit.
The associated thrust bearing assembly will often support the entire weight of
the electric machine, plus the coupled unit (e.g. case history 4).
In addition to generic unbalance, eccentricity, resonance, alignment, stability, and assembly problems common to most machines electric machines are
subjected to additional excitations due to the presence of magnetic fields. In
many cases, the dilemma encountered during vibration analysis of electric
machinery occurs with differentiating, or distinguishing, between mechanical
and/or electrical problems. In simple cases, the vibratory evidence clearly points
towards a mechanical, or a purely electrical malfunction. In more complex situations, the mechanical and electrical excitations become intertwined, and the
diagnostician may be misled into a set of erroneous conclusions.
Before examining the telltale characteristics of electrical machinery malfunctions, it is desirable to define the basic differences between synchronous and
induction machines. In most applications, synchronous machines are used for
high horsepower, and slow operating speeds. The rotational speed of a synchronous motor or generator is independent of load, and it is strictly governed by the
line frequency and the number of poles. The following equation (10-42) is used to
determine the synchronous frequency, which is equal to the rotational speed for
these machines:
120 F line
F sync = ----------------------------Np
where:
(10-42)
Fsync = Synchronous Speed (Revolutions / Minute)

Fline = Electrical Line Frequency (Cycles / Second or Hertz)
Np = Number of Poles (Dimensionless Integer)
This equation is widely publicized as the method to calculate synchronous

speed. In reality, this is merely a frequency conversion expression where the line
frequency in hertz (cycles/second) is converted to cycles per minute by multiplying by 60. Division of the frequency in cycles per minute by the number of pole
pairs yields the synchronous speed. In general practice, the actual number of
poles are used rather than the number of pole pairs, and the additional factor of
2 is included. For the normally encountered line frequencies of 60 and 50 Hz,
Table 10-3 summarizes the synchronous speed for various numbers of poles:
Large electric machines are built with a large number of poles, and they
508
Chapter-10
Table 103 Synchronous Speed As A Function Of Number Of Poles And Line Frequency
Number of Poles
60 Hertz
Line Frequency
50 Hertz
Line Frequency
3,600 RPM
3,000 RPM
1,800 RPM
1,500 RPM
1,200 RPM
1,000 RPM
900 RPM
750 RPM
10
720 RPM
600 RPM
12
600 RPM
500 RPM
run at slow speeds. For instance, a 26 pole generator will run at 277 RPM,
whereas a 60 pole unit will only turn at a synchronous speed of 120 RPM.
Machines in this category are typically used as generators coupled to hydro turbines (similar to case history 32).
The second major category of electrical machines consists of induction
motors and generators. These machines do not operate at synchronous speed,
and they change speed in accordance with the load. An induction motor operates
close to synchronous speed under a no-load condition. As load is applied to the
induction motor, the rotating speed decreases. Similarly, an induction generator
operates close to synchronous speed under a no-load condition. As mechanical
power is applied to the induction generator, the rotating speed increases as the
power output increases.
Induction motors and generators are physically similar, and in some cases
identical. For example, in some pump storage facilities, the electric machine is
used as a generator during the day to produce electricity. Each night it is used as
a motor to pump water back to an elevated reservoir (lake).
The difference between rotating speed and synchronous speed is the slip
between the stationary and rotating field. The general equation for determination of slip frequency for an induction machine is presented as follows:
F rotor
F slip = F line 1 ----------------

F sync
where:
(10-43)
Fslip = Slip Frequency (Cycles / Minute)

Fline = Electrical Line Frequency (Cycles / Minute))
Frotor = Shaft Rotating Speed (Revolutions / Minute)
This expression may also be stated in terms of the number of poles by substituting equation (10-42) with consistent units of CPM into (10-43) as follows:
F rotor
F sync N p
Np
F slip = ------------------------------ 1 ---------------- = -------- { F sync F rotor }
2
F sync
(10-44)
509
This synchronous speed is the frequency applied to the stationary stator

coil. The rotative speed is obviously the rotational speed of the rotor. For a two (2)
pole machine, the line frequency is equal to the synchronous speed, and equation
(10-44) simplifies to the following:
2
F slip = --- { F sync F rotor } = { F sync F rotor }
2
As an example, consider a two pole, 60 Hz, induction motor running under
load at an actual running speed of 3,580 RPM. The synchronous speed is 60 Hz,
or 3,600 CPM, and the slip frequency is easily calculated as:
F slip = { F sync F rotor } = 3, 600 3, 580 = 20 Cycles/Minute
The slip frequency is an important parameter, since it provides one of the
tools used to distinguish between various types of malfunctions in electric
machines. Unfortunately, the slip frequency is a fairly small value, and it could
easily go undetected in many types of routine data acquisition and analysis. For
instance, consider the spectrum plot presented in Fig. 10-26. This FFT data was
Fig. 1026 Typical Spectrum Plot Of Electric Motor

Shaft Vibration Signal At
Frequency Resolution Of
30 CPM Per Filter Line
processed with a respectable 400 line resolution across the 12,000 Cycle/Minute
span of this spectrum plot. The frequency resolution for this plot is equal to
12,000 CPM divided by 400 lines, or 30 CPM per line. This FFT resolution does
not provide any visibility of the 20 CPM slip frequency. If this plot was the only
piece of information available, the diagnostician would be hard pressed to accurately diagnose the origin of the single frequency component. However, if the
vibration data is viewed in the time domain, the vibration signal amplitude will
appear to pulsate with passing time. In fact, this same signal will display a distinct amplitude modulation as shown in Fig. 10-27.
This time domain plot covers a total of 8.0 seconds, and the vibration amplitude has passed through many high frequency cycles, plus more than 2 cycles of
510
Chapter-10
Fig. 1027 Typical Time

Domain Plot Of Electric
Motor Shaft Vibration Signal For 8.0 Seconds
a low frequency beat. This low frequency may be determined by measurement of

the period, and computation of the associated beat frequency as follows:
Period = 6.0781 3.0625 = 3.0156 Seconds
From equation (2-1) the period may be easily determined as follows:
1 Cycle
60 Sec.
1
Frequency = ------------------- = --------------------------- ---------------- = 19.897 CPM 20 CPM
3.0156 Sec.
1 Min.
Period
Hence, the beat frequency is equal to the slip frequency of the example
induction motor. Although this activity was not evident in the averaged FFT plot
shown in Fig. 10-26, it is quite visible in the time domain trace. This type of signal was identified as an amplitude modulation. As discussed in chapter 7, the
characteristic beat frequency is the difference between two fundamental frequencies. Thus, further analysis of this signal requires precise identification of
the two signals that contribute to the beat frequency.
Fig. 1028 Translated

Spectrum Plot Of Electric
Motor Shaft Vibration Signal At Resolution Of 0.94
CPM Per Filter Line Displaying A Mechanical Fault
511
It is clear from the FFT plot in Fig. 10-26 that the signal energy is contained around the rotational frequency. To allow detailed examination of this
narrow frequency range, the analysis bandwidth is reduced from 12,000 CPM
(200 Hz) to 375 CPM (6.25 Hz) using the frequency translator capabilities of the
DSA (zoom transform). The frequency content between 3,400 and 3,775 CPM
appears in Fig. 10-28. Note that the majority of the activity (3.55 Mils,p-p) occurs
at the rotational speed of 3,580 RPM. The other frequency component at the line
frequency of 3,600 CPM is only 0.95 Mils,p-p. Hence, in this example, the diagnostician would probably concentrate on examination of mechanical problems
rather electrical phenomena.
The opposite situation is depicted in the translated spectrum plot in Fig.
10-29. Again the frequency content between 3,400 and 3,775 CPM is examined,
and again two major peaks are visible. In this case, the majority of the activity
(3.58 Mils,p-p) occurs at the synchronous line frequency of 3,600 CPM. The small
component is the rotational speed of 3,580 RPM at an amplitude of 0.97 Mils,p-p.
In this case, the machinery diagnostician would concentrate on examination of
electrical rather mechanical problems.
Fig. 1029 Translated

Spectrum Plot Of Electric
Motor Shaft Vibration Signal At Resolution Of 0.94
CPM Per Filter Line Displaying An Electrical Fault
It should be mentioned that both of these example plots would appear identical in the original 12,000 CPM span spectrum plot in Fig. 10-26. They would
appear to be virtually indistinguishable on an oscilloscope time base trace. However, if the scope is triggered with a Keyphasor signal, and the majority of the
motion becomes locked or frozen on the CRT screen, this would be indicative of a
dominant running speed component. Conversely, if the oscilloscope is triggered
off line frequency (i.e., 60 Hz), and the majority of the motion becomes locked on
the screen, the conclusion of a major component at the electrical synchronous
line frequency is correct.
In this example, the beat frequency of 20 CPM is the slip frequency for this
motor. In most cases, the slip frequency does not appear as a separate low frequency component. Usually, the effect of the slip, i.e. the beat frequency, is
512
Chapter-10
observed on the field vibration data. The same scenario applies to induction generators where a full load slip frequency of 30 or 40 CPM will not be directly visible, but an amplitude modulation with a beat frequency of 30 to 40 CPM will be
quite evident in the vibration signals.
Due to the nature of electric machinery, it is often necessary to run additional field tests to determine the origin of a vibration problem. As with most
machines, data should be obtained during initial cold, plus normal hot steady
state data under load. The standard transient startup and coastdown characteristics should also be documented. There are mechanisms that occur during transient speed conditions that are not evident during full speed operation. For
example, synchronous electric motors display an oscillating torque during startup. This torque oscillation may excite any natural torsional resonance(s)
present in the mechanical system. The occurring frequency of this torsional oscillation is easily determined from the following expression:
F tor = N p { F sync F rotor }
where:
(10-45)
Ftor = Torsional Oscillation Frequency (Cycles / Minute)
At full operating speed, the shaft rotational speed Frotor is equal to the synchronous speed Fsync, and the effective torsional oscillation frequency is zero.
However, during startup of a synchronous motor, the torsional excitation frequency decreases as machine speed increases. This oscillating torque is often sufficient to excite the torsional criticals. This is particularly true on a machinery
train that includes a gear box coupled to the synchronous motor. The gear box
provides a natural mechanism to translate the oscillatory torque into a significant lateral vibration. An example of this behavior is shown in case history 34.
It is also desirable to examine other conditions on electric machines. For
instance, a motor may be run solo (i.e., uncoupled), and the shaft vibration data
reviewed during initial cold versus final warm solo operation. Observation of the
coastdown behavior of an electric machine, particularly the response due to a
termination of input power, may be very useful. If the vibration data is tape
recorded in conjunction with a contact closure (e.g., 9 volt battery power source),
the actual sequence of events from the trip point may be accurately determined.
As discussed in other parts of this text, the field tests should be designed to
examine specific mechanisms, or to eliminate potential problems.
Examination of problems such as mass unbalance on electric machines
will be subjected to the same criteria and characteristics discussed in chapter 11.
Obviously, the majority of the shaft vibratory motion should occur at rotative
speed, and it should change in accordance with any system balance resonance(s).
The 1X vibration response should exhibit minimal change with respect to operating temperature or load. Furthermore, any slip frequency beat should be minimal or nonexistent, and a solo power shut off test should reveal no immediate
change as the power is terminated.
It should also be noted that many electric motors are designed as stiff shaft
machines. Motors and generators often contain bearings and couplings that are
513
insulated from ground. This isolates the rotating assembly, and minimizes any
chance of bearing damage due to electrical currents. However, physical degradation of the insulating material, or increased bearing clearances will reduce the
support stiffness of the rotor assembly. In many documented cases, this has
resulted in a decrease of the rotor lateral critical speed back into the operating
speed range. Once this is recognized, replacement of the damaged bearings or
electrical insulators will raise the lateral critical, and restore normal stiff shaft
operation to the electric machine.
An induction motor with broken rotor bars will exhibit a slip frequency
beat to the 1X component when the unit is operated solo. Typically, this beat will
cause minor amplitude changes during a solo run, and the ensuing coastdown
will exhibit only small variations. In most cases, broken rotor bars generally cannot be detected during a motor solo run. However, operating a motor with broken
rotor bars under load will result in a significant increase in vibration amplitudes.
Although it is not necessarily a linear relationship, it has been observed that
increased load will produce increased vibration. In addition, the signal characteristics will change. The dominant vibration amplitude will still occur at running speed, but upper and lower sidebands, at number of poles times slip
frequency, will appear. Thus, for a two pole motor, sidebands at twice slip frequency should appear on both sides of the 1X component. These sidebands will
disappear with power removal.
The problem of high resistance rotor bars can be confirmed by connecting a clip on current probe to one leg of the motor power line. The current signal
is then subjected to a spectrum analysis similar to Fig. 10-29. If high values of
the slip frequency side bands (and slip frequency harmonics) around the line frequency are observed, the problem is generally associated with rotor bars. Obviously, the final condition of this behavior is the development of broken rotor bars.
One of the most common problems on electric machines is the mis-positioning of the rotor within the stator to yield an uneven air gap between the rotor
and stator. On electric machines, a uniform radial air gap around the circumference of the rotor is considered mandatory. If this air gap is distorted by misalignment of the rotor, bearing damage, deformed frame and stator coil, or any other
physical mechanism the vibratory behavior of the electric machine will suffer.
Assuming a concentric rotor, the primary characteristic of an uneven air gap is a
significant excitation at twice line frequency. Thus, a 60 Hz induction motor
would display a major vibration component at 7,200 CPM (irrespective of the
number of poles). This excitation would appear under solo operation at no load,
plus fully loaded conditions. Since this excitation is due to variations in the magnetic flux between rotor and stator, the twice line frequency vibration component
will disappear as soon as the power is cut off (solo or loaded). This dynamic
behavior is mimicked by an electric machine with unbalanced line voltages.
The difference between these malfunctions is usually identified by measuring
the voltages and currents for each respective phase.
If an eccentric rotor is combined with a round stator that is concentric
with the rotor centerline, another type of air gap fault in encountered. In this situation, the rotational speed vibration is high, and it is generally modulated by
514
Chapter-10
the slip frequency. In some motors, the modulation occurs at twice the slip frequency, and a discrete frequency component at twice slip speed may be visible.
This eccentric rotor behavior appears during unloaded solo operation as well as
fully loaded and coupled. The modulation does disappear immediately upon cutting power in either the solo or the loaded condition.
An eccentric rotor should be readily detectable during rotor assembly, and
final runout checks. Furthermore, any type of air gap variation between rotor
and stator should be identifiable by measuring the assembled machine air gaps.
This static measurement is obtained with feeler gauges during final assembly of
the electric machine. On small units, it may only be possible to obtain air gap
measurements every 90 (4 points). On larger machines, it is desirable to obtain
these readings at 8 points (every 45), or at 12 points (every 30). These air gap
measurements should be taken at both ends of the electric machine, and they
should be retained as part of the maintenance documentation.
The question of correct radial rotor positioning, and the acceptable variation of air gap measurements is often difficult to address. It is generally troublesome to make precise air gap measurements due to the physical construction of
some machines. Hence, the machinery diagnostician is referred back to the original OEM specifications for guidance in this critical area. Since some specifications change with time, it might also be advisable to contact the OEM directly for
an update of current procedures and tolerances.
In years past, the requirement for variable speed electric machines has
been satisfied by large DC units. The number of these units are steadily fading
due to attrition, and DC electric machines are seldom encountered within the
industrial community. In situations where variable speed motors are required
due to process considerations or energy conservation, the trend has been towards
variable speed AC motors.
From equation (10-42) it is apparent that synchronous speed is dependent
on the line frequency, and the number of poles. Since the number of poles is fixed
for any particular machine configuration, the only way to vary the motor speed is
to change the line frequency. In fact, this is exactly what the newest motor control systems accomplish. These motor control devices are often called frequency
converters, or Variable Frequency Drives (VFDs). Functionally, they vary the
input frequency to the motor to provide variable speed operation.
These variable speed AC motors add another degree of complexity to the
business of analyzing motor vibratory characteristics. Since the motor stator frequency is now a variable parameter instead of a constant 60 Hz, the diagnostician must be even more careful during data acquisition, processing, and
analysis. In some cases, the inclusion of motor currents via clamp-on transducers, plus coil temperature distributions may be highly beneficial in identifying
and solving a problem on an electric machine. As always, the shaft motion and
position should be observed. On flexible supports, the bearing housing motion
should be measured, and good engineering judgment applied.
Finally, the views and opinions on electric machines presented in the last
few pages are based upon a variety of field experiences by the senior author.
Within this chapter there has been no effort to provide any detailed explanations
515
of the electrical forcing functions. This is a complex topic, and the reader is referenced to documents such as the tutorial on motors by James Baumgardner12.
Within this article, Baumgardner provides a detailed discussion of the vibration
characteristics of three phase, squirrel cage, induction motors. The explanation
of motor electrical characteristics and forces are both clear and accurate.
The reader is also cautioned against fully believing some of the motor analysis charts and tables that appear in the literature. Some of these guides contain
considerable inaccuracies, and they can result in more confusion than positive
assistance. As a general rule, the technical information provided by the OEMs
and knowledgeable end users will typically be solid and reliable troubleshooting
information. Some of the other published sources should be used carefully. Dont
get stuck in the groove of trying to find a troubleshooting chart that matches
your problem symptoms.
Case History 33: Motor With Unsupported Stator Midspan
Many large motors have long rotors, and the companion stators are equally
endowed with physical length. Many of these units are built with support rails
that run the full length of the casing, and are designed for support along the
entire length of the stator. For example, consider the induction motor driving single stage booster compressor depicted in Fig. 10-30. The 3,000 HP motor was
only supported at the four corners. During uncoupled operation, the shaft vibration (from probes 1Y through 2X) was dominated by rotational speed motion at
an average value of 1.2 Mils,p-p for all four probes. The shaft runout varied
between 0.2 and 0.3 Mils,p-p, and it was generally in-phase with the synchronous
vectors at full speed. Hence, the runout compensated shaft vibration was approximately 1.0 Mil,p-p at all four measurement points. For a 3,600 RPM motor this
was somewhat higher than desirable for an unloaded spin test, but it was
accepted by the end user.
Installation of this machinery train was complicated by a poor grouting job.
The initial problems were eventually corrected, and a proper full contact epoxy
grout was used between the foundation and the sole plate. Examination of the
transient startup behavior revealed nothing unusual. However, as load was
applied the shaft vibration began to increase, and it eventually reached amplitudes in the vicinity of 3.0 Mils,p-p. Furthermore, the vibration levels did not
remain constant, and they continually pulsated back and forth between 3.0 and
0.5 Mils,p-p. As observed on an oscilloscope, the behavior was clearly an amplitude modulation similar to the diagram shown in Fig. 10-27. With a stopwatch, it
was determined that the vibration amplitudes were continually changing at a
rate of every seven seconds. That is, the time from one peak amplitude, through
the vibration decrease, and back to the next peak amplitude was approximately
seven seconds. Simultaneously, the rotational speed phase angles varied by nom12 James Baumgardner, Tutorial Session on Motors, Proceedings of the Eighteenth Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University System, College Station,
Texas (October 1989).
516
Chapter-10
Discharge
K
Induction Motor - 2 Pole
1Y
Rated 3,000 HP @ 3,580 RPM
Booster
Compressor
4,160 Volts - 3 - 60 Hz
1 Stage
1X
2Y
45 45
CW
Suction
2X
45 45
Rotation and Probe

Orientation Viewed
From Motor Outboard
3Y
3X
45 45
4Y
4X
45 45
Fig. 1030 Machinery Arrangement For Motor Driven Single Stage Compressor
inally 100 in a consistent manner with the oscillations in shaft vibration. This
behavior occurred at a running speed of 3,591 RPM. From equation (10-42), it
was clear that the slip frequency for this two pole motor was 9 cycles per minute.
The period of 9 CPM is 6.7 seconds. Hence, the observed amplitude and phase
variation period of approximately seven seconds was really the period of the
motor slip frequency.
Another perspective of the dynamic motion is presented in the orbital data
in Fig. 10-31. This information consists of eight shaft orbits acquired at 1 second
intervals. The data was band-pass filtered at the running speed of 3,591 RPM.
However, the bandwidth of the filter included the motor line frequency at 3,600
Fig. 1031 Consecutive Shaft Orbits At Induction Motor Outboard Bearing
517
CPM (60 Hz). Note that rotation is clockwise as viewed from the outboard end of
the motor towards the compressor. For normal machinery behavior, the shaft
precession should be in the same direction as shaft rotation. However, orbits
and in Fig. 10-31 exhibit a distinct reverse precession. This is followed by
orbits through with a normal clockwise precession. If this data was viewed
live on an oscilloscope, the shaft orbit would be in constant motion, and it would
consecutively repeat the patterns documented in Fig. 10-31. This dynamic behavior is due to the interaction of two closely spaced frequencies.
The respective amplitudes at 3,591 and 3,600 CPM cannot be properly identified with a 12 CPM band-pass filter. It is necessary to employ the frequency
expansion capabilities of a DSA over a suitably small frequency span. In this
case, a 6.25 Hz (375 CPM) span was selected between 3,400 and 3,775 CPM. For
a 400 line resolution, the time record length was 64 seconds, and the resolution
for a 400 line display would be 0.9375 CPM per filter (from Table 7-1). Perform
this analysis for 4 averages, the results are summarized in Table 10-4.
Table 104 Summary Of Motor Vibration Amplitudes At Running Speed And Line Frequency
Motor Vibration Probe
Amplitude At
Running Speed
of 3,591 RPM
Amplitude At
Line Frequency
of 3,600 CPM
Outboard - 1Y
1.64 Mils,p-p
1.59 Mils,p-p
Outboard - 1X
1.88 Mils,p-p
1.68 Mils,p-p
Coupling - 2Y
1.17 Mils,p-p
1.63 Mils,p-p
Coupling - 2X
0.91 Mils,p-p
1.67 Mils,p-p
The running speed vibration amplitudes at 3,591 RPM are higher than the
solo levels of nominally 1.0 Mil,p-p. This could be due to a variety of reasons, but
the significant result of Table 10-4 is the constant excitation of 1.6 Mils,p-p at the
synchronous or line frequency of 3,600 CPM. Based on the previous discussion
within this chapter, the diagnostician would certainly suspect an electrically or
magnetically induced excitation. Since the 3,600 CPM shaft vibration amplitudes are essentially the same at both ends of the machine, it is reasonable to
expect a mechanism that influences the entire rotor in a uniform or evenly distributed manner. The slip frequency behavior of this motor does not directly
match any of the traditional symptoms mentioned earlier in this chapter. The
documented assembly information on this motor revealed proper bearing clearances, uniform air gap at both ends of the motor, plus a reasonable alignment to
the single stage booster compressor. Overall, there was no evidence in the
mechanical assembly information to indicate any abnormalities.
A wise mechanical engineer once told the senior author that you will never
understand a machine unless you go and look at it. This is certainly an unpopular approach in the modern world of powerful computers, remote information
518
Chapter-10
Motor Stator
Motor
Outboard
To Coupling &
Compressor
Motor Rotor
Fig. 1032 Diagram Of

Induction Motor With
Exaggerated Stator Sag
Baseplate and Pedestals
transfer, and beautiful graphics generated with a variety of analytical and data
processing programs. However, even advanced 1s and 0s technology must
always succumb to the realities of the physical installation. In this particular
case history, a walk around the unit revealed that the motor was only supported
at the four corners. Although a full length support surface under the stator was
provided by the OEM, over 90% of this vertical support surface was unused.
Additional field optical measurements revealed that the center of the stator
support was about 15 Mils lower than the ends. In essence, the diagram shown
in Fig. 10-32 describes the rotor and stator position with only end supports for
the stator. In this condition, the air gap was not uniform along the length of the
rotor. That is, a straight rotor was running within the confines of a dropping stator. The uneven magnetic forces due to this distorted air gap were primarily
responsible for the 1.6 Mil,p-p vibration component at 3,600 CPM. Placing a midspan support under the stator relieved the problem, and 60 Hz excitations were
reduced to levels of less than 0.2 Mils,p-p at both motor bearings. This midspan
stator support also resulted in a reduction of the 1X running speed vibration
amplitudes.
As a side note to this problem, recall that the air gap measurements at each
end of the motor were acceptable, and within the OEM specifications. As shown
in the exaggerated stator deflection diagram of Fig. 10-32, it is quite possible to
have reasonably even air gaps around the ends of the rotor combined with an
eccentric air gap at the center of the rotor. In this situation, the rotor appears to
be properly centered in the stator, but midspan deflection of the stator due to its
own weight results in an uneven air gap problem at the center of the rotor.
In virtually all cases, the OEMs recommendations for machine support
should be followed. If this information is not supplied in the installation manual,
then the OEM should be contacted, and requested to provide their recommended
support configuration for the model and size of the machine under consideration.
If machine support issues still exist after talking to the OEM, then best engineering judgment should be applied. Remember that few machines have ever
experienced problems due to large foundations and rigid supports. However,
many machines have been damaged or destroyed due to poor foundations,
improper supports, or insufficient supports.
519
Case History 34: Torsional Excitation From Synchronous Motor

The machinery train shown in Fig. 10-33 consists of a 6 pole, 2,500 HP synchronous motor driving a speed increasing gear box. The pinion output speed is
8,016 RPM, and it is directly coupled to a 5,000 HP hot gas expander turbine.
The combined power output from the motor and the expander are used to drive
the process air compressor located at the end of the train. Of particular note on
this unit is the coupling between the motor and gear box. This 1,200 RPM grease
packed gear coupling has a high torsional stiffness of 110,000,000 Inch-Pounds
per Radian. This machinery train has been in service for many years, and high
vibration amplitudes are generally encountered during startup conditions on the
pinion and the adjacent expander bearing.
Exhaust
Discharge
Suction
CW
Normal
Thrust
6X

Viewed From The Motor Outboard
Km
Synchronous
Motor
Kc
45
8,016 RPM
Hot Gas
Expander
Gear
Box
5,000 HP
Process Air
Compressor
6.680:1
2,500 HP - 6 Pole
4,160 Volts - 3 - 60 Hz
1,200 RPM
7Y
7X
45 45
Hot
Inlet
8Y
8X
45 45
9Y
9X
45 45
10Y
10X
45 45
CCW
135
1Y
135
2Y
135
CW
3Y
Fig. 1033 Machinery Arrangement Of Motor, Gear Box, Expander, And Compressor
After the plant is up and running, and a full heat soak has been achieved by
all of the cases, shaft vibration levels are generally low and acceptable. However,
the initial cold startup generates significant noise in the gear box, plus the high
transient vibration levels previously mentioned. Fortunately, the high startup
pinion vibration levels only affects bearing 7 (probes 7Y and 7X). Expander bearing 8, and the entire air compressor are generally not influenced by the high pinion vibration amplitudes.
A typical pinion startup is characterized in the transient capture data
exhibited in Fig. 10-34. The entire startup from zero to full speed is achieved in
slightly less than 23 seconds. It is clear from the time domain trace of Fig. 10-34
that an initial excitation occurred shortly after rolling, and a major response of
6.58 Mils,p-p occurs at approximately 15 seconds into the ramp-up. Although the
overall severity of the pinion vibration is described by Fig. 10-34, this type of
information does not provide much visibility as to the origin or frequency composition of the vibration signal.
If the same data is processed in the frequency domain, the cascade plot
520
Chapter-10
Fig. 1034 Transient Capture Time Domain Plot Of

Pinion Vibration During A
Typical Startup
shown in Fig. 10-35 may be produced. This diagram displays a series of spectrum
plots at 200 RPM increments between 2,000 and 8,000 RPM. At a pinion speed of
2,000 RPM, the bull gear and motor speed is 300 RPM. At the full pinion speed of
8,016 RPM, the associated motor and bull gear speed is equal to 1,200 RPM.
These two end points are connected, and the line labeled as bull gear and motor
speed on Fig. 10-35. It is clear that throughout the majority of the speed range,
the rotational speed vibration amplitudes at motor and bull gear speed are negligible.
The same argument may be applied to the amplitudes associated with the
pinion rotating speed. Since the cascade plot shown in Fig. 10-35 is limited in fre-

of Pinion Vibration During
A Typical Startup
521
quency span to 6,000 CPM, the pinion rotational frequency line only extends
from 2,000 to 6,000 RPM. This line is labeled as pinion speed, and it is clear that
vibration amplitudes along this line are minimal.
The major activity throughout the speed domain occurs along a line at 900
CPM (15 Hz). This is the first torsional resonance of the machinery train, and it
initially appears between pinion speeds of 2,400 and 4,000 RPM. Logically, this
is the initial amplitude burst shown in the first few seconds of Fig. 10-34. The
next, and largest component on Fig. 10-35 occurs at the intersection of 900 CPM
and the line labeled as the oscillating torque frequency. The maximum component amplitude at this intersection is 5.0 Mils,p-p that is consistent with the peak
value of 6.58 Mils,p-p displayed on Fig. 10-34.
This excitation is the pulsating torque originating from the synchronous
motor. The resultant torsional frequency generated conforms to equation (10-45).
In this particular case, the number of poles Np is equal to 6. The synchronous
speed of the motor Fsync is 1,200 RPM, and the torsional resonant frequency Ftor
is 900 CPM. If these values are substituted into (10-45), the expression may be
solved for the motor rotor speed Frotor that provides a pulsating torque at the torsional resonant frequency of 900 CPM.
F tor
900 CPM
F rotor = F sync ----------- = 1, 200 CPM ---------------------- = 1, 200 150 = 1, 050 RPM
Np
6
Thus, at a motor or bull gear speed of 1,050 RPM, the oscillating torque frequency will be 900 CPM. Multiplying the bull gear speed of 1,050 RPM by the
gear box speed ratio of 6.68 yields a pinion speed of 7,014 RPM. Clearly, this is
consistent with the data displayed in the cascade plot of Fig. 10-35.
Simultaneous with this high startup vibration on the pinion, the motor and
bull gear reveal only minor traces of this torsional motion. It is reasoned that the
oscillating torque from the motor is transmitted directly through the torsionally
hard coupling to the bull gear. Since there is minimal torsional to lateral crosscoupling in either the motor or the bull gear, the lateral vibration at the torsional
frequencies are minimal. However, the tooth contact forces transmitted across
the mechanical link between the bull gear and the pinion teeth contain a vector
component that translates torsional to lateral motion. Hence, the pinion is forced
to vibrate at the oscillating motor torque frequency. This is detrimental for both
the pinion and bearings, plus the expander bearing located across the high speed
coupling. During one documented startup, new expander bearings with a 6 Mil
diametrical clearance experienced a shaft centerline position change of 9 Mils.
Hence, the high torsional startup vibration effectively hammered the expander
bearing, and increased the clearance by 50%.
This situation could be significantly improved by reducing the torsional
stiffness of the coupling between the motor and bull gear. This would isolate the
oscillating torque behavior to the motor, and would probably influence the effective torsional resonant frequency. However, due to years of successful operation,
the end user elected not to correct this problem.
522
Chapter-10
RECIPROCATING MACHINES
Reciprocating machines have been in service for many years prior to the
introduction of centrifugal units. In many circles, reciprocating machinery is considered to be less sophisticated, or perhaps less elegant than centrifugal units. In
reality, reciprocating machines are more complicated, and contain more individual parts than centrifugal units. Reciprocating machine installations often
require customized piping simulation, and pressure pulsation suppression
devices such as pulsation bottles or restriction orifices. This technology is seldom
required on centrifugal installations. In many cases, reciprocating machines
demand larger foundations, and they are less tolerant of supporting structure
degradation. In most facilities, the long-term maintenance costs on recips are
significantly greater than centrifugals.
In spite of these drawbacks, the total quantity of positive displacement
reciprocating engines, pumps, and compressors exceed the number of centrifugal
units. From an application standpoint, there are many situations where centrifugal units cannot provide the necessary differential head, or they cannot efficiently operate over the required flow rates demanded by the process. In these
cases, reciprocating machines are far superior to their centrifugal counterparts,
and the advantages strongly outweigh the obvious disadvantages.
As mentioned throughout this text, the technology required for the measurement and diagnosis of machinery problems on centrifugal units has progressed at a rapid rate during the past three decades. Unfortunately, there has
not been similar progress for the analysis of reciprocating machines. In the
majority of cases, the reciprocating machinery condition is evaluated based upon
process measurements such as suction and discharge temperatures and pressures. These measurements are often supplemented by periodic external measurements such as external valve temperatures, or casing vibration of the frame
or crankcase.
Measurement and trending of valve temperature has proven to be an
effective predictive tool on many reciprocating compressors. In these units the
valves are accessible from the outside of the cylinder, and an in-operative valve
assembly may be easily replaced. This configuration applies to plate valves, poppet, or channel valves. Some machines are equipped with permanently mounted
temperature sensors on each valve, plus an associated scanning recorder. This
allows the trending of valve temperatures with time, and provides a good perspective of valve problem development.
Additional perspective of cylinder behavior may be obtained by running a
pressure-volume curve (PV diagram, also know as an indicator card). The volume is determined by the stroke position, and cylinder pressure is obtained with
a dynamic pressure sensor. This provides a good overview of the entire expansion-compression cycle, and problems such as valve chatter and piston ring rattle
are discernible. It is also meaningful to observe the time domain pressure pulsation data, and examine the pressure pulsations as a function of both time and
stroke position. However, any type of pressure measurement on a cylinder
requires a direct physical connection between the interior portion of the cylinder
Reciprocating Machines
523
and the externally mounted pressure pickup. In some cases, this is a drilled hole
through the cylinder wall, with a tapped external connection for a pressure
pickup. This type of installation minimizes the passage length between the cylinder and the transducer, but it does necessitate shutting down the machine for
installation and removal of the pressure pickup. In other installations, a nipple
and block valve are screwed into the hole that is drilled into the cylinder wall.
This approach allows the pressure pickups to be installed and removed from the
machine during operating. However, equipping a large machine with ten or
twenty small overhung valves may be undesirable from a safety standpoint.
These small valves and associated nipples are subject to mechanical damage
from people working around the machines, and they are also prone to fatigue
failures due to vibration of the cantilevered valves. These type of measurement
ports may be acceptable for air or nitrogen compressors, but they should be cautiously applied on reciprocating units handling combustible materials.
On some reciprocating engines, it is possible to purchase spark plugs with
pressure pulsation probes that are an integral part of the spark plug. This type
of installation is very effective to study detonation or misfiring problems in the
engine. In some cases, it is also meaningful to measure casing vibration in conjunction with the pressure pulsation data.
Frame or crankcase vibration measurements are common on reciprocating units. These may be periodic measurements obtained with a portable data
collector, or they may be permanently installed casing transducers connected to a
monitoring system, computer-based DCS, or machinery trending system. Before
the appearance of cost-effective piezoelectric transducers, these external frame
measurements on recips were generally obtained with velocity coils. Unfortunately, the high vibration levels typically encountered on reciprocating machines
results in premature failure of the coil springs in a velocity pickup. Hence, one
day the velocity coil has an output of 0.12 IPS,o-p, and the next day the level has
dropped to essentially zero. In this example, the velocity coil has failed, and operations personnel become further convinced that this stuff doesnt work.
With the advent of cost-effective piezoelectric accelerometers, and integral
transducer electronics to integrate the acceleration signals to velocity, many of
the poor measurement reliability issues have been corrected. Hence, solid state
vibration transducers may be installed on the external frames or crankcases of
reciprocating machines, and they will successfully operate for extended periods
of time. However, most end users are still reluctant to use this type of vibration
measurement for automatic shutdown of the machinery.
A more meaningful and reliable vibration measurement involves the use of
proximity probes. These transducers are mounted on stationary parts of the
machine, and they observe the relative motion of moving machine elements. For
example, a compressor drive motor such as the 200 RPM synchronous motor
shown in Fig. 10-36 (case history 35) may be equipped with X-Y proximity probes
to observe the motor shaft vibration relative to the stationary bearing housings.
This application is obviously identical to the approach used on any other piece of
critical centrifugal machinery. The concept of measuring rotational motion may
be extended to the crankshaft main bearings. In some installations, X-Y proxim-
524
Chapter-10
ity probes have been installed to measure crankshaft vibration relative to the
stationary main bearings. The axial clearances around these main bearings are
often limited, and the mounting of proximity probes may be quite difficult.
Although there are benefits to be gained from this measurement of crankshaft
vibration and position, the probe installation is often challenging. In some cases,
it is possible to install the probes in the bottom half of the main bearing, and
route the extension cables out the side of the crankcase. This location may
require boring through the babbitt, and the OEM should be consulted on any
potential adverse effect this may have upon the bearing.
From the crankshaft, connecting rods may go directly to the individual pistons, or they may be attached to a crosshead. This is the point in the machine
where the rotational motion is converted to reciprocating motion. In the compressor case shown in Fig. 10-36 (case history 35), a main crosshead is mounted
at the crankcase for each throw. An auxiliary crosshead is then driven back and
fourth by a pair of horizontal drive rods from each main crosshead.
At this point, the primary vibration measurement consists of proximity
probes mounted on the seals or cylinders observing the reciprocating piston
rod or plunger. In a simple installation, one vertical probe is mounted on each
rod or plunger. This transducer may be directly above, or directly below, the moving element, but it is important that the probe be in a true vertical direction. By
observing and trending DC gap voltages from the proximity probe, it is possible
to measure the average horizontal running position of the piston rod. As wear
occurs on the piston rings due to normal attrition, the piston rod will move downward, and the vertical proximity probe will detect this event by virtue of a
change in probe gap voltage. If the probe is mounted above the piston rod, the
DC gap voltage will increase with piston ring wear. Conversely, the gap voltage
will decrease if the vertical probe is installed below the piston rod. This is sometimes referred to as the rod drop measurement, and it provides actionable
maintenance information on this part of the reciprocating machine.
Additional machinery information may be acquired by installing a horizontal probe in conjunction with the true vertical probe as shown in Fig. 10-36. This
orthogonal transducer provides information in the horizontal plane, and in many
cases this may be more significant than the vertical motion. Furthermore, the
use of perpendicular vertical and horizontal proximity probes allows the machinery diagnostician the capability of observing the combined vertical and horizontal motion of the piston rod or plunger as function of stroke. More specifically, on
a rotating machine, the probes observe the same circumferential shaft surface as
the shaft rotates, and it is customary to examine the shaft orbit. This orbital
motion may be examined during a single turn, or for multiple rotations. On a
reciprocating machine, the probes observe the same angular position on the rod
or plunger as the element reciprocates back and fourth. When observed from the
end of the cylinder, the resultant motion is a Lissajous figure of the rod centerline during one or multiple strokes. In essence, this is a two-dimensional representation of a three-dimensional event.
The sequential timing of events throughout the machine is achieved by a
once-per-cycle Keyphasor probe. This timing transducer may be installed on
525
the motor, crankshaft, or one of the drive rods (as shown in Fig. 10-36). To make
the analysis of data as straightforward as possible, it is always desirable to have
the trigger point coincident with the top dead center of one of the cylinders
within the machine. By positioning this timing probe with respect to a physical
event on one cylinder, the entire machine may be phase related.
The electronic tools used to diagnose the behavior of rotating machines are
fully applicable to the analysis of the transducer signals generated by reciprocating units. Although the mechanics are different, the processing techniques
applied to the electronic signals are virtually identical. From a frequency analysis standpoint, the dominant frequency observed on a reciprocating machine is
the fundamental speed of the machine. For instance, on the machine depicted in
Fig. 10-36, the motor speed is 200 RPM, and this is directly coupled to the main
crankshaft at 200 RPM. The main and auxiliary crossheads, plus the plungers
all reciprocate at a frequency of 200 CPM. Since a constant frequency exists
throughout the machine, this may complicate any malfunction analysis, since
everything is moving at the same frequency. In addition, there are often other
cylinders, and other throws that are generating the same frequency, and a variety of interactions plus rotational speed harmonics are often observed.
This measurement of piston rod or plunger motion provides considerable
information on the dynamic behavior of the mechanical system. Deflection of the
rod or plunger, plus the influence of various preloads, are detectable with these
displacement transducers. This dynamic measurement is sometimes referred to
as rod runout, which is a misnomer. Basically, these machine elements do not
have runout, but they do respond to changes in position and dynamic forces. The
timing relationships between cylinders, and the position of each individual piston rod or plunger with respect to top dead center and bottom dead center are
important parameters. With some malfunctions, the relative position of elements
are important. With other mechanical problems, the Lissajous patterns
described by the V-H probes provide the necessary clues. In still another class of
abnormal behavior, the simple time domain examination of the proximity probe
signals will provide useful information. By understanding the characteristic
behavior of these dynamic signals, the diagnostician will have additional useful
tools for meaningful evaluation of the machinery.
Lower pressure reciprocating machines are equipped with piston rods, pistons, piston rings, plus a variety of seal configurations. These machines may also
have manual or automatic valve unloaders, and other appurtenances such as cylinder clearance pockets. These general types of machines may be difficult to diagnose simply due to the mechanical complexity and the associated multiple
degrees of freedom of the unit. A somewhat simpler version of the reciprocating
machine is the high pressure hyper compressor discussed in case history 35. In
this type of machine, a solid tungsten carbide plunger is used as a combined piston rod, piston, and piston rings. This plunger is connected directly to the auxiliary crosshead, and it functions as a rigid member that reciprocates in and out of
the cylinder with stationary packing. This type of assembly does simplify some of
the system dynamics, but other issues do develop that make the examination of
this type of machinery quite interesting.
526
Chapter-10
Case History 35: Hyper Compressor Plunger Failures

The machinery depicted in Fig. 10-36 represents one of four trains installed
in a new low density polyethylene plant. The original plant contained a total of
four processing lines. Each line consisted of a primary compressor, a high pressure hyper compressor, plus the reactor and associated product handling equipment. The primary compressors in each line are traditionally configured
reciprocating machines that boost the ethylene pressure to nominally 4,000 Psi.
Depending on product mix, the high pressure compressors provide a reactor inlet
pressure that varies from 30,000 to 45,000 Psi. Obviously, it doesnt matter if
these pressures are specified as gauge or absolute, these are very high pressures.
In order to meet these pressure demands, the high stage compressors are a
unique variety of machine that was specifically designed for ultra high pressure
operation. Although the drive, crankcase, and main crosshead arrangements are
fairly typical of most reciprocating machines a significant change in technology occurs in the auxiliary crossheads, plungers, and cylinders. In this type of
machine the piston rod, piston, and piston rings are replaced with a solid tungAY
AX
45 45
CCW
Synchronous Motor
Main Crosshead
10,000 HP - 36 Pole - 200 RPM

13,200 Volts - 3 - 60 Hz
BY
BX
45 45
Auxiliary Crosshead
Stationary Tie Rods
Cylinder 1
4th Throw
Suction
Cylinder 4
Typical
Cylinder
Piping
Cylinder 3
Discharge
4V
90
Typical
Probe
Installation
2nd Throw
Crank Case - 5 Throws
Cylinder 5
Cylinder 6
Tungsten Carbide Plungers
Cylinder 7
Cylinder 8
Reciprocating Drive Rods
5th Throw
Cylinder 2
1st Throw
3rd Throw
Auxiliary
Crosshead
with Slipper
Cylinder 9
Cylinder 10
4H
Fig. 1036 Plan View Of High Pressure Reciprocating Compressor And Motor Driver
527
sten carbide plunger. Two plungers are attached to each auxiliary crosshead with
a patented resilient connection that incorporates a spherical seat joining into a
thrust block. These plungers are mated with high pressure cylinders that are
mounted on each auxiliary crosshead yoke assembly as shown in Fig. 10-36. The
44.5 inch long plungers have diameters that vary from 2.25 to 3.75 inches, with a
total stroke of 15 inches. It is meaningful to note that the tungsten carbide
plunger material has a modulus of elasticity of 80,000,000 pounds per inch2. The
tensile strength is equal to 120,000, and the compressive strength is 570,000
pounds per inch2. This is also a very hard material with Rockwell C values in the
vicinity of 84. Tungsten carbide will handle enormous compressive loads, but it is
much weaker when subjected to tension or bending. In fact, the remnants of
plunger failures revealed a brittle fracture of the material.
The suction and discharge piping to each cylinder consists of high pressure
tubing. In many respects this piping appears to be a cannon bore construction
with large outer diameters, thick walls, and small inner diameters. As shown in
Fig. 10-36, the suction piping enters at the outboard end of each cylinder. The
discharge exits from a connection at the cylinder inboard, close to the stationary
yoke assembly. During operation, the cylinders visibly moved back and fourth
with each stroke.
The plunger in each cylinder is equipped with several rows of segmented
bronze packing that ride directly on the tungsten carbide plungers. The amount
of gas leakage past the seals is monitored, and historically this has been used as
a partial indicator of packing condition. These units also incorporate plunger
coolant circulation around the outer plunger seal assembly.
These unique machines were sequentially started as the construction and
commissioning on each line was completed. In the spring of the startup year, all
four lines were successfully operating, and the prognosis of continued operation
seemed to be excellent. However, in June of that same year a series of three
plunger failures occurred within an 18 day period. Total machine operating time
to each of the failures varied between 2,400 and 3,000 hours. Fortunately, the
first two failures just involved equipment damage. However, the third broken
plunger resulted in personnel injuries in addition to fire and mechanical damage.
This type of situation in a chemical plant is difficult for all parties involved.
A variety of activities are immediately initiated to determine the root cause of
the failures, plus the implementation of suitable corrective and preventative
measures. In addition to the complex technical problems, and the personnel protection issues, the plant was faced with a termination of fire insurance on these
machines. In the event of a fourth failure, and the possibility of a major fire, the
personnel hazard and financial implications were potentially devastating.
In all three mechanical failures, a plunger broke, and high pressure ethylene was released to the atmosphere. One of the candidates that could be responsible for the plunger breakage was misalignment of the plungers to their
respective cylinders. Although each plunger was carefully aligned with dial indicators, this only covered the static condition as the machine is barred over. There
was no method to measure the plunger position during actual operation of the
528
Chapter-10
machine. It was suggested that air gauges be directed at the plungers, but this
concept was rejected due to the limited range of this measurement (20 Mils). The
only dynamic transducer that made any sense was the proximity probe.
Although this sensor would measure the average distance between the probe and
the plunger (DC gap), it would also measure the reciprocating motion of the
plunger (AC portion of the signal). Initially, it was believed that the reciprocating
motion was minimal, and the majority of the information would be derived from
changes in probe DC gap voltage with the machine in full operation.
At this point in history, only a handful of people had ever installed proximity probes on reciprocating machines. Most of the industrial contacts that had
attempted this measurement were either working on considerably different
machines, or they were extremely reluctant to share their experiences. Hence,
the installation of proximity probes on tungsten carbide plungers was virtually a
new application of an existing technology by the senior author.
The first step in this investigation consisted of determining if proximity
probes will work with tungsten carbide. To verify this point, and allow the development of calibration curves, a test fixture was constructed to hold a proximity
probe plus a section of a fractured plunger. It was quickly determined that the
measurement was quite possible, and a normal calibration curve could be generated from this dense material. On typical compressor shafts made out of 4140
steel, the proximity probes yield a calibration of 200 millivolts per Mil. However,
on tungsten carbide plungers, this value increased to 290 millivolts per Mil. Furthermore, the resultant calibration curves were exceptionally clean, the points
easily fell into a straight line, and consistency between plungers was exceptional.
In retrospect, this was logically due to the high quality of the plungers, and the
uniformity of the metallurgy.
Since the proximity probe measurement worked in the shop, it was now
necessary to determine if probes could be installed on the compressors. As it
turned out, the transducers were mounted on the outer packing flange. One
probe was installed in a true vertical direction, and the companion probe was
mounted in a true horizontal direction. A typical probe installation is shown in
Fig. 10-36 with the vertical transducer at the 12 oclock position, and the horizontal probe mounted at the 3 oclock location. Due to various interferences with
existing hardware, this typical probe installation was occasionally changed by
installing the vertical probe below the plunger looking up, and/or mounting the
horizontal probe on the left side of the plunger. In all cases, the proximity probes
were maintained in a true vertical and a true horizontal orientation.
During the first test of this instrumentation, only six probes were installed
on three different plungers. Temporary Proximitors were connected to the
probes, and a jury-rigged power supply and patch panel were used to drive the
probes. The test equipment at this stage consisted of an oscilloscope, an oscilloscope camera, a digital voltmeter, and a four channel FM tape recorder. The
ensuing startup was quite remarkable since it was clear that DC probe gap was
useful information, but the AC or dynamic motion of the plungers was substantially more than anticipated. It was also amazing to watch a startup of this 200
stroke per minute machine, and observe the plunger static plus dynamic motion
529
on a DC coupled oscilloscope. During the first few strokes of the drive rods, the
plunger would rise in the cylinder. This was followed by an increase in the vertical motion, and a general decrease in the horizontal excursions.
Following this simple beginning, vertical and horizontal probes were
installed on all of the tungsten carbide plungers on this machinery train. The
Proximitors were mounted in suitable enclosures, and the wiring was properly
encased in conduit. Routine surveillance then began, and various problems were
correlated to changes in plunger position and dynamic motion. Additional measurements were made on the high pressure compressors, including deflection
and position changes of the drive rods between the main and auxiliary crossheads. Casing vibration measurements were also acquired along the length of
the cylinders, plus various locations on the crank case. Of all the measurements,
the proximity probe signals of the relative plunger motion proved to be the most
meaningful, and representative of mechanical condition.
During this investigation, it was evident at an early stage that timing
between events must be quantified. This was necessary for understanding characteristics of the machine, plus sorting-out of the relative motion of each respective plunger. Since the majority of the signals occurred at 200 CPM, it was
reasonable to establish a 200 CPM trigger pulse that could be used to relate the
reciprocating to the stationary system. This would be in the same manner that a
Keyphasor probe would be used on a centrifugal machine. In this case, a hose
clamp was attached to one of the drive rods, and a proximity probe mounted on
the yoke housing (Fig. 10-36). The head of the hose clamp was positioned beneath
the probe, and the axial position of the hose clamp was coincident with top dead
center (TDC) of the plunger in #1 cylinder. Hence, when #1 plunger was at TDC,
the Keyphasor probe would fire. This proved to be an enormously useful measurement, and highly beneficial towards understanding of the machinery.
The analytical instrumentation was also expanded to include a multi channel brush recorder, tunable filters, and a vector filter. With this additional instrumentation it became possible to dissect the probe signals with even more detail.
For instance, the plunger Lissajous patterns shown in Fig. 10-37 were obtained
Cylinder 1
Cylinder 2
Cylinder 3
Filtered at Reciprocating Frequency of 200 CPM
Cylinder 4
1.0 Mil per Division
Fig. 1037 Direct Plunger Lissajous Patterns For Cylinders 1 Through 4
530
Chapter-10
from the first two throws, and the associated cylinders 1 through 4. This data
was filtered at 200 CPM, and the previously referenced reciprocating Keyphasor mark was superimposed to indicate that cylinder 1 was at TDC. At the
same time cylinder 2 was at the opposite end of the stroke, and it was at bottom
dead center (BDC). Note that the Keyphasor mark at the top of the pattern for
cylinder 1, and at the bottom of the Lissajous for cylinder 3.
It is convenient to refer to plunger axial position in terms of degrees where
one stroke equals 360. Thus, TDC is located at 0 or 360, and BDC occurs halfway through the stroke, or 180. Using this approach it is possible to chart peak
plunger motion in terms of a digital vector filter phase angle. Table 10-5 provides
a typical summary of reciprocating speed vectors for the entire array of ten cylinders with vertical and horizontal measurements on each plunger. A minor comTable 105 Summary of Reciprocating Speed Vectors Plus Throw Correction
Cylinder Plunger and
Probe Location
Recip. 1X
Amplitude
Direct
Phase
Phase
Correction
Corrected
Phase
Cylinder #1 Vertical
3.60 Mils,p-p
350
350
Cylinder #1 Horizontal
1.15 Mils,p-p
180
180
2.43 Mils,p-p
185
180
0.72 Mils,p-p
62
180
242
2.25 Mils,p-p
264
+72
336
0.20 Mils,p-p
45
+72
117
1.55 Mils,p-p
101
-108
353
0.85 Mils,p-p
278
-108
170
2.07 Mils,p-p
-36
332
0.37 Mils,p-p
233
-36
197
3.10 Mils,p-p
191
+144
335
2.03 Mils,p-p
194
+144
338
2.98 Mils,p-p
303
+36
339
1.48 Mils,p-p
151
+36
187
1.60 Mils,p-p
140
-144
356
1.23 Mils,p-p
289
-144
145
3.83 Mils,p-p
239
+108
347
1.03 Mils,p-p
45
+108
153
1.15 Mils,p-p
36
-72
324
1.59 Mils,p-p
244
-72
172
531
plication occurs when the other nine cylinders on the machine are referenced to
the same Keyphasor. It then becomes necessary to correct the direct phase
angle by the appropriate crank location and plunger orientation. For the five
throw machine shown in Fig. 10-36, the angular position between throws must
be 72 (=360/5). For a four throw machine, this incremental step would be 90.
For the example under consideration, the measured phase angles for cylinder 2 should be corrected by 180. For throw 2, cylinder 3, a +72 correction
would be necessary, and a -108 correction would be required for cylinder 4. The
phase corrections for the remaining six plungers are shown in Table 10-5. A measured phase angle may be easily corrected by the addition or subtraction of the
appropriate angle. This will yield a corrected angle that relates plunger position
with respect to TDC of that particular plunger. For example, in Table 10-5, the
corrected angles are summarized in the right hand column. Note that all of the
vertical probes reveal a 1X phase angle in the direct vicinity of 0. The actual
spread varies from 324 to 5, but it is clear that the timing mark is essentially
coincident with the peak of the vertical motion.
Another way to represent this behavior is to apply the phase corrections to
the Lissajous patterns previously displayed in Fig. 10-37. The results of this
transform are shown in Fig. 10-38. Note that the timing mark appears at the top
Cylinder 1
Cylinder 2
Cylinder 3
Filtered at Reciprocating Frequency of 200 CPM
Cylinder 4
1.0 Mil per Division
Fig. 1038 Plunger Lissajous Patterns For Cylinders 1 Through 4 With Corrected Phase
of each pattern. This is a graphical display of one of the fundamental interrelationships between mechanical components in this machine.
In the overview, it has been documented that the plunger will normally
exhibit a 2 to 6 Mil elevation in plunger elevation between zero speed and the
normal running position. The plunger will then reciprocate along this new centerline position until acted upon by external forces. As shown in Fig. 10-38, the
plunger Lissajous pattern will typically be elliptical, with the vertical motion
exceeding the horizontal. In most cases, the vertical displacement at reciprocating frequency varies between 2.0 and 5.0 Mils,p-p. In the horizontal plane, the
plunger will typically move between 1.0 and 3.0 Mils,p-p.
532
Chapter-10
The plunger motion in every observed normal case has been downhill. That
is, as the plunger enters the cylinder it drops in elevation. In fact, the plunger
will reciprocate at a slight angle with respect to a true horizontal plane. This
same motion occurs at both plungers connected to an auxiliary crosshead. It is
clear from the data that as one plunger is dropping, the companion plunger on
the opposite side of the crosshead is rising. This observation yields the conclusion
that a rocking effect is present on the auxiliary crosshead. It is generally
agreed that this behavior is a function of the following physical occurrences:
1. The surf boarding effect of the auxiliary crosshead on the slipper oil film
combined with a semi-compliant plunger connection to the crosshead.
2. The differential thermal expansion between the suction and discharge high
pressure tubing will tend to elevate the yoke end of the cylinder.
3. Vertical restraint upon the auxiliary crosshead is less than the horizontal.
4. Based on the direction of crankshaft rotation, the main crossheads for
throws 1 and 2 may be lifted by the connecting rods. This oscillation at the
main crosshead may impart a partial rocking of the auxiliary crosshead.
The rocking motion of the auxiliary crosshead, and the associated plunger
motion occurs at the fundamental reciprocating frequency of 200 CPM. This
behavior is very consistent on a normal, well-aligned machine. When malfunctions occur, the normal motion will be interrupted. In the final analysis, the precise cause of the 3 plunger failures could not be precisely established. However,
the evidence points to changes in plunger alignment during operation. This can
be effected by mechanisms such as a loose plunger connection to the crosshead,
wear of the crosshead slipper, improperly installed packing, or worn packing.
When any of these malfunctions are active, the normal dynamic motion of the
plunger is directly influenced. The plunger Lissajous figure will often reveal the
preload, and a shift in plunger centerline position is generally observed. If no
mechanical damage is caused by the misalignment, then the plunger may be successfully realigned, and normal motion restored. Obviously, when mechanical
damage has occurred, realignment will only provide temporary relief. The damage initiated by the misalignment will usually continue to deteriorate, and corrective measures will be required.
For reference purposes, several of the primary mechanical malfunctions
detected on this class of machine are summarized as follows.
The detection of plunger packing failures (radial or tangential rings) is
best described in terms of the restraining effect of the packing. Since the high
pressure packing does function with a close clearance to the plunger, the packing
must act as one of the lateral plunger restraints (spring). Any increase in packing clearance will decrease the effective packing spring constant, and allow
increased plunger motion in the direction of increased clearance. It is usually difficult to separate alignment and packing malfunctions since they often display
the same symptoms. It is certainly reasonable for this to occur since alignment
and packing difficulties are often coexistent. Therefore, the inception of misalignment may cause packing deterioration, and vice versa. One factor that does occur
533
only with terminal packing failures is a significant increase in leak gas rate.
Some packing failures result in the impregnation of bronze into the
tungsten carbide plungers. These small particles of bronze are highly visible on
an oscilloscope time base as sharp spikes originating from the basic plunger
sinusoidal waveform. In most observed cases, when the plunger physically contacts the bronze rings, rubbing occurs, and the impregnated bronze will increase
as a function of time.
Babbitt loss on the auxiliary crosshead slipper will effect both plungers on a single throw. This mechanism will reveal itself as a significant drop in
the centerline position on both attached plungers, combined with a substantial
change in plunger dynamic motion. Although the reciprocating amplitudes generally increase when this problem occurs, it is also possible to experience a brief
period of reduced motion before amplitudes begin to increase. It should also be
noted that this slipper babbitt loss may be due to attrition, or it may be caused
by improper lubrication of the slipper.
Loose plunger connection to the auxiliary crosshead is difficult to detect
during normal operation, due to the large compressive forces acting between the
plunger and the crosshead connection. This problem is usually detected during
startup when lube oil is applied to the auxiliary crosshead, and coolant oil is
directed to the packing gland. During this condition, a loose connection will show
up as a substantial and erratic variation in probe DC gap voltage. This change
will appear on one, and in some cases both proximity probes observing the loose
plunger. To identify this problem, the probe DC gap voltages should be documented prior to the initiation of oil, and they should be rechecked throughout the
startup sequence to be sure that large gap voltage changes have not occurred.
Occasionally, this problem may also be visible after startup and before loading of
the machine. Running essentially unloaded a loose plunger connection will
appear as very high peak to peak displacement amplitudes.
The work described in this case history was performed during a 4 month
period. The correlation between machinery integrity and the vibration and position data obtained from the plunger proximity probes was self-evident. The next
step was to install permanent X-Y vibration monitors for each plunger. These
monitors were equipped with a 15.0 Mil,p-p full scale. Alert levels were set at 8.0
Mils,p-p, and an automatic trip of the machine was set at 10.0 Mils,p-p. This trip
was suppressed during startup, but was fully active after the machine was at
speed and pressure. For accurate documentation of the probe DC gap voltages, a
DC voltmeter and selector switch was also incorporated with the original monitor racks. Most of the initial data was obtained with clip boards and photographs
of the oscilloscope traces. In a current rendition, this type of system would be logically handled in a digital monitoring system combined with a computer-based
data acquisition and storage system.
The success or benefit of any machinery analysis project is seldom measured on a short-term basis. The real measure of success is the ability to safely
operate year after year with no failures, and no surprises. For this particular
project, the initial 3 plunger failures appeared in the vicinity of 3,000 hours of
534
Chapter-10
run time (about 4 months). In the early days of this plant, the time interval
between major overhauls of these machines was planned for approximately
every 5,000 hours (about 6 months). However, since the conclusion of this study,
and the installation of the machinery protection instrumentation, there have
been no additional plunger failures. Furthermore, the time interval between
major overhauls of these machines has been extended to over 20,000 hours (over
2 years), with some reported instances of machines operating more than 30,000
hours (greater than 3 years) between overhauls.
BIBLIOGRAPHY
1. Alban, Lester E., Systematic Analysis of Gear Failures, Metals Park, Ohio: American
Society for Metals, 1985.
2. Baumgardner, James, Tutorial Session on Motors, Proceedings of the Eighteenth
Turbomachinery Symposium, Turbomachinery Laboratory, Texas A&M University
System, College Station, Texas (October 1989).
3. Blevins, Robert D., Flow-Induced Vibration, Second Edition, New York: Van Nostrand Reinhold, 1990.
4. Corley, James E., Tutorial Session on Diagnostics of Pump Vibration Problems,
Proceedings of the Fourth International Pump Symposium, Turbomachinery Laboratory, Texas A&M University, College Station, Texas (May 1987).
5. Dudley, Darle W., Gear Handbook, New York: McGraw-Hill Book Company, 1962.
6. Harris, Cyril M., Shock and Vibration Handbook, Fourth edition, pp. 29.1 to 29.19,
New York: McGraw-Hill, 1996.
7. Nelson, W.E. (Ed), and J.W. Dufour, Pump Vibrations, Proceedings of the Ninth
International Pump Users Symposium, Turbomachinery Laboratory, Texas A&M
University, College Station, Texas (March 1992), pp. 137-147.
8. Schiavello, Bruno, Cavitation and Recirculation Troubleshooting Methodology,
Proceedings of the Tenth International Pump Users Symposium, Turbomachinery
Laboratory, Texas A&M University, College Station, Texas (March 1993), pp. 133156.
9. Schwartz, Randal E., and Richard M. Nelson, Acoustic Resonance Phenomena In
High Energy Variable Speed Centrifugal Pumps, Proceedings of the First International Pump Symposium, Turbomachinery Laboratories, Texas A&M University,
College Station, Texas (May 1984), pp. 23-28.
10. Spotts, M.F., Design of Machine Elements, 6th Edition, Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1985.
11. Winterton, John G., Component identification of gear-generated spectra, Orbit,
Vol. 12, No. 2 (June 1991), pp. 11-14.
Rotor Balancing
11
11
ass unbalance in a rotating system

often produces excessive synchronous forces that reduces the life span of various
mechanical elements. To minimize the detrimental effects of unbalance, turbo
machinery rotors are balanced with a variety of methods. Most rotors are successfully balanced in slow speed shop balancing machines. This approach provides good accessibility to all correction planes, and the option of multiple runs to
achieve a satisfactory balance. It is generally understood that balancing at slow
speeds with the rotor supported by simple bearings or rollers does not duplicate
the rotational dynamics of the field installation. Other rotors are shop balanced
on high speed balancing machines installed in vacuum pits or evacuated chambers. These units provide an improved simulation of the installed rotor behavior
due to the higher speeds, and the use of bearings that more closely resemble the
normal machinery running bearings. In these high speed bunkers, the influence
of rotor blades and wheels are substantially reduced by operating within a vacuum. In general, this is a desirable running condition for correcting rotor unbalance, and for studying the synchronous behavior of an unruly rotor.
Some machines, such as large steam turbines, often require a field trim balance due to the influence of higher order modes, or the limited sensitivity of the
low speed balance techniques. There is also a small group of machines that contain segmented rotors that are assembled concurrently with the stationary diaphragms or casing. In these types of machines, the final rotor assembly is not
achieved until most of the stationary machine elements are bolted into place.
Machines of this configuration almost always require some type of field trim balance correction.
In the overview, virtually all rotating machinery rotors are balanced in one
way or another. As stated in chapter 9, this is a fundamental property of rotating
machinery, and it must be considered in any type of mechanical analysis. Furthermore, it is almost mandatory for the machinery diagnostician to fully understand the behavior of mass unbalance, and the implications of unbalance
distribution upon the rotor mode shape, and the overall machinery behavior. If
the diagnostician never balances a rotor during his or her professional career,
they still must understand the unbalance mechanism to be technically knowledgeable and effective in this business.
535
536
Chapter-11
Rotor balancing is often considered to be a straightforward procedure that

is performed in accordance with the instructions provided by the balancing
machine manufacturer. Although this is true in many instances of shop balancing, field balancing is considerably more complicated. It must always be recognized that the rotor responds in accordance to the mechanical characteristics of
mass, stiffness, and damping. Thus, a rotor subjected to a low speed shop balance
does not necessarily guarantee that field operating characteristics will be acceptable. In most instances, simple rotors may be acceptably shop balanced at low
speeds. In some cases, complex rotor systems, or units with sophisticated bearing
or seal arrangements, may require a field trim balance at full operating speeds,
with the rotors installed in the actual machine casing.
In either situation, the synchronous 1X response of the rotor must be
understood, and the influence of balance weights must be quantified. Shop balancing machines typically perform the full array of vector calculations with their
internal software. However, field balancing requires the integration of various
transducers with vector calculations performed at one or more operating speeds.
In order to provide an improved understanding of rotor synchronous motion, the
influence of higher order modes, and the typical field balance calculation procedure, chapter 11 is presented.
BEFORE BALANCING
There are several considerations that should be addressed prior to the field
balancing of any rotor. The fundamental issue concerns whether or not the vibration is caused by mass unbalance or another malfunction. A variety of other
mechanisms can produce synchronous rotational speed vibration. For example,
the following list identifies problems that initially can look like rotor unbalance:
Excessive Bearing Clearance

Bent Shaft or Rotor
Load or Electrical Influence
Gear Pitch Line Runout
Misalignment or Other Preload
Cracked Shaft
Soft Foot
Locked Coupling
Gyroscopic Effects
Compliant Support or Foundation
Thus, the first step in any balancing project is to properly diagnose the root
cause of mechanical behavior. The machinery diagnostician must be reasonably
confident that the problem is mass unbalance before proceeding. If this step is
ignored, then the balancing work may temporarily compensate for some other
malfunction; with direct implications for excessive long-term forces acting upon
the rotor assembly.
Before Balancing
537
Balancing speed, load, and temperature are very important considerations. The balancing speed should be representative of rotational unbalance at
operating conditions yet free of excessive phase or amplitude excursions that
could confuse either the measurements, or the balance calculations. This means
that the rotational speed vectors should remain constant within the speed
domain used for balancing. It is highly recommended that balancing speeds be
selected that are significantly removed from any active system resonance. This is
easily identified by examination of the Bode plots, and the selection of a speed
that resides within a plateau region where 1X amplitude and phase remain constant. It should be recognized that in some cases, the field balancing speed may
not be equal to the normal operating speed. Again, this can only be determined
by a knowledgeable examination of the synchronous transient speed data.
In most instances, the transmitted load and operating temperature are
concurrent considerations. Balancing a cold rotor under no load may produce
quite different results from balancing a fully heat soaked rotor at full process
load. In many cases, the machine will be reasonably insensitive to the effects of
load and temperature. On other units, such as large turbine generator sets, these
effects may be appreciable. In order to understand the specific characteristics of
any machinery train, the synchronous 1X vibration vectors should be tracked
from full speed no load, to full speed full load operation at a full heat soak. This
should be a continuous record that includes process temperatures and load information. If discrete 1X vectors are acquired at the beginning and end of the loading cycle, the diagnostician has no visibility of how the machine changed from
one condition to the other. Hence, the acquisition of a detailed time record (probably computer-based) is of paramount importance. In some cases, the field balance corrections will be specifically directed at reducing the residual unbalance
in the rotor(s). In other situations, the installation of field balance weights may
compensate for a residual bow, or the effects of some load or heat related mechanism. These should be knowledgeable decisions obtained by detailed examination of the synchronous response of the machine during loading.
Mechanical configuration and construction of the rotor must be reviewed to
determine the mode shape at operating speed, plus the location and accessibility of potential balance planes. The mode shape must be understood to select
realistic balance correction planes, and to provide guidance in the location of correction weights. As discussed in chapter 3, the mode shape may be determined
by field measurements, by analytical calculations, or a suitable combination of
the two techniques. If a modally insensitive balance plane is selected, the addition of field balance weights will be totally ineffective. In some situations, weight
changes at couplings, or holes drilled on the outer diameter of thrust collars may
never be sufficient for field balancing a machine. In these cases, the field balancing efforts are futile, and the machine should be disassembled for shop balancing
of the rotor (low or high speed) at modally sensitive lateral locations.
Field weight corrections are achieved by various methods depending
upon the machine, and the available balance planes. For example, it is common
to add or remove balancing screws, add or remove sliding weights, add washers
to the coupling, weld weights on the rotor, or drill/grind on the rotor element. The
538
Chapter-11
use of balance screws in an OEM balance plane is usually a safe correction. It is

good practice to use an anti-seize compound on the screw threads (mandatory for
stainless steel weights screwed into a stainless steel balance disk). In some
cases, steel weights may not be heavy enough for the required balance correction.
In these situations, consider the use of tungsten alloy balance weights that have
nearly twice the density of steel weights.
The sliding weights employed on the face of many turbine wheels fit into a
circumferential slot, and are secured in place with a setscrew. These weights
have a trapezoidal cross section to fit semi-loosely into the trapezoidal slot. The
setscrew passes through the balance weight, and into the axial face of the turbine wheel. Tightening the setscrew locks the weight between the angled walls of
the slot and the wheel. Normally, each balance slot of this type has only 2 locations for insertion of the weights. Depending on the weights already installed in
the slot, it may be easier to install the weights from one side versus the other.
The addition of coupling washers carries disadvantages, such as the loss of
the washers during future disassembly, or the mis-positioning of the washers
during future re-assemblies. For the most part, addition of coupling washers represents a temporary balance weight correction measure. This may be the most
appropriate way to get a machine up and running in the middle of the night; but
more permanent corrections should be made to the coupling or rotor assembly
during the next overhaul.
The installation of U-Shaped weights is a common practice on units such as
induced or forced draft fans. These weights are temporarily attached to the outer
diameter of the fan center divider plate, or a shroud band, using an axial setscrew. After verification that the weights are correct, the balance weights are
typically welded to the rotor section. In all cases, welding balance weights on the
rotor should be performed carefully. On sensitive machines, the weight of the
welding rod (minus flux) should be included in the total weight for the balance
correction. In addition, the ground connection from the welding machine must be
attached to the rotor close to the location of the balance weight. Under no circumstance should the ground wire be connected to the machine casing, bearing
housing, or pedestal. This will only direct the welding machine current flow
through the bearings, with a strong probability of immediate bearing damage.
Furthermore, the machinery diagnostician must always be aware of the
metallurgy of the fan rotor, and the balance weights. On simple carbon steel
assemblies, virtually any qualified welder will be able to do a good job with commonly available welding equipment. On more exotic metal combinations, the
selection of the proper rod, technique, and welding machine must be coupled
with a fully qualified welder for that physical configuration.
For drilling or grinding on a machinery rotor, low stress areas must be
selected. Mechanical integrity of the rotor should never be compromised for a
balance correction. The rotor material density should be known, so that the
amount of weight removed can be computed by knowing the volume of material
removed. Finally, the location of angles for weight corrections should be the
responsibility of the individual performing the balancing work. It is easy to misinterpret an angular orientation and drill the right hole in the wrong place. Mis-
Standardized Measurements and Conventions
539
takes of this type are expensive, and they are totally unnecessary.
Finally, there are individuals who firmly believe that balancing will provide
a cure for all of their mechanical problems. The attitude of lets go ahead and
throw in a balance shot is prevalent in some process industries. Obviously this
philosophy will be correct when the problem really is mass unbalance. However,
this can be a dangerous approach to apply towards all conditions. Basically, if the
problem is unbalance, then go balance the rotor. If the problem is something else,
then go figure out the real malfunction.
STANDARDIZED MEASUREMENTS AND CONVENTIONS

Before embarking on any discussion of balancing concepts, it is highly
desirable to establish and maintain a common set of measurements and conventions. These standardized rules will be applied throughout this balancing chapter, and they are consistent with the remainder of this text. As expected, vectors
are used for the 1X response measurements and calculations. Vectors are
described by both a magnitude and a direction. For instance, a car driven 5 miles
(magnitude) due West (direction) defines the exact position of the vehicle with
respect to the starting point. For vibration measurements, a running speed vector quantity stated as 5.0 Mils,p-p (magnitude) occurring at an angle of 270
(direction) defines the amount, and angular location of the high spot. In any balancing discipline, both quantities are necessary to properly define a vector.
As discussed in chapter 2, circular functions, exponential functions, and
inphase-quadrature terms may be used interchangeably. Although conversion
from one format to another can be performed, this does unnecessarily complicate
the calculations. Within this chapter, vectors will be expressed as a magnitude,
with an angle presented in degrees (1/360 unit circle). Angular measurements in
radians, grads, or other units will not be used. Vector amplitudes will vary with
the specific quantity to be described. For shaft vibration measurements, magnitudes will be presented as Mils (1 Mil equals 0.001 Inches), peak to peak. This
will generally be abbreviated as Mils,p-p. Casing measurements for velocity will
carry magnitude units of Inches per Second, zero to peak, and will be abbreviated as IPS,o-p. Finally, casing vibration measurements made with accelerometers will be shown as Gs of acceleration (1 G equals the acceleration of gravity),
and zero to peak values will be used. Acceleration magnitudes will generally be
abbreviated as Gs,o-p.
All balance weights, trial weights, and calibration weights must be
expressed in consistent units within each balance problem. Typically, small
rotors will be balanced with weights measured in grams. Large rotors will generally require larger balance weights, and units of ounces (where 1 ounce equals
28.35 grams), or pounds (where 1 pound equals 16 ounces or 453.6 grams) will be
used. It is always desirable to both calculate the weight of a correction mass
(density times volume) plus place the correction mass on a calibrated scale
and weight it directly. This is certainly a belt and suspenders approach, but it
540
Chapter-11
does prevent embarrassing mistakes, and the potential installation of the wrong
weight at the right location.
In the following pages, balance sensitivity vectors will be calculated. The
magnitude units for these vector calculations will consist of weight (or mass)
divided by vibration. For instance, units of Grams/Mils,p-p would be used for sensitivity vectors associated with most rotors. Occasionally, these vectors may be
inverted to yield units of Mil,p-p/Gram. This format sometimes provides an
improved physical significance or meaning. However, the diagnostician should
always remember that these are vector quantities. If you invert the magnitude,
then the angle must also be corrected to be mathematically correct. In addition,
the balance equations are totally interlocked to the balance sensitivity vectors. If
someone begins to casually invert sensitivity vectors, the equation structure will
become completely violated.
In some situations, a radial length may be included to define the radius of
the balance weight from the shaft centerline. This allows more flexibility in
selection of the final weight and associated radius. For instance, if a rotor balance sensitivity is 50 Gram-Inches/Mil,p-p, and the measured vibration amplitude is 2.0 Mils,p-p; then the product of these two quantities would be a balance
correction of 100 Gram-Inches. This may be satisfied by a correction weight of 20
grams at a radius of 5 inches, or by a weight of 4 grams at a 25 inch radius.
Generally, the magnitude portion of the vector quantities is easily understood and applied. The major difficulty usually resides with the phase measurements and the associated angular reference frame. Part of the confusion is
directly related to the function and application of the timing mark, or trigger
point. In all cases, the timing signal electronics provide nothing more than an
accurate and consistent manner to relate the rotating element back to the stationary machine. Within this text, the majority of the synchronous timing signals
will be based upon a proximity probe observing a notch in the shaft, or a projection such as a shaft key. In either case, the resultant signal emitted by the proximity timing probe will be a function of the average gap between the probe and
the observed shaft surface.
As discussed in chapter 6, a Keyphasor probe will produce a negative
going pulse when the transducer is positioned over a notch or keyway as shown
in Figs. 11-1 and 11-2. In a similar manner, the Keyphasor proximity probe will
generate a positive going pulse when it observes a projection or key as shown in
Figs. 11-3 and 11-4. The actual trigger point is a function of the instrument that
receives the pulse signal. This device may be a synchronous tracking Digital Vector Filter (DVF), a Dynamic Signal Analyzer (DSA), or an oscilloscope. All of
these traditional instruments require the identification of a positive or a negative slope for the trigger, plus the trigger level within that slope. In many cases,
the devices are set for an Auto Trigger position, which automatically sets the
trigger at the halfway point of the selected positive or negative slope.
The physical significance of the trigger point is illustrated by the diagrams
in Figs. 11-1 through 11-4. In all eight cases, the actual trigger point is established by the coincidence of the physical shaft step and the proximity probe. For
541
Keyphasor
Trigger
Point
Keyphasor
Trigger
Point
Trigger At
Leading
Edge of Slot
CC
Trigger At
Trailing
Edge of Slot
CW
W R o t a ti o n
R ot a ti o n
Fig. 111 Negative Trigger Slope With Slot
CC
CW
W R o t a ti o n
R ot a ti o n
Fig. 112 Positive Trigger Slope With Slot
an instrument set to trigger off a negative slope, the Keyphasor probe is essentially centered over the leading edge of the notch as shown in Fig. 11-1. For a
trigger off a positive slope, the Keyphasor probe is centered over the trailing
edge of the notch as shown in Fig. 11-2.
This positioning between the stationary and the rotating systems is not
that critical for machines with large shaft diameters. However, on rotors with
small shaft diameters, the establishment of an accurate trigger point is mandatory. For example, on a 2 inch diameter shaft, if the trigger point is off by 1/4
inch, this is equivalent to a 14 error. If this error is encountered during the
placement of a balance correction weight, the results would probably be less than
desirable. Hence, the establishment of an accurate trigger point is a necessary
requirement for successful balancing.
The diagrams presented in Figs. 11-3 and 11-4 describe the trigger condition for a positive going pulse emitted by a timing probe observing a projection or
Keyphasor
Trigger
Point
Keyphasor
Trigger
Point
Trigger At
Trailing
Edge of Key
Trigger At
Leading
Edge of Key
CC
W R o t a ti o n
CW
R ot a ti o n
Fig. 113 Positive Trigger Slope With Key
CC
W R o t a ti o n
CW
R ot a ti o n
Fig. 114 Negative Trigger Slope With Key
542
Chapter-11
other raised surface such as shaft key. Again, the trigger point is established by
the coincidence of the shaft step and the proximity probe. For an instrument set
to trigger off a positive slope, the Keyphasor probe is essentially centered over
the leading edge of the key as in Fig. 11-3. For a trigger off a negative slope, the
Key probe is centered over the trailing edge of the key as shown in Fig. 11-4.
As noted, each trigger point example is illustrated with a clockwise and a counterclockwise example. Typically, the machine rotation is observed from the driver
end of the train, and the appropriate Keyphasor configuration (i.e., notch or
projection), is combined with an instrument setup requirement for a positive or a
negative trigger slope. This combination of parameters allows the selection of
one of the eight previous diagrams as the unique and only trigger point for the
machine to be balanced.
In passing, it should also be mentioned that the use of an optical Keyphasor observing a piece of reflective tape on the shaft will produce a positive going
with most optical drivers. Hence, the optical trigger signals will be identical to
the drawings shown in Fig. 11-3 and 11-4. Also be advised that reflective tape
will not adhere to high speed rotors. Depending on the shaft diameter, a limit of
15,000 to 20,000 RPM is typical for acceptable adhesion of most reflective tapes.
For balancing of units at higher speeds that require an optical Key, the use of
reflective paint on the shaft is recommended. Additional contrast enhancement
may be obtained by spray painting the shaft with dull black paint or layout bluing. This dark background combined with the reflective paint or tape will yield a
strong pulse signal under virtually all conditions.
Regardless of the source of the Keyphasor, the diagnostician must always
check the clarity of the signal pulse on an oscilloscope. A simple time domain
observation of this pulse will identify if the voltage levels are sufficient to drive
the analytical instruments (typically 3 to 5 volts, peak). Next, the time domain
signal will reveal if there are any noise spikes or other electronic glitches in the
signal. Most of these interferences are due to some problem with the transducer
installation, and will have to be corrected back at the timing probe.
There are conditions where baseline noise on the presumably flat part of
the trigger curve may be corrected with external voltage amplifiers. In this common manipulation, the direct pulse signal is passed through a DC coupled voltage amplifier, and the bias voltage adjusted (plus or minus) to flatten out the
baseline. Next, the signal is passed through an AC coupled voltage amplifier and
the signal gain is increased to provide a suitable trigger voltage. Naturally, the
outputs of both amplifiers should be observed on an oscilloscope to verify the
proper results from both amplifiers (e.g., Fig. 8-8). This same procedure may be
used to clean up a signal from a tape recorder. The final objective must be a clean
and consistent trigger relationship between the machinery and the electronics.
Once a unique trigger point has been established, the rotor is physically
rolled under the Key probe to satisfy this trigger condition. At this point, an
angular coordinate system is established from one of the vibration probes. For
example, consider the diagram presented in Fig. 11-5 for vibration probes
mounted in a true vertical orientation. In all cases, the angular coordinate system is initiated with 0 at the vibration probe, and the angles always increase
0
315
90
225
135
45
W R o t a ti o n
315
90
270
135
180
CC
45
0
45
270
543
225
180
CW
R o t a ti o n
Fig. 115 Angular Convention With True

Vertical Vibration Probes
0
315
135
270
CC
315
90
180
0
45
225
90
180
135
225
W R o t a ti o n
270
CW
R o t a ti o n
Fig. 116 Angular Convention With Vertical Probes At 45 Left of Vertical Centerline
against rotation. Fig. 11-5 describes the angular reference system for both a
counterclockwise, and a clockwise rotating shaft. Another way to think of this
angular coordinate system is to consider the progression of angles as the shaft
rotates in a normal direction. Specifically, if one observes the rotor from the perspective of the probe tip, and the shaft is turning in a normal direction, the
angles must always increase. This type of logic is mandatory for a proper correlation between the machine, and the resultant polar plots of transient motion, and
orbit plots of steady state motion.
If the vibration probes are located at some other physical orientation, the
logic remains exactly the same. The 0 position remains fixed at the vibration
transducer, and the angles are laid off in a direction that is counter to the shaft
rotation. For probes mounted at 45 to the left of a true vertical centerline, the
angular coordinate system for a counterclockwise and a clockwise rotating shaft
are presented Fig. 11-6.
All vibration vector angles from slow roll to full speed are referenced in this
manner. All trial weights, calibration weights, and balance correction weights
are referenced in this same manner. The mass unbalance locations are also referenced with this same angular coordinate system. Differential vectors and balance sensitivity vectors also share the exact same angular coordinate system.
Although this may seem like a trivial point, it is an enormous advantage to
maintain the same angular reference system for all of the vector quantities
involved in the field balancing exercise.
A minor variation exists when X-Y probes are installed on a machine. If all
of the angles are reference to the Vertical or Y-axis probes, there will be a phase
difference between measurements obtained from the X and the Y transducers. If
the machine exhibits forward circular orbits, a 90 phase difference will be exhibited at each bearing. This normal phase difference causes only a minor problem.
It is recommended that one set of transducers, for example the Y probes, be used
as the 0 reference as previously discussed. The vibration vectors measured by
the X probes would be directly acquired, and measured angles used in the balancing equations. Since the calibration weights are referenced to the Y probes,
the results from the X probes will be self-corrected. This concept will be demon-
544
Chapter-11
strated in the case history 37 presented immediately after the development of

the single plane balancing equations. Additional explanation of this characteristic will also be provided in case history 38 on a five bearing turbine generator set.
The final concept that must be understood is the relationship between the
Heavy Spot and the High Spot. As discussed in previous chapters, the rotor
Heavy Spot is the point of effective residual mass unbalance for the rotor (or
wheel). As shown in Figs. 11-7 and 11-8, the Heavy Spot maintains a fixed angular position with respect to the vibration probe. Regardless of the machine speed
(i.e., below or above a critical speed), a constant relationship exists between the
vibration probe and the Heavy Spot. This is the angular location where weight is
removed to balance the rotor (110 in this example). Weight may also be added at
290 in this example to correct for the Heavy Spot unbalance at 110. This is the
logic behind the low speed proximity probe balancing rule of:
At speeds well below the critical,
remove weight at the phase angle,
or add weight at the phase angle plus 180
The Heavy Spot and the High Spot are coincident at slow rotational speeds
that are well below any shaft critical (balance resonance) speeds. Whereas the
Heavy Spot is indicative of the lumped residual unbalance of the rotor, the High
Spot is the high point measured by the shaft sensing proximity probe. This is the
angular location that maintains the minimum distance between the rotor and
the shaft during each revolution.
As rotor speed is increased, the unit passes through a balance resonance
(critical speed). For a lightly damped system, the observed High Spot passes
through a nominal 180 phase shift. This change is due to the rotor performing a
self-balancing action where the center of rotation migrates from the original geometric center of the rotor to the mass center, as shown In Fig. 11-8. The proxim-
Probe
@ 0
High
Spot
Phase
= 110
Add
Weight
Location
Geometric
Center
Add
Weight
Location
Center of
Rotation
270
90
Probe
@ 0
Phase
= 290
Geometric
Center
Center of
Rotation
90
270
Mass
Center
Mass
Center
Heavy
Spot
High
Spot
Remove
Weight
Location
180
CCW Rotation
Fig. 117 Heavy Spot Versus High Spot At

Speeds Well Below The Critical Speed
Heavy
Spot
Remove
Weight
Location
180
CCW
Rotation
Fig. 118 Heavy Spot Versus High Spot At

Speeds Well Above The Critical Speed
Combined Balancing Techniques
545
ity probe can only measure distances, and it observes the shift through the
critical speed as a change of 180 in the High Spot. The probe has no idea of the
location of the mass unbalance, it only responds to the change in distance. This
behavior is the logic behind the traditional proximity probe balancing rule of:
At speeds well above the critical,
add weight at the phase angle,
or remove weight at the phase angle minus 180
Clearly, a comparison of Fig. 11-7 with 11-8 reveals that the Heavy Spot has
remained in the same location at 110. Removing an equivalent weight at 110 or
adding an equivalent weight to 290 will result in a balanced rotor. For additional explanation of this classic behavior through a critical speed, the reader is
referenced back to the description of the Jeffcott rotor presented in chapter 3 of
this text. In addition, the presence of damping and mechanisms such as combined static and couple unbalance will alter the above general rule.
COMBINED BALANCING TECHNIQUES

The measurements and standardized conventions provide the basis for discussing unbalance corrections. It would be convenient if these concepts could be
formed into a set of balancing procedures that are universally applicable to all
machines and situations. Unfortunately, such a panacea does not exist, and additional consideration must be given to the actual field balancing techniques. Over
the years, various successful techniques have been developed within the machinery community. In 1934, the original influence coefficient vector approach was
described by E.L. Thearle1 of General Electric. Ronald Eshleman2 of the Illinois
Institute of Technology, completed his initial work on methods for balancing flexible rotors in 1962. Balancing with shaft orbits was published in 1971 by Charles
Jackson3 of Monsanto. The combination of modal and influence coefficient techniques was presented in 1976 by Edgar Gunter4, et. al., University of Virginia.
Variable speed polar plot balancing was introduced by Donald Bently5, Bently
Nevada in 1980. Certainly there have been many other contributors to this field;
and today there are a variety of balancing techniques available.
1 E.L. Thearle, Dynamic Balancing of Rotating Machinery in the Field, Transactions of the
American Society of Mechanical Engineers, Vol. 56 (1934), pp. 745-753.
2 R. Eshelman, Development of Methods and Equipment for Balancing Flexible Rotors,
Armour Research Foundation, Illinois Institute of Technology, Final Report NOBS Contract 78753,
Chicago, Illinois (May 1962).
3 Charles Jackson, Balance Rotors by Orbit Analysis, Hydrocarbon Processing, Vol. 50, No. 1
(January 1971).
4 E.J. Gunter, L.E. Barrett, and P.E. Allaire, Balancing of Multimass Flexible Rotors, Proceedings of the Fifth Turbomachinery Symposium, Gas Turbine Laboratories, Texas A&M University, College Station, Texas (October 1976), pp. 133-147.
5 Donald E. Bently, Polar Plotting Applications for Rotating Machinery, Proceedings of the
Vibration Institute Machinery Vibrations IV Seminar, Cherry Hill, New Jersey (November 1980).
546
Chapter-11
The real key to success resides in selecting the techniques most applicable
to the machine element requiring balancing and performing that work in a
timely and cost-effective manner. There is an old adage that states: if your only
tool is a hammer, then all of your problems begin to resemble nails. This is particularly true in the field of onsite rotor balancing. If you only use one specific technique, your options are very limited, and you have no recourse when a machine
misbehaves. The balancing techniques used for high speed rotors should integrate the concepts of modal behavior, variable and constant speed vibration measurements, plus balance calibration of the rotor to yield discrete corrections.
Although these topics may be considered as separate entities, they are all
addressing the same fundamental mass distribution problem. The integrated
balancing approach discussed herein attempts to use the available information
to provide a logical assessment of field balance corrections.
Initially, a correct understanding of the modal behavior is important for two
reasons. First, it helps to identify balance planes with suitable effectiveness
upon the residual unbalance. Secondly, it provides direction as to whether the
weight correction should be added or removed at a particular phase angle. The
mode shape can be determined analytically, or experimentally by vibration measurements. Ideally, the analytical calculations should be substantiated by variable speed field vibration measurements to confirm the presence, and location of
system critical speeds.
The next step consists of using vibration response measurements to help
identify the lateral and angular location of rotor unbalance. In a case of pure
mass unbalance, the runout compensated vibration angles will be indicative of
the angular location of the unbalance. In the presence of other forces, amplitude
and angular variations will occur. However, the relative vibration amplitudes
will help to identify the offending lateral correction planes, and the vector angles
provide a good starting point for angular weight locations.
Unless previous balancing information is available, it is usually difficult to
anticipate the amount of unbalance. For this reason, many field balancing solutions gravitate towards the Influence Coefficient method for calculation of correction weights. Applying this technique, the mechanical system is calibrated with a
known weight placed at a known angle. Assuming a reasonably linear system (to
be discussed), the response from the calibration or trial weights are used to compute a balance correction that minimizes the measured vibration response
amplitudes at the balancing speed.
It should be recognized that the balancing calculations are precise, but they
are based upon values that contain different levels of uncertainty. Hence, it is
always best to run the calculations with the best possible input measurements,
and then make reasonable judgments of the actual corrections to be implemented on the machine. In some cases contradictions will appear in the results,
and the individual performing the balancing will have to exercise judgment in
selecting corrections that make good mechanical sense.
The following sections in this chapter will address the typical balancing calculations that can be performed. The presented vector balancing equations can
be programmed on pocket calculators or personal computers. In fact, operational
Linearity Requirements
547
programs have been available for many years. The use of portable personal computers equipped with spreadsheet programs are ideal for this type of work. It is
acknowledged that the calculator programs or computer spreadsheets are only
as good as the balancing software. It is always desirable to fully understand the
software package, and test it with previously documented balance calculations
and/or a mechanical simulation device (e.g. rotor kit), where the integrity and
operation of the software can be verified in a noncritical environment.
The final point in any field balance consists of documentation for future reference. In some cases, if a unit requires a field balance, chances are good that
periodically this machine may have to be rebalanced. If everything is fully documented, the knowledge gained about the behavior of this particular machine will
be useful during the next balance correction. The engineering files should contain all of the technical information, and notes that were generated during the
execution of the balancing. This file should be complete enough to allow reconstruction of the entire balancing exercise.
Again, it must be restated that successful field balancing really requires an
unbalanced rotor. The mechanical malfunctions listed at the beginning of this
chapter will exhibit many symptoms that may be interpreted as unbalance.
However, careful examination of the data will often allow a proper identification
of the occurring malfunction, and treatment of the actual mechanical problem.
LINEARITY REQUIREMENTS
Traditional balancing calculations generally assume a linear mechanical
system. For a system to be considered linear, three basic conditions must be satisfied. First, if a single excitation (i.e., mass unbalance) is applied to a system, a
single response (i.e., vibration) can be expected. If the first excitation is removed,
and a second excitation applied (i.e., another mass), a second response will
result. If both excitations are simultaneously applied, the resultant response will
be a superposition of both response functions. Hence, a necessary condition for a
system to be considered linear is that the principle of superposition applies.
The second requirement for a linear system is that the magnitude or scale
factor between the excitation and response is preserved. This characteristic is
sometimes referred to as the property of homogeneity, and must be satisfied for a
system to be linear. The third requirement for a linear system considers the frequency characteristics of dynamic excitations and responses. If the system excitations are periodic functions, then the response characteristics must also be
periodic. In addition, the response frequency must be identical to the excitation
frequency; and the system cannot generate new frequencies.
Most rotating machines behave in a reasonably linear fashion with respect
to unbalance. Occasionally, a unit will be encountered that violates one or more
of the three described conditions for linearity. When that occurs, the equation
array will fail by definition, and a considerably more sophisticated diagnostic
and/or analytical approach will be necessary. However, in many instances a
direct technique may be used to determine the unbalance in a rotating system.
548
Chapter-11
Case History 36: Complex Rotor Nonlinearities
Journal 4.000"
5 Pads - LOP
445# Static Load
Exhaust
700F
430F Discharge
450F Discharge
1,250F Inlet
The machinery discussed in case history 12 will be revisited for this example of nonlinear machinery behavior. Recall that this unit consisted of an overhung hot gas expander wheel, a pair of midspan compressor wheels, and three
overhung steam turbine wheels6 as originally shown in Fig. 5-10. For convenience, this same diagram is duplicated in Fig. 11-9. A series of axial through
bolts are used to connect the expander stub shaft through the compressor
wheels, and into the turbine stub shaft. In this machine, the rotor must be built
concurrently with the inner casing. Specifically, the horizontally split internal
bundle is assembled with the compressor wheels, stub shafts, plus bearings, and
seals. The end casings are attached, the expander wheel is bolted into position,
and the turbine stages are mounted with another set of through bolts.
The eight rotor segments are joined with Curvic couplings. Although each
of the segments are component balanced, any minor shift between elements will
produce a synchronous force. Since this unit operates at 18,500 RPM, a slight
unbalance or eccentricity will result in excessive shaft vibration. Furthermore,
the distribution of operating temperatures noted on Fig. 11-9 reveals the complex thermal effects that must be tolerated by this rotor. The 1,250F expander
inlet is followed by compressor temperatures in excess of 430F. The steam turbine operates with a 700F inlet, and a 160F exhaust.
By any definition, this must be considered as a difficult unit. As discussed
in case history 12, the rotor passes through seven resonances between slow roll
and normal operating speed. These various damped natural frequencies were
summarized in Table 5-4. This rotor normally requires field trim balancing after
every overhaul. Previous field balancing activities were successful when a two
step correction process was applied. The first step consisted of balancing at a pro-
Journal 4.500"
5 Pads - LOP
465# Static Load
Ambient
Air Suction
220F Suction
Turbine
Compressor
Stub Shaft
Curvic #4
1st Stage
Compressor
Curvic #3
2nd Stage
Stub Shaft
Curvic #2
Expander
Wheel
Curvic #1
Exhaust
160F
Steam Inlet
700F
Expander
Thrust
Faces
1st Stage Turbine

2nd Stage Turbine
3rd Stage Turbine
Curvic #5
Curvic #6
Curvic #7
Balance Plane #1
20 Axial Holes
Balance Plane #2
20 Radial Holes
1Y
30
Balance Plane #3
20 Radial Holes
30
60
1X
Rotation and Angular
Coordinates Viewed
from the Expander
CCW
Rotation
2Y
Balance Plane #4
30 Axial Holes
60
2X
Rotor Weight = 910 #

Rotor Length = 77.00"
Bearing Centers = 45.68"
Fig. 119 Combined Expander-Air Compressor-Steam Turbine Rotor Configuration

6 Robert
C. Eisenmann, Some realities of field balancing, Orbit, Vol.18, No.2 (June 1997), pp. 12-17.
549
cess hold point of 14,500 RPM using the outboard balance planes #1 and #4. This
was followed by a final trim at 18,500 RPM on the inboard planes #2 and #3
located next to the compressor wheels. It had been repeatedly demonstrated that
if the rotor was not adequately balanced at 14,500 RPM, it probably would not
run at 18,500 RPM. Hence the plant personnel were committed to performing a
field balance at 14,500 as well as 18,500 RPM.
Although the high speed balance at full operating speed was readily achievable, the intermediate speed balance at 14,500 RPM was always difficult. In an
effort to improve the understanding of this machinery behavior, the historical
balancing records were reviewed, and transient vibration data was examined. In
addition, the damped critical speeds plus associated mode shapes were computed
as previously discussed in case history 12.
One of the interesting aspects of this machine was the variation in balance
sensitivity vectors at 14,500 RPM. As discussed in this chapter, the balance sensitivity vectors provide a direct relationship between the rotor mass unbalance
vectors and the vibration response vectors. These vectors are determined by
installation of known trial or calibration weights at each of the balance planes,
and measuring the resultant shaft vibration response. Suffice it to say, these balance sensitivity vectors must remain reasonably constant in order for the vector
balancing calculations to be correct. For this particular rotor, three sets of sensitivity vectors were computed from the available historical information at 14,500
RPM, and the results of these vector calculations are summarized in Table 11-1.
Since this rotor contains two measurement planes, and four balance correction planes, a total of eight balance sensitivity vectors were computed using
equation (11-17). The first balance sensitivity vector identified as S11 in Table
11-1 defines the vibration response at measurement plane 1, with a calibration
weight installed at balance plane 1. Similarly, sensitivity vector S12 specifies the
vibration response at measurement plane 1, with a weight at balance plane 2,
Table 111 Balance Sensitivity Vectors Based On Steady State Data At 14,500 RPM
Data Set #1
Data Set #2
Data Set #3
S Vector
(Grams/Mil,p-p @ Deg.)
S11
20.3 @ 139
16.2 @ 179
Not Available
S12
48.1 @
34a
22.6 @ 309
76.4 @ 233
S13
42.6 @ 211
14.7 @ 177
24.6 @ 200
S14
34.1 @ 259
41.7 @ 289
16.2 @ 305
S21
18.3 @ 308
13.4 @ 345
Not Available
S22
19.1 @ 168
18.5 @ 142
18.6 @ 160
S23
32.3 @ 258
20.2 @ 269
31.2 @ 221
S24
24.2 @ 83
20.3 @ 147
14.9 @ 142
aShaded
vectors of questionable accuracy due to small differential vibration vectors with weights.
550
Chapter-11
and so forth throughout the remainder of the tabular summary. The two shaded
vectors in Table 11-1 are of questionable accuracy due to the fact that the differential vibration vector was less than 0.1 Mils,p-p. This small differential vibration is indicative of minimal response to the applied weight, and the validity of
the particular balance sensitivity vector is highly questionable. On much larger
machines, the validity of the sensitivity vectors would be considered marginal if
the differential vibration vectors were less than 0.5 or perhaps 1.0 Mil,p-p. However, for this small, high speed rotor, a differential shaft vibration value of 0.1
Mils,p-p was considered to be an appropriate lower limit.
Examination of the remaining S vectors in Table 11-1 reveals some similarities, but the overall variations are significant. For instance, the magnitude of
S12 varies from 22.6 to 76.4 Grams per Mil,p-p, and a 76 angular difference is
noted. On S24 the amplitudes change from 24.1 to 14.9 Grams per Mil,p-p, but the
angles reveal a 59 spread. At this point, a preliminary conclusion might be
reached that this rotor is indeed nonlinear and cannot be field balanced.
Further review showed that the vibration response vectors used for balancing were acquired at a process hold point of 14,500 RPM. Under this condition,
the machine speed was held constant, but rotor and casing temperatures were
changing as the process stabilized. This could be a major contributor to the
spread in sensitivity vectors in Table 11-1. Attempting to balance a machine with
these variable coefficients is difficult at best, and many runs are required to
attain a barely acceptable balance state.
Phase Lag (Degrees)
200
Expander Bearing
Probe #1Y
250
300
350
40
400
Turbine Bearing
Probe #2Y
450
90
4,000
5.0

Y-Axis Probes During A
Typical Machine Startup
6,000
8,000 10,000 12,000 14,000 16,000 18,000

Mode @
Full Speed @
7,800 RPM
18,500 RPM
4.0
Expander Bearing
Probe #1Y
3.0
Process Hold @
14,500 RPM
2.0
1.0
Turbine Bearing
Probe #2Y
0.0
4,000
6,000
8,000
10,000 12,000 14,000 16,000 18,000
Rotational Speed (Revolutions/Minute)
551
Variable speed vibration response vectors were extracted from the historical database, and a typical startup Bode is presented in Fig. 11-10. This data displays the Y-Axis probes from both measurement planes. Both plots are corrected
for slow roll runout, and the resultant data is representative of the true dynamic
motion of the shaft at each of the two lateral measurement planes. The major
resonance occurs at approximately 7,800 RPM, which is consistent with the analytical results discussed in case history 12. It is significant to confirm that the
process hold point at 14,500 RPM displays substantial amplitude and phase
excursions. This is logically due to the heating of the rotor and casing, plus variations in settle out of the operating system (i.e., pressures, temperatures, flow
rates, and molecular weights). Although this process stabilization is a necessary
part of the startup, the variations in vibration vectors negates the validity of this
information for use as repetitive balance response data.
At speeds above 14,500 RPM, there are additional vector changes, and a
desirable plateau in the amplitude and phase curves does not appear. The only
consistent area of essentially flat levels occurs in the vicinity of 14,000 RPM. To
test the validity of this conclusion, individual vectors at 14,000 RPM were
extracted from the historical transient startup files. These displacement vectors,
in conjunction with the installed weights, were used to re-compute the balance
sensitivity vectors with equation (11-17). The results of these computations are
presented in Table 11-2, and are directly comparable to Table 11-1.
By observation and comparison, it is clear that the consistency of S vectors
between the three data sets is far superior in the results presented in Table 11-2.
This applies to both the magnitude and direction of the computed balance sensitivity vectors. Hence, the repeatability and associated linearity of the balance
sensitivity vectors, plus the predictable balance response of the mechanical system was significantly improved by selecting a stable data set for computation of
the balance sensitivity vectors.
Table 112 Balance Sensitivity Vectors Based On Transient Data At 14,000 RPM
Data Set #1
Data Set #2
Data Set #3
S Vector
S11
24.7 @ 144
86.2 @ 188
Not Available
S12
125. @
333a
38.2 @ 228
42.4 @ 228
S13
31.8 @ 207
20.8 @ 184
22.9 @ 176
S14
21.8 @ 322
36.1 @ 321
30.7 @ 301
S21
24.9 @ 313
21.6 @ 311
Not Available
S22
17.3 @ 167
23.7 @ 175
18.9 @ 178
S23
78.7 @ 299
22.2 @ 221
27.4 @ 232
S24
27.6 @ 147
25.8 @ 166
28.3 @ 164
aShaded
vectors of questionable accuracy due to small differential vibration vectors with weights.
552
Chapter-11
SINGLE PLANE BALANCE

The simplest form of mass unbalance consists of weight maldistribution in
a single plane. This type of unbalance is characterized by an offset of the mass
centerline (inertia axis) that is parallel to the geometric centerline of the rotating assembly. This is often called a static unbalance, and in some cases it may be
detected by placing a horizontal rotor on knife edges, and allowing gravity to pull
the heavy spot down to the bottom of the assembly.
A static unbalance may occur in a thin rotor element such as a turbine disk
or a compressor wheel. The mass correction for this condition would probably be
very close to the actual plane of the unbalance. A static unbalance may also occur
in a long rotor such as a turbine or a generator rotor. In this situation, the correction plane may, or may not, be coincident with the location of the static unbalance. In any balance problem, the following basic questions must be addressed:
Lateral Correction Plane To Be Used?
Amount of the Balance Weight?
Angular Location of the Balance Weight?
The answer to the first question resides in an evaluation of the type of

unbalance, combined with the specific rotor configuration, deformed mode shape,
and accessible balance planes. The proper weight installed at the wrong plane
will not help the rotor. In all cases, the diagnostician must determine the modally effective balance planes, and then narrow the choices down to physically
accessible locations. For instance, a generator rotor may display a pure static
unbalance that ideally should be corrected at midspan. In reality, the placement
of a midspan weight in the generator rotor is not feasible, but the generator
unbalance may be corrected by weights installed at the accessible end planes.
The second basic balancing question addresses the amount or size of the
balance weight to be used. Ideally, previous information on the specific rotor, or a
similar unit, would be available to guide the diagnostician. In field balancing situations where historical data is unavailable, it is customary to install calibration
weights that produce centrifugal forces in the vicinity of 5% to 15% of the rotor
weight. Machines such as motors or expanders that rapidly accelerate up to
speed are candidates for initial weights that produce centrifugal forces in the
vicinity of 5% of the rotor weight. A more aggressive approach is often applied
towards machines such as steam turbines that may be started up slowly, and the
unit tripped if the weight is incorrect. For these types of machines, the installation of an initial weight that produces a centrifugal force equal to 15% of the
rotor weight has proven to be an effective starting point.
The third major balancing question of angular location of the weight is
often the most difficult to address. Some individuals take the approach of installing the initial balance weight at any angular location, and then computing the
vector influence. It must be recognized that this is an extremely dangerous
practice that can result in serious mechanical damage. In virtually all
cases, the weight should be installed to reduce the residual unbalance, and lower
Single Plane Balance
553
Outboard
End
315
0
45
270
90
270
225
225
135
180
Fig. 1111 Single Mass

Rotor Kit
R
C W ot
315
Bearing
Journal
Coupling End
the associated vibration amplitudes.

The correct angular location of the initial weight is again dependent upon
the type of unbalance, the specific rotor configuration, the deformed mode shape,
and the accessible balance planes. For a single plane balance and a simple mode
shape, the process is considerably simplified. For demonstration purposes, consider Fig. 11-11 of a one mass rotor kit. The mass is supported between bearings,
and proximity probes are mounted inboard of the bearings at both ends of the
rotor. Since this is only a single mass system, the dominant viable shaft mode
shape is a pure translational mode. At the critical speed of approximately 5,000
RPM the physical orientation of elements, and the maximum shaft deflection is
depicted in Fig. 11-12 of an undamped mode shape diagram.
Fig. 1112 Mode Shape

Of Single Mass Rotor Kit
At Translational First Critical Speed of 5,000 RPM
It is clear that the maximum deflection occurs at midspan, and that the
center mass is in a modally sensitive location. It is also apparent that the proximity probe locations will yield information that is representative of the synchronous 1X response. Therefore, a knowledge of the phase characteristics should
provide the information necessary for a logical angular weight placement at the
midspan mass.
One of the easiest ways to determine the proper location for a balance
weight was proposed by Charles Jackson7 in his article entitled Balance Rotors
by Orbit Analysis. Quoting directly from this paper, Jackson states that:
554
Chapter-11
the orbit represents a graphical picture of the shaft motion pattern. The
key-phase mark represents where the shaft is at the very instant the notch passes
the probeBelow the first critical, the mark on the orbit represents the location of
heavy spot of the shaft relative to that bearing. This point is difficult to see, yet
simple once it is understood; i.e., the shaft must be wherever it is because of either
external forces or mass imbalance. Limiting this discussion to imbalance, the
shaft is displaced by imbalance. The mark shows where the shaft is at that precise
instant when the notch passes the probe.
Therefore, if one would stop the machine and turn the shaft until the notch
lines up with the probe, the angular position of the shaft is satisfied. Then, laying
off the angle from the pattern taken on the CRT gives the heavy spot for correction. Weight can either be subtracted at this point or added at a point 180 degrees
diametrically opposite, on the shaftThe orbit diameter will reduce as correction
is applied. Should too much weight change be given, the mark will shift across the
orbit, indicating the weight added is now the greatest imbalance.
Above the first critical, the rules change. The key-phase mark would have
shifted approximately 180 degrees when the shaft mode of motion
changedTherefore, the phase mark will appear opposite the actual heavy spot
and the weight addition would be on the key-phase mark position
These statements are consistent with the previous discussion presented
earlier in this chapter. It is appropriate to apply these techniques to the single
mass rotor shown in Fig. 11-11. This rotor kit was run at speeds below and above
the first critical. The top portion of Fig. 11-13 displays the 1X filtered orbit and
Fig. 1113 Orbit And

Time Base Plots For An
Unbalanced Single Mass
Rotor Kit Running At 2,620
RPM Which Is Below The
Critical, And At 8,060 RPM
Which Is Above The Rotor
First Critical Speed
7 Charles Jackson, Balance Rotors by Orbit Analysis, Hydrocarbon Processing, Vol. 50, No. 1
(January 1971).
555
time domain plots that were extracted from the outboard X-Y probes at 2,620
RPM. This operating speed is below the 5,000 RPM balance resonance (critical
speed). The bottom set of 1X filtered orbit and time domain plots in Fig. 11-13
were obtained from the outboard X-Y probes at 8,060 RPM, which is well above
the translational critical speed.
Both sets of data reveal forward and reasonably circular orbits at the outboard end of the rotor kit. This data is vectorially corrected for shaft runout, and
the presented plots are representative of the true dynamic motion of this single
mass rotor. Similar behavior was observed by the coupling end X-Y probes. The
coupling end data was not included, since it would be redundant to the outboard
plots. However, during an actual field balance, vibration data would always be
obtained at both ends of the machine.
For purposes of completeness, both the low and high speed shaft orbits will
be evaluated for proper balance weight angular location. The diagram shown in
Fig. 11-14 describes the low speed shaft orbit at 2,620 RPM. The vertical probe
phase angle documented in Fig. 11-13 was 106. Thus, moving 106 in a counter
rotation direction (i.e., clockwise) from the vertical probe locates the high spot.
This high spot is coincident with the heavy spot in this simple example, and the
angular location is identical to the Keyphasor trigger point.
Add
Weight
Location
Fig. 1114 Shaft Orbit
And Probe Locations For
Rotor First Critical Speed
v=106
CC W
R otn
h=10
Remove
Weight
Location
The horizontal probe phase angle shown in Fig. 11-13 was 10. Rotating 10
in a clockwise direction from the horizontal probe in Fig. 11-14 locates the same
high spot (i.e., Keyphasor trigger). Thus, both probes have identified essentially
the same angular location, and this point is the high spot, and the Key trigger
point. Since this information was obtained below the first critical speed (translational resonance); the identified point would logically be coincident with the
residual heavy spot on the disk. Thus, weight should be removed at the Keyphasor dot (heavy spot) at nominally 3:30 oclock angular position. Alternately,
weight could be added to the 9:30 oclock position to correct for the unbalance at
the 3:30 oclock position.
In passing, it should be mentioned that the difference between the angular
location identified by the vertical and horizontal phase angles is not exactly the
same. In this example, a 6 difference is noted between the two angular locations. This is quite common behavior due to the fact that the orbit is not perfectly
circular. In chapter 7, it was shown that a perfectly circular orbit would appear
556
Chapter-11
only if the amplitudes in the orthogonal directions were equal, and the phase
varied by 90 between the two probes. In the example shown in Fig. 11-14, and in
most field balancing situations, the orbit is somewhat elliptical, and the measured phase angles differ from the pure 90 value. Hence, it makes sense to make
a weight correction between the two positions, and attempt to satisfy the vertical
as well as the horizontal vibration response.
Above the critical speed, the phase should increase by approximately 180,
and the Keyphasor dot should shift to the other side of the orbit. In fact, this
anticipated behavior was displayed by the bottom set of orbit and time base plots
presented on Fig. 11-13. This data acquired at 8,060 RPM, which is considerably
above the 5,000 RPM critical speed. Extracting the orbit from the high speed
data set at the bottom of Fig. 11-13, and including the measured phase angles,
the diagram in Fig. 11-15 was generated.
v=274
Add
Weight
Location
Fig. 1115 Shaft Orbit

And Probe Locations For
RPM Which Is Above The
Rotor First Critical Speed
C C W R ot n
Remove
Weight
Location
h=169
The high speed vertical probe phase angle was 274. Moving 274 in a
counter rotation direction (i.e., clockwise) from the vertical probe locates the high
spot (coincident with the Key trigger). Similarly, the horizontal probe phase
angle shown in Fig. 11-13 was 169. Moving 169 in a clockwise direction from
the horizontal probe locates basically the same high spot. Thus, both probes have
identified essentially the same angular location above the resonance. Since this
information was obtained above the first critical speed, the identified point
would be opposite to the residual heavy spot. Thus, weight should be added at
the Keyphasor dot around the 9 oclock position. Alternately, weight could be
subtracted at the 3 oclock position.
In retrospect, the data above the critical speed (Fig. 11-14) is identifying
the same general angular location as the data obtained below the critical (Fig.
11-13). The high speed orbit indicates a heavy spot at 3 oclock, and the low speed
orbit reveals a heavy spot that is somewhat lower at 3:30 oclock. The two values
would be identical if both orbits were perfectly round (circular), and a precise
180 phase change occurred through the critical speed range. However, these two
ideal conditions seldom occur on real machines, and the discussed data is representative of typical machinery behavior.
557
In many respects, field balancing consists of a series of compromises. In this

case, the average (low to high speed) weight removal should be in the vicinity of
100 clockwise from the vertical probe. However, the available balance holes in
this portion of the disk were empty, and there was no opportunity to easily
remove additional weight. The next compromise would be to add weight at the
light spot at approximately 280 (i.e., 100 plus 180). It was noted that weights
already filled the balance hole at 270, and the only remaining empty hole was at
292. A total of 0.5 Grams was installed at this 292 hole, and the resultant
response due to this single plane weight addition is presented in Fig. 11-16. As
before, the top set of orbit and time base plots at 2,620 RPM depict the shaft
vibration below the critical speed. The bottom set of plots in Fig. 11-16 were
acquired at 8,060 RPM, which is above the rotor balance resonance frequency.
Fig. 1116 Orbit And

Time Base Plots For A
Balanced Single Mass
Critical, And At 8,060 RPM
Which Is Above The Rotor
First Critical Speed
It is clear that the 0.5 Gram weight at 292 significantly reduced the synchronous 1X unbalance response. It should also be clear that the position of the
Keyphasor dot on the orbit is representative of the high spot. This concept is
fundamental to balancing, as well as the analysis and understanding of the
behavior of any rotating system.
These concepts might seem to be somewhat different from the automated
instrumentation installed on most low speed shop balancing machines. In actuality, the concept is the same, but there other significant differences. For instance,
during most shop balancing work, it is inexpensive to make a run, and there is
little physical risk to the machinery or the operator. In the case of field balancing, it is often difficult to change weights, and it is generally expensive to make a
full speed run. Furthermore, if an incorrect weight is used in a field balance, the
results may be hazardous to the machinery, and the health of the operator.
Each field balance shot should be a meaningful move, and it should contrib-
558
Chapter-11
ute to the overall database describing the behavior of the machine. Field balancing generally requires the quantification of the basic relationship between the
shaft response and the applied force as commonly expressed by:
Force
This general expression has been stated several times in this text due to
the fact that it has many specific applications in rotor dynamics. Within the balancing discipline, response is the measured shaft or casing vibration vector. The
applied force is represented by the unbalance vector, and the restraint may be
thought of as a stiffness vector. In balancing applications, this variable may be
considered as a spring-type parameter of a specific unbalance producing a specific deflection or rotor vibration. Another way to view this restraint term is to
consider it as the sensitivity of the machine to rotor unbalance. If these balancing terms are substituted for the equivalent values in the previous expression,
the following equation (11-1) evolves:
Unbalance
Vibration = -------------------------------Sensitivity
(11-1)
All variables in (11-1) are vector quantities. Each parameter carries both a
magnitude and a direction. If the initial rotor vibration is described by the A vector with amplitude in Mils,p-p, and the unbalance is defined by the U vector with
units of Grams, then the balance sensitivity S vector must carry units of Grams
per Mil,p-p. Using these designations, equation (11-1) may be rewritten as:
U
A = ----S
where:
(11-2)
A = Initial Vibration Vector (Mils,p-p at Degrees)

U = Mass Unbalance Vector (Grams at Degrees)
S = Sensitivity Vector to Unbalance (Grams/Mil,p-p at Degrees)
This expression may be easily remembered as the U = S A equation. In

either format, the vibration vector may be measured directly, and the technical
problem resolves to one of determining the mass unbalance vector based upon
some unknown sensitivity. This sensitivity vector may be experimentally determined by adding a known calibration weight at a known angular location to the
rotor, and measuring the vibration response vector. Assume that the calibration
weight vector is defined by W, and the resultant rotor vibration is identified as
the B vector. If the machine is re-run at the same speed and operating condition,
and if the system exhibits linear behavior, then equation (11-2) may be expanded
into the following expression:
559
U+W
B = ----------------S
where:
(11-3)
B = Vibration Vector with Calibration Weight (Mils,p-p at Degrees)

W = Calibration Weight Vector (Grams at Degrees)
Expansion of equation (11-3), and substituting (11-2) yields the following:

U+W
U W
W
B = ----------------- = ----- + ----- = A + ----S
S S
S
or
W
B A = ----S
From this expression, the balance sensitivity vector may be computed as:
W
S = ---------------
B A
(11-4)
The unbalance may now be determined from equation (11-2). It should be

noted that the measured vibration vector must represent actual dynamic motion
of the rotor. Hence, the vector must be corrected for any electrical and/or
mechanical runout. This is achieved by vector subtraction of the slow roll from
the vibration vector measured at balancing speed. The following equation (11-5)
is applicable for all measurements that require a slow roll or runout correction or
compensation (e.g., proximity probes). Note that other transducers (e.g., casing
accelerometers) do not require this type of vector correction.
Ac = A E
where:
(11-5)
A c = Runout Compensated Initial Vibration Vector (Mils,p-p at Degrees)

E = Slow Roll Runout Vector (Mils,p-p at Degrees)
Thus, the proper expression for calculation of the mass unbalance is now
easily derived from equations (11-2) and (11-5) as follows:
U = S Ac
(11-6)
As previously noted, the vibration vector amplitudes are measured in

Mils,p-p, the weight units may be expressed in Grams or Gram-Inches, and the
balance sensitivity vectors would carry the units of Mils/Gram or Mils/GramInch respectively. In all cases, the angular orientation is against rotation (i.e.,
560
Chapter-11
phase lag), from each respective vibration transducer. In addition, the trigger
point is established by the coincidence of the physical shaft trigger location, and
the center of the Keyphasor timing probe.
It should be mentioned that the plane of unbalance, the correction plane,
and the measurement plane are not defined as coincidental. They may be, and
usually are, separate planes in a machine assembly. It is important to recognize
that the previous equations are directed towards achieving a minimum value of
the vibration vector. That is, the calculations will yield a balance weight, located
at the correction plane, that is sized to minimize the vibration response at the
measurement plane to the actual mass unbalance distributed in the rotor. In the
majority of cases, this is both acceptable and agreeable. However, it is always
good practice to consider the shaft mode shape, and verify that the correction
weight does not aggravate deflections at other points along the rotor, while
reducing vibration at the measurement plane.
This same equation array is applicable to calculations performed both
above and below a critical speed. In all cases, the balance computation solves for
a zero response amplitude. A similar set of expressions may be developed for
multiple measurement and correction planes. It should be restated that the
equation set used for the calculations presumes a linear response of the mechanical system to mass unbalance. The presence of significant shaft preloads (due to
misalignment, gear contact forces, etc.), thermal effects, fluidic forces, bearing
instabilities, and various other mechanisms may render these calculations ineffective. However, for many conditions of rotational mass unbalance, the mechanical system will behave in a reasonably linear fashion.
In many instances, a single plane calculation is not totally adequate, and
cross-coupling between two or more correction planes must be considered. This is
achieved by expansion of the previously discussed equation set to multiple
planes. However, before addressing any additional complexity of the equation
structure, it would be advisable to examine the field application of these single
plane calculations to the following case history of a forced draft.
Case History 37: Forced Draft Fan Field Balance
A direct application of single plane balancing occurs on simple rotor systems that contain essentially one plane of unbalance. For example, consider the
forced draft fan rotor described in Fig. 11-17. This is a fully symmetrical rotor
that is nominally 177 inches long, and weights approximately 7,100 pounds. This
forced draft fan is driven by a steam turbine through a speed decreasing gear
box. Normal speed varies between 1,470 and 1,540 RPM. Unfortunately, the fan
translational resonance (critical speed) exists at 1,500 RPM, and the resonance
bandwidth is approximately 400 RPM. Thus, the first critical resonance persists
between speeds of 1,300 RPM [=1,500-(400/2)] and 1,700 RPM [=1,500+(400/2)].
It is apparent that under normal operating conditions, the fan runs within the
bandwidth of the first critical.
The situation is further complicated by the fact that the machine is
installed in an undesirable environment that allows ingestion of coke, plus other
561
Balance Plane
Journal
15" Long
6"
15"
8.5"
Journal
15" Long
6"
45 45
X
45 45
3/4" Thick Center Plate

58 " Outer Diameter
Total Weight = 7,100 Pounds Overall Length = 177 Inches
Fig. 1117 Rotor Configuration For Steam Boiler Forced Draft Fan
foreign objects into the fan. Hence, the fan blades are always under attack, and
mechanical damage to the blades results in changes to the balance characteristics at the middle of the rotor. Since the fan runs within the translational critical
speed domain, and since balance changes due to physical damage continually
occur at the midspan, this unit is quite susceptible to rapid changes in vibration.
When these undesirable events occur, the only reasonable solution is to correct
the change in mass unbalance by field balancing.
The fan is equipped with X-Y proximity probes at each bearing as shown in
Fig. 11-17. The probes are installed at 45 from true vertical, and a Keyphasor
is mounted at the outboard stub end of the fan shaft. It should be mentioned that
the bearings are supported on tall pedestals attached to a flexible baseplate. The
entire support structure is quite soft, and considerable casing motion occurs. To
include this information into the balance calculations, a casing vibration probe is
installed on each fan bearing housing. These casing probes are mounted in line
with the X-Axis shaft sensing proximity probes. This orientation allows the summation (electronic or by calculation) of the relative shaft signal with the absolute
casing vibration signal to obtain absolute shaft vibration.
The fan in question experienced substantial damage during one particularly violent ingestion of a large icicle. Two blades were damaged, and one bearing had a babbitt breakdown. These mechanical problems were corrected, and
the unit was restarted. Understandably, the blade repairs were responsible for a
major mass unbalance condition, and the startup was terminated at 900 RPM
with shaft vibration amplitudes in excess of 7.0 Mils,p-p.
Based on previous experience, an 1,190 gram (42 ounces) correction weight
was welded to the center plate at an angle of 28. The machine was restarted,
and the correction proved to be effective. This time a desirable balancing speed of
1,650 RPM was achieved. It is usually unwise to try and acquire balance
response vibration data close to a critical speed. This is due to the fact that small
changes in speed will result in significant changes in the 1X vectors. Hence,
when a machine runs in the vicinity of a resonance, it is usually good practice to
try and obtain the balance data in the plateau region above the resonance.
562
Chapter-11
In this case, the maximum attainable fan speed was 1,650 RPM, and this
speed was used for the remaining balance runs. Response data at both bearings
was similar, and for purposes of brevity, only the outboard bearing will be discussed. In this case, the runout compensated X-Y proximity probe data, and the
horizontal casing motion will be used for the balancing calculations. The Y-Axis
proximity probe was used as the zero degree reference for all of the weights. The
initial vibration response vectors at the outboard bearing at 1,650 RPM are summarized in the middle column of Table 11-3. The addition of a 567 gram calibration weight at the fan center plate at an angle of 40 produced the vibration
vectors in the right hand column of Table 11-3.
Table 113 Forced Draft Fan - Initial Plus Calibration Weight Vibration Vectors
Initial Vibration
(A Vector)
Vibration With Calibration

Weight Installed (B Vector)
Outboard Shaft Y-Axis
5.60 Mils,p-p @ 322
7.54 Mils,p-p @ 226
Outboard Shaft X-Axis
6.08 Mils,p-p @ 163
9.82 Mils,p-p @ 81
Outboard Casing X-Axis
3.85 Mils,p-p @ 144
5.79 Mils,p-p @ 48
Measurement Location
Unfortunately, vibration amplitudes have increased at all locations, and it

is clear that the calibration weight was placed at the wrong angle. It was mistakenly assumed that the calibration weight should go in the same angular location
as the 1,190 gram correction weight that was added to allow full speed operation.
In actuality, the 567 gram calibration weight should have been placed in the
vicinity of 320 instead of the 40 position. The measured amplitudes were high,
but they were within the tolerable range for a short duration run. A complete set
of information is now available to perform a single plane balance calculation. The
first step is to determine the balance sensitivity vectors from equation (11-4).
The calculations for the Y-Axis shaft proximity probe are shown as follows:
567 Grams 40
W
567 Grams 40
S = --------------- = ----------------------------------------------------------------------------------- = --------------------------------------
Mils
226
5.60
Mils
322
7.54
9.85 Mils 192

B A
S = 57.56 Grams/Mil,p-p 208
This value for the balance sensitivity vector may now be combined with the
initial vibration vector to determine the mass unbalance using equation (11-6):
U = S A c = { 57.56 Grams/Mil,p-p 208 } { 5.60 Mil,p-p 322 } = 322 Grams 170
The vector multiplication was performed in accordance with the rules
established by equation (2-39). Performing the same calculations for the X-Axis
shaft probe and the casing velocity sensor yields Table 11-4. This table summarizes the calculated mass unbalance for this fan as a function of each of the individual vibration measurement probes.
563
Table 114 Forced Draft Fan - Summary Of Calculated Weight Corrections

Unbalance Weight
Weight To Add
322 Grams @ 170
322 Grams @ 350
319 Grams @ 156
319 Grams @ 336
300 Grams @ 168
300 Grams @ 348
The center column of Table 11-4 identifies the magnitude of the unbalance
in grams, plus the location of the unbalance in degrees. If this rotor was to be
balanced by weight removal, then approximately 320 grams would be removed at
nominally 170 counter rotation from the vertical probe. On the other hand, if
balancing will be accomplished by weight addition, then the angular location
must be modified by 180 to determine the weight add vectors as shown in the
right hand column of Table 11-4. Thus, a weight of approximately 320 grams
should be added at an angle of nominally 350 counter rotation from the vertical
probe. In actuality, a 342 gram weight was welded onto the center plate at about
350. This initial correction provided a significant improvement to the 1X synchronous vibration response of this fan. Since additional time was available, a
small 30 gram trim correction was installed at 330. After this trim, the initial
and the final response vectors at 1,650 RPM are summarized in Table 11-5.
Table 115 Forced Draft Fan - Initial Versus Final Vibration Vectors
Initial Vibration Vectors
Final Vibration Vectors
5.60 Mils,p-p @ 322
0.46 Mils,p-p @ 104
6.08 Mils,p-p @ 163
0.68 Mils,p-p @ 60
3.85 Mils,p-p @ 144
0.23 Mils,p-p @ 246
The balance corrections identified by these vector calculations were quite

close together as demonstrated in Table 11-4. This type of calculated weight distribution makes it easy to select an appropriate and effective correction. Unfortunately, many machines do not produce such a consistent array of required
weight corrections. For these units, the diagnostician must carefully examine the
vibration response data, and either discount or completely ignore any questionable data. In many cases, field balancing has no unique solution. It is really a
compromise of balance weight selection based upon the calculations, plus a further compromise as to what can really be mounted on the machine.
As a final comment, it should be mentioned that the weights, and the vector
angles are all referenced to the vertical probe as 0 in accordance with the previous discussions. The horizontal probe calculations are also referenced to this
same location, and correct results are obtained. This is due to the self-canceling
nature of the calibration and correction weights within the final equation. The
validity of this self canceling relationship is demonstrated in case history 38.
564
Chapter-11
TWO PLANE BALANCE

The next level of unbalance complexity consists of weight maldistribution
in two separate geometric planes. The simplest form of this type of unbalance is
generally referred to as a couple unbalance. A pure couple is characterized by a
mass centerline (principal inertia axis) that intersects the geometric centerline
of the rotating assembly at the rotor center of gravity. This type of unbalance
produces a rocking or pivotal motion in the rotor assembly. Phase angles are separated by 180 across the rotor, and this type of excitation is most detrimental to
rotor response through the pivotal balance resonance.
In most process machinery, a pure couple seldom exists by itself. The couple
unbalance is usually combined with some type of static unbalance that may, or
may not, occur at the same planes as the couple. This weight distribution further
complicates the balancing problem, and corrections at more than two planes may
be necessary. The term Quasi-Static is often used to describe an unbalance condition where the mass centerline (principal inertia axis) intersects the geometric
centerline of the rotating assembly away from the rotor center of gravity. This
type of unbalance distribution may be sufficient to excite multiple critical speeds,
and the machinery diagnostician may experience difficulty in correcting a quasistatic unbalance. The problems associated with this type of unbalance range
from recognition of the behavior, to modal separation, and determination of
proper angular correction locations. For demonstration purposes, consider Fig.
11-18 illustrating of a two mass rotor kit.
Outboard
End
0
45
315
0
90
270
45
315
270
Bearing
135 Journal
90
270
225
180
225
135
180
Fig. 1118 Two Mass

Rotor Kit
R
C W ot
315
Coupling End
This is the same device used for discussion of the single plane unbalance,
with the addition of a second disk. Again, X-Y proximity probes are mounted
inboard of the bearings, and their location allowed detection of synchronous
motion. The two masses were positioned at the quarter points of the rotor to
enhance the pivotal mode response. The undamped mode shape at the translational first critical speed is depicted in Fig. 11-19 at a speed of 4,800 RPM. This
mode shape is virtually identical to the single plane mode shown in Fig. 11-12. In
both cases, the maximum deflection occurs at the midspan of the rotor, and modally effective balance weight corrections should be made at the middle of each
Two Plane Balance
565

For Two Mass Rotor Kit At
Translational First Critical
Speed Of 4,800 RPM
rotor for this first translational mode.

The calculated undamped mode shape at the pivotal or second critical speed
is displayed in Fig. 11-20. This mode occurs at a measured speed of 8,000 RPM,
and is characterized by the midspan nodal point, plus the out-of-phase behavior
across the nodal point. From these two mode shape diagrams, it is clear that a
static weight correction (same angular location) at both planes would influence
the first mode, but would be essentially self-canceling at the second critical. Furthermore, a pure couple correction between the planes would be ineffective at the
first mode, but it would have a significant influence upon the pivotal critical.

For Two Mass Rotor Kit At
Pivotal Second Critical
Speed Of 8,000 RPM
The actual transient speed behavior of this two mass rotor kit is depicted in
the Bode plot presented in Fig. 11-21. In this data array of 1X amplitude and
phase versus rotational speed, the response from the vertical probes at both ends
of the machine are documented. It is self-evident that a resonant response is
detected by both vertical proximity probes at 4,800 RPM, and 8,000 RPM. The
casual observer might identify these resonances as the first and second criticals,
and assume that they are the translational and pivotal modes for the rotor. In
this case, this assumption is totally correct, but the Bode plot does not directly
offer this information. The runout corrected 1X vectors in Fig. 11-21 display the
566
Chapter-11

Two Mass Rotor Kit Before
Balancing
same phase through the 4,800 RPM mode, and it is reasonable to conclude that a
translational mode is in progress. By the same token, the vertical probe phase
angles diverge by 180 through the 8,000 RPM resonance, and a pivotal mode
may be deduced. On a more complex machine, reaching these conclusions from a
single Bode plot would become much more difficult.
Another approach towards examination of this transient speed vector data
was presented by Donald E. Bently 8in his paper entitled Polar Plotting Applications for Rotating Machinery. Within this paper, Bently addressed the benefits
associated with a polar coordinate presentation of the variable speed vectors.
These obvious advantages included assistance in modal separation, plus
improved visibility during balancing. Replotting the transient vectors from the
Bode plot of Fig. 11-21 into a polar coordinate format produces the plots shown in
Figs. 11-22 (coupling end) and 11-23 (outboard end). From these two plots it is
clear that both ends of the rotor are moving together, or translating together, as
the rotor passes through the first balance resonance (first critical) at 4,800 RPM.
As rotational speed increases, the ends of the rotor begin to move in opposite directions as the unit enters, and then passes through the resonance at
8,000 RPM. From this runout compensated data display it is apparent that the
rotor is pivoting through this region, and the conclusion of a pivotal balance resonance is based upon factual evidence. These polar plots also display vector
directions for weight additions to correct the pivotal response at 8,000 RPM.
These vectors are indicated by thick lines, and they were determined by evaluating the rotor response specifically associated with the pivotal resonance. In most
8 Donald E. Bently, Polar Plotting Applications for Rotating Machinery, Proceedings of the
Vibration Institute Machinery Vibrations IV Seminar, Cherry Hill, New Jersey (November 1980).
Two Plane Balance
Fig. 1122 Coupling End Polar Plot Of

Two Mass Rotor Kit Before Balancing
567
Fig. 1123 Outboard End Polar Plot Of

Two Mass Rotor Kit Before Balancing
cases, the resonances should be viewed individually, and balance corrections or

other analysis performed on each individual resonance loop. The weight add
location is determined by drawing a vector from the start of the resonance loop,
to the conclusion of the resonance loop. The direction of this vector will identify
the weight location required to correct the particular mode under examination.
For example, the coupling end polar plot in Fig. 11-22 displays a desired directional correction in the vicinity of 147 (start to end of 2nd mode loop). Simultaneously, the outboard end polar plot in Fig. 11-23 requires a directional
correction in the vicinity of 320 (start to end of 2nd mode loop). The 173 difference between these two weight add locations is indicative of a pivotal resonance,
and the presence of a couple unbalance.
The first mode response will not be corrected in this example, and the balancing effort will be directed to a pure couple correction of the two mass rotor.
Based upon the available balancing holes in both masses, it was determined that
a couple weight set could be easily installed. At the inboard coupling end of the
rotor, a 0.5 gram weight was installed at 135 (147 desired from polar plot). At
the outboard mass, another 0.5 gram weight was also installed. At this location,
the weight was placed at the 315 hole (320 desired from polar plot). At both
ends of the rotor, the weights were installed as close as possible to the desired
angular locations, and a pure 180 couple was maintained between the same
magnitude weights installed at the same radius.
The validity of this approach is demonstrated in Figs. 11-24 and 11-25 that
display the polar plots following this couple balance correction. The initial plots
required a 20 Mil,p-p plot radius to contain the high response through the second
mode. However, following the couple correction, the vibration amplitudes were
significantly reduced. To obtain maximum visibility of the post balance data, the
568
Chapter-11

Two Mass Rotor Kit After Balancing

Two Mass Rotor Kit After Balancing
coupling end plot in Fig. 11-24 was placed on a 6.0 Mil,p-p radius, and the outboard end data in Fig. 11-25 was presented on a 4.0 Mil,p-p radius. It should be
noted that the vibration response through the second critical has been significantly reduced (almost eliminated), and the motion through the translational
first critical has been virtually unaffected by the couple balance shot. This is reasonable and expected behavior based upon the mode shape plots previously discussed for this two mass machine.
The Bode plot of this balanced condition is shown in Fig. 11-26. Again, note
the similar behavior through the first mode, and the virtual elimination of the

Two Mass Rotor Kit After
Balancing 2nd Mode
Two Plane Balance
569
pivotal second mode response. If this Bode plot was viewed on a larger amplitude
and phase scale, the uninitiated might not recognize the existence of a pivotal
mode on this machine.
Polar plotting is extraordinarily useful due to the direct relationship
between the plots and the physical mechanical system. In other sections of this
text it is always recommended that the diagnostician maintain a clear and consistent angular relationship between all elements. Thus, all calibration and correction weights, all vibration vectors, all hole locations, and all polar plots should
be referenced to zero degrees (0) at the probe. In all cases, the angular coordinate systems are laid off from the zero degree reference at the probe in a counter
rotation direction. Thus, the clockwise rotating system will have angles that
progress in a counterclockwise direction. The counterclockwise rotating system
will display angles that progress and increase in a clockwise direction. Failure to
maintain and enforce this simple angular convention can drive the simplest balance correction into the realm of the unattainable.
Polar plots are also useful for examining closely spaced resonances. This
includes split vertical and horizontal criticals, as well as closely coupled but distinctively different modes. Without polar plots the discrimination and relationship between many of these resonant responses may be easily misinterpreted. In
addition, secondary resonances such as structural or acoustic resonances may be
identified and properly separated from the major rotor balance resonances. Overall, polar plots should be used for improved field balancing, and for final documentation of the results. On very simple machines this may not be necessary;
but for complex machinery, the use of polar plots is considered to be mandatory.
In many cases, the balancing problem will be examined in terms of specific
vector quantities. Although a complex matrix solution could be constructed for
any number of planes, the reality of the field situation typically restricts the
allowable corrections to two independent planes. To simplify discussion of the
equation set, the measurements will be referenced to bearings 1 and 2 respectively; and weight corrections will be referenced to balance correction planes 1
and 2. Using this nomenclature, the response at each measurement plane is
equal to the vector summation of the unbalance response at each balancing
plane. This may be expressed by the following traditional two plane vector equations for the initial unbalance response of a linear mechanical system:
U1 U2
A 1 = --------- + ---------
S 11 S 12
(11-7)
U1 U2
A 2 = --------- + ---------
S 21 S 22
(11-8)
570
Chapter-11
where:
A 1 = Initial Vibration Vector at Bearing 1 (Mils,p-p at Degrees)

S 11 = Sensitivity Vector at Bearing 1 to Weight at Plane 1 (Grams/Mil,p-p at Degrees)
U 1 = Mass Unbalance Vector at Plane 1 (Grams at Degrees)
For purposes of clarification, the weights in these equations are expressed

in units of grams. Obviously, this is a metric mass unit, and not a weight unit in
the English system. The diagnostician may elect to use weights in terms of
ounces or pounds. If these units are used consistently, the vector equation structure will not be influenced. However, from a weight measurement standpoint,
most operating plants have a laboratory scale, or a triple beam balance that provides accurate weights in units of grams. Hence, grams will be used throughout
this text, and the associated examples and case histories.
For a linear system, the addition (or removal) of a calibration weight W1 at
plane 1 should vectorially sum with the existing unbalance U1. This presumed
linear weight summation may be included with the previous equations (11-7)
and (11-8) to produce the following new pair of vector equations:
U1 + W1 U2
= ---------------------- + ---------
S 11 S 12
(11-9)
U1 + W1 U2
B 21 = ---------------------- + ---------
S 21 S 22
(11-10)
B 11
where:
B 11 = Vibration Vector at Bearing 1 with Weight W1 at Plane 1 (Mils,p-p at Degrees)

W 1 = Calibration Weight Vector at Plane 1 (Grams at Degrees)
Removal of the calibration weight W1 at balance plane 1, plus the addition

(or removal) of a calibration weight W2 at balance plane 2 produces the following
pair of vector equations based upon the initial expressions (11-7) and (11-8):
Two Plane Balance
where:
571
B 12
U1 U2 + W2
= --------- + ----------------------
S 11 S 12
(11-11)
B 22
U1 U2 + W2
= --------- + ----------------------
S 21 S 22
(11-12)

In each of these six vector equations, the first subscript defines the measurement plane, and the second subscript describes the correction or balance
plane. Hence, this six equation array contains eight known vector quantities, i.e.,
the six vibration vectors, plus the two calibration weights. The calculation procedure initially solves for the four unknown balance sensitivity vectors, and finally
the two mass unbalance vectors are calculated.
The S11 vector is determined by first expanding equation (11-9), and then
substituting equation (11-7) in the following manner:
U1 + W1 U2
U1 W1 U2
W1
B 11 = ---------------------- + --------- = --------- + --------- + --------- = A 1 + ---------
S 11 S 12
S 11 S 11 S 12
S 11
This expression may now be solved for the first balance sensitivity vector:
S 11
W1
= ----------------------
B 11 A 1
(11-13)
S21 is determined in a similar manner by first expanding equation (11-10),

and then substituting equation (11-8) as follows:
B 21
U1 + W1 U2
U1 W1 U2
W1
= ---------------------- + --------- = --------- + --------- + --------- = A 2 + ---------
S 21 S 22
S 21 S 21 S 22
S 21
This expression may now be solved for the second sensitivity vector:
572
Chapter-11
S 21
W1
= ----------------------
B 21 A 2
(11-14)
Similarly S12 is determined by first expanding equation (11-11), and then

substituting equation (11-7) as follows:
U1 U2 + W2
U1 U2 W2
W2
B 12 = --------- + ---------------------- = --------- + --------- + --------- = A 1 + ---------
S 11 S 12
S 11 S 12 S 12
S 12
This expression may now be solved for the third balance sensitivity vector:
S 12
W2
= ----------------------
B 12 A 1
(11-15)
Finally, S22 is determined by first expanding equation (11-12), and then

substituting equation (11-8) as follows:
B 22
U1 U2 + W2
U1 U2 W2
W2
= --------- + ---------------------- = --------- + --------- + --------- = A 2 + ---------
S 21 S 22
S 21 S 22 S 22
S 22
This expression may now be solved for the fourth balance sensitivity vector:
S 22
W2
---------------------=
B 22 A 2
(11-16)
There is a clear pattern to the development of the balance sensitivity vectors. If these calculations are generalized, the following expression provides a
general solution for balance sensitivity vectors for a multilplane solution:
Wp
S mp = ---------------------------
B mp A m
(11-17)
In equation (11-17), the subscript m specifies the measurement plane, and

the subscript p identifies the weight correction plane. This general expression is
identical to the previously developed equations (11-13) through (11-16). It may
be applied in any balance situation where a weight is installed, vibration data is
Two Plane Balance
573
acquired at speed, and the weight is then removed prior to the installation of the
next calibration weight. For example, the balance sensitivity vectors summarized in Tables 11-1 and 11-2 were computed using equation (11-17).
Combining the solutions for the four balance sensitivity vectors within the
initial equations (11-7) and (11-8) yields the following result for mass unbalance
at both correction planes:
( S 12 A 1 ) ( S 22 A 2 )
U 1 = ------------------------------------------------------------S S
12 22
-------- --------

S
11 S 21
( S 21 A 2 ) ( S 11 A 1 )
U 2 = ------------------------------------------------------------S S
21 11
-------- --------

S 22 S 12
(11-18)
(11-19)
Correcting for proximity probe slow roll runout at both measurement

planes, the above two equations should be more properly expressed in terms of
runout compensated initial vibration vectors as follows:
S A S A
1 c 22
2 c
12
U 1 = -----------------------------------------------------------------S S
12 22
-------- --------

S 11 S 21
S A S A
2 c 11
1 c
21
U 2 = -----------------------------------------------------------------S S
21 11
-------- --------

S
22 S 12
(11-20)
(11-21)
This final pair of equations (11-20) and (11-21) may be used to calculate a
two plane balance correction. The individual expressions for balance sensitivity
vectors may also be used separately to compare balance response characteristics
between different rotors, or for repetitive calculations on the same rotor.
It is important to recognize that the previous array of balancing equations
are predicated upon an explicit application sequence of the calibration weights
(W1 and W2). Specifically, the first calibration run is performed with weight W1
mounted at balance plane 1. At the conclusion of this first calibration run, the
weight W1 is removed from the machine, and a calibration weight W2 is attached
574
Chapter-11
at balance plane 2. Following the conclusion of the second calibration run, weight
W2 is also removed from the machine. The unbalance calculations presented in
equations (11-20) and (11-21) represent the effective rotor unbalance at each correction plane, irrespective of the balance calibration weights.
The calculated mass unbalance vectors (U1 and U2) represent the amount
of weight that should be used at each balance correction plane. The angles associated with these unbalance vectors represent the angular location of the mass
unbalance. Hence, weight can be removed at the calculated angles, or an equivalent weight may be added at the opposite side of the rotor. That is, if weight must
be added, the weight addition angle would be equal to the calculated mass unbalance vector angle plus or minus 180. Although it is generally desirable to
remove weight from a rotor during balancing, there are many situations when it
is proper to add balance correction weights.
In some mechanical configurations, balance weights cannot be installed on
the machinery, and corrections must be performed by weight removal. In these
cases, the weight changes are performed by grinding, or drilling balance holes. In
other situations, the applied calibration weights may significantly reduce the
vibration amplitudes, and there might be a reluctance to remove weights that
provided a positive influence on the machinery. In both of these scenarios, the
applied calibration weights become part of the final correction weights. Furthermore, the computation of balance sensitivity vectors must be modified to accommodate this change in weight attachment or removal sequence.
In recent years, balance calculations have been significantly improved by
using smaller and faster personal computers. Machines such as the Apple Macintosh, IBM, or Compaq provide substantially more capability. All of these computers come in portable laptop configurations (less than 8 pounds) that can
easily be carried to a plant site for field balancing work. The utilization of spread
sheet programs such as Microsoft Excel allow rapid data entry combined with
almost instantaneous calculations. Hard copy documentation of the balancing
data and calculations is easily achieved with a variety of small LaserJet or Ink
Jet printers. In essence, these improvements in portable computers have allowed
the machinery diagnostician to concentrate on the actual balance problem
instead of the intricacies of the multiple vector manipulations.
The same scenario applies to the vibration data acquisition and processing
portion of the balancing work. Modern digital systems allow the capture of transient startup and coastdown data, plus steady state information at various loads
or heat soak conditions. The integration of these digital instrumentation systems
with the laptop computer and the portable printer provides excellent capability
for acquiring and printing the Bode and polar plots, plus the constant speed orbit
and time base data. Hence, the diagnostician has many tools to acquire a variety
of data, perform many complicated calculations, and generate the necessary
hard copy documentation in the field. This not only improves the quality of the
machinery balancing, it also minimizes the time required to perform the work,
and provides improved confidence to the selected balance shots.
With all of the available tools, the machinery diagnostician may have some
Two Plane Balance
575
fundamental questions like what do I use, and when do I use it? This is not a
casual issue, it is a very serious question regarding the application of the available tools and techniques. Unfortunately, some people try to address field balancing with a cookbook approach that can be used on any machine, at any time.
These types of canned techniques are destined to failure from the beginning. In
all cases, the diagnostician must examine the various facets of the machinery
problem, and select the appropriate measurements, instruments, and calculations that will solve the problem. This means that you have to know what youre
doing instead of blindly following some general procedure.
In an effort to provide some realistic direction, the following three case histories 38, 39, and 40 are presented for consideration. These are three different
configurations of generator drives that exhibit three different balance problems.
The discussions associated with each of these field case histories are quite
detailed, and the specific balancing logic is presented for each unit.
Case History 38: Five Bearing, 120 MW Turbine Generator Set
IP Exhaust to LP Inlet
IP Inlet from Reheat
HP Inlet at 1,900 psi
HP Exhaust
to Reheat
The five bearing turbine generator set depicted in Fig. 11-27 consists of two
turbine cases driving a hydrogen cooled, two pole synchronous generator. The
electric generator produces 3 phase, 60 Hz power at 13,800 volts. It is rated at
134,000 KVA, with a power factor of 0.85. The high pressure HP turbine accepts
superheated inlet steam at 1,900 Psi, and it exhausts to a reheat section before
readmission into the intermediate pressure IP turbine. Exhaust from the IP turbine is directed to a double flow LP turbine that exhausts to a surface condenser
at 2 inches of mercury absolute. As shown in Fig. 11-27, the HP and IP rotors are
contained in one case, and the inlet steam to the double flow IP is at the center of
the machine. This configuration places the large diameter low pressure turbine
Wheel Balance
Plane#3
Shaft
Balance
Plane#2
105
CCW
Rotation
Generator
Double Flow LP
Steam Turbine
HP-IP Steam Turbine

A
Cpl.
B
Cpl.
134,000 KVA
13,800 Volts & 5,600 Amps
3,600 Rpm, 3, 60 Hz
Transducers & Rotation

Viewed From The Turbine
Looking To The Generator
1Y
1X
45 45
2Y
2X
3Y
45 45
3X
4Y
4X
5Y
5X
45 45
45 45
45 45
Bearing #3
Journal =13.0
Clearance:
Vert.= 16 Mils
Horiz.= 27 Mils
Bearing #4
Journal =13.0
Clearance:
Vert.= 19 Mils
Horiz.= 32 Mils
Bearing #5
Journal =13.0
Clearance:
Vert.= 16 Mils
Horiz.= 27 Mils
LP Exhaust at
2" Hg Abs.
Bearing #1
Journal =8.0
Clearance:
Vert.= 10 Mils
Horiz.= 17 Mils
Bearing #2
Journal =15.0
Clearance:
Vert.= 21 Mils
Horiz.= 36 Mils
Fig. 1127 Machinery Arrangement Of Five Bearing Turbine Generator Set
576
Chapter-11
wheels at both exterior ends of the LP casing.

All five journal bearings are elliptical, and their respective vertical and horizontal clearances are shown in Fig. 11-27. The largest clearances are located at
the 15 inch diameter journal at bearing #2. In this location, the vertical diametrical clearance of 21 Mils is combined with a 36 Mil horizontal clearance. This is
the largest journal diameter in the train, and it must also accommodate some
significant thermal changes. The A coupling between the HP-IP and the double
flow LP turbine, plus the B coupling between the LP turbine and the generator
are solid couplings. These two couplings must be carefully assembled in conjunction with a sling check at each location.
Prior to the peak generating season, this machinery train was subjected to
a major six week outage for a complete maintenance overhaul. During disassembly, it was discovered that the #1 bearing and journal were damaged, and the
remaining four bearings had excessive clearances. A set of new buckets were
installed in one row of the IP rotor, and the last stage blades on both ends of the
LP rotor were replaced with a new and longer blade design. Several new diaphragms were installed, and various other stationary items were repaired or
replaced as required. An 8.0 Mil bow was documented on the IP rotor, but it was
not corrected. During reassembly, the LP turbine was balanced in-place at a
speed of 250 to 300 RPM by OEM personnel.
The rebuilt machinery train experienced a variety of routine mechanical
and electrical problems that were sequentially corrected. After eight days of various runs, the turbine generator set was finally placed on-line. The shaft vibration monitors indicated generally acceptable displacement amplitudes. The
dominant motion at all measurements points occurred at the shaft rotational
speed of 3,600 RPM. Specifically, the shaft 1X running speed orbits at each of the
five main bearings are shown on Fig. 11-28 at a constant load of 112 megawatts.
Fig. 1128 Initial Runout Compensated Orbits At Each Train Bearing At 112 Megawatts
These 1X orbits are all compensated for shaft runout, and they are considered to be representative of the absolute shaft vibration relative to the proximity
probes mounted at each main bearing. A normal forward precession combined
with low vibration amplitudes are displayed by bearings #1 and #2 on the HP-IP
turbine. The Y-Axis probe amplitude on bearing #3 approaches 4.7 Mils,p-p, but
the horizontal bearing clearance is 27 Mils. Hence, the runout corrected shaft
vibration is only 17.4% of the available diametrical clearance. Typically, a shaft
Two Plane Balance
577
vibration of 4.7 Mils,p-p would be a cause for concern on smaller units. However,
on a large machine such as this turbine generator set, this shaft vibration level
may be higher than desired, but it is still acceptable for operation.
Both generator orbits were predominantly vertical and a reverse precession
was noted on the inboard bearing #4. These generator vibration levels were also
higher than anticipated, but they were considered to be manageable. It should be
mentioned that the radial shaft vibration levels were lower than the values
logged before the overhaul. From the standpoint of the panel mounted vibration
monitor readings, the machinery train was in good condition. Unfortunately, a
significant casing or structural vibration was emitted between the LP turbine
and the generator. The severity of this excitation was such that the control room
operators were acutely aware of this problem since it produced a physically
uncomfortable sensation in the main control room.
This unusual vibratory behavior occurred predominantly at shaft rotational
speed, and casing measurements revealed that it was strongest in the axial
direction at the #3 bearing. Various load and speed variation tests were performed, and it was repeatedly demonstrated that this casing 1X component varied between 9.0 and 18.0 Mils,p-p axially. For instance, the expanded Bode plot
from one cold startup is displayed in Fig. 11-29. This diagram depicts the 1X
amplitude and phase from the horizontal #3X and vertical #3Y shaft vibration
140
BBBBBBBBB
Phase Angle (Degrees)
160
HJ HJ HJ HJ J J
H H HJ HJ HJ HJ
HJ HJ HJ HJ HJ HJ HJ J J J
200
H H H HJ HJ HJ J J J
H H H HJ J
220
HHJ
111111
11111111111111
240
11111
1 1 11
8.0
BBBBBBBB
B
180
260
3 000
10.0
Fig. 1129 Expanded

Bode Plot Of Synchronous
1X Shaft And Casing
Vibration At The #3 Bearing On The LP Turbine
BBBBBBBBBBB
3 200
H
J
1
B
3 400
Case Axial
Vert Case
Axial Case
Vert Shaft #3Y
JJ
Horiz Shaft #3X
J
J
3 600
JJ
JJ
JJ
B BB
B B B B B B BJ BJ BJ BJ BJ BJ B B BJ BJ B B B B B B B B B B
J
4.0
J
J
JJ
H H H H HH
JJJ
H 1H 1H 1H 1H 1H 1H 1H 1H 1 1 1 1 11
H
H
H
H
H
H
H
H
H
1
H
H
H
1
1
H
1H 1 1 1 1 1 1 1 1 1 1 1
6.0
2.0
3,000
3,200
3,400
3,600
578
Chapter-11
probes between 3,000 and 3,550 RPM. It also includes the 1X casing motion measured in the vertical and axial planes at the same bearing. This casing data was
acquired on the bearing cap with medium frequency range accelerometers, and
double integrated to obtain casing displacement.
The phase data at the top of Fig. 11-29 reveals nothing unusual. It is noted
that the phase differential between the X and Y shaft probes remains fixed at
about 85. Furthermore, all four transducers exhibit a nominal 20 change across
the plotted frequency range. This startup is typical for this class of machine, and
it is not considered to be representative of any type of resonant response.
Examination of the 1X amplitude data presented in the bottom half of Fig.
11-29 reveals moderate changes in synchronous motion for the three radial
transducers. However, the axial casing vibration increased from 3.9 Mils,p-p at
3,240 RPM to an amplitude of 9.7 Mils,p-p at 3,550 RPM. Hence, a 10% speed
change was accompanied by a 250% increase in axial casing vibration. Although
this appears to be a resonant-like response, the companion phase data does not
support the presence of a casing resonance or rotor critical speed.
Another peculiarity was noted on bearing #3 when oil drain temperatures
were checked. The temperatures on the #1 and #2 bearings on the HP-IP turbine
consistently ran between 150 and 155F. The two generator elliptical bearings
operated between 140 and 146F. However, the LP turbine #3 bearing oil drain
thermocouple never exceeded 126F. This thermocouple was replaced, and the
reading confirmed with a local dial thermometer. Any way the measurement was
made, the oil drain temperature from bearing #3 remained around 126F.
It was speculated that perhaps bearing #3 was running unloaded. This
would account for the lower than average oil drain temperatures, plus the higher
than average shaft vibration amplitudes. Unfortunately, this concept was not
supported by the shaft centerline position data. For instance, Table 11-6 summarizes the vertical shaft position of the journals within each bearing.
Since this train was equipped with elliptical bearings at all locations, the
total shaft centerline position change consisted of a horizontal shift in the direction or rotation, plus a vertical lift. The data presented in Table 11-6 summarizes
the vertical clearance at each elliptical bearing, and the measured vertical shaft
lift in Mils (based on X-Y proximity probe DC gap voltages). To provide a comparTable 116 Vertical Position Of Journals With Respect To Each Bearing At 112 Megawatts
Bearing Location
Vertical Bearing
Clearance
Vertical Shaft
Rise
Percent Journal
Rise In Bearing
HP Turbine Bearing #1
10 Mils
4 Mils
40%
IP Turbine Bearing #2
21 Mils
8 Mils
38%
LP Turbine Bearing #3
16 Mils
4 Mils
25%
Generator IB Bearing #4
19 Mils
7 Mils
37%
Generator OB Bearing #5
16 Mils
7 Mils
44%
Two Plane Balance
579
ison of lift versus clearance, the fourth column in Table 11-6 was used for the
ratio of these two parameters. Clearly, the journal at bearing #3 is sitting the
lowest in its respective bearing, with a total lift equal to only 25% of the available clearance. Hence, the shaft centerline position data indicated a loaded,
rather than an unloaded #3 bearing.
Obviously the oil temperature and shaft vibration indicators were in conflict with the journal centerline position data. In an effort to resolve this contradiction, the #3 bearing was elevated by 5 Mils with stainless steel shims. After a
restart, and a normal load and temperature stabilization, it was evident that
vibration amplitudes had decreased slightly at the #3 bearing. Unfortunately,
the shaft vibration had increased at each of the other four bearings. Although the
casing axial vibration was temporarily reduced to 6.4 Mils,p-p, the increases at
the other machine train bearings were considered to be unacceptable.
Once again, the oil drain temperature at #3 bearing was about 20F lower
than the other four bearings, and the array of shaft centerline positions
remained virtually identical to the behavior described on Table 11-6. Hence, the
evidence in support of an unloaded #3 bearing was beginning to dwindle. This
issue was finally put to rest when the operating logs from the past few years
were examined. Within this database, it was clear that the drain temperature
from bearing #3 was always about 20F lower than the other bearings. Hence,
the lower oil drain temperature on #3 bearing could not be associated with an
unloaded LP turbine bearing.
Since the generator still displayed flat elliptical orbits, and since the generator vibration had increased with the 5 Mil rise on the #3 bearing alignment
across the B coupling was questioned. At this point, the OEM elected to raise
the generator by 8 Mils, and see if that helped the situation. After the train was
restarted, it was clear that the elevated generator was contributing to the problem. Under this condition, axial casing vibration amplitudes at the #3 bearing
exceeded 18.0 Mils,p-p. Obviously, the project was moving in the wrong direction,
and everyone was growing weary of this problem.
External lagging was removed from the LP turbine, and additional vibration readings were acquired on the turbine casing and associated structural elements. This data did not identify any specific mechanical component that could
be a contributor or source of the high axial casing motion. Next, the machine was
shutdown and allowed to cool. The elevation shims were removed from underneath the #3 bearing (5 Mils), and the generator (8 Mils). This restored these elements back to their original vertical alignments. The LP turbine outer casing
was removed, and each strut and cross brace were examined for any indication of
a loose structural member within the turbine shell. This inspection extended
down into the condenser, and no loose or broken structural members were found.
At this stage, it was clear that there were no contributing structural problems. Furthermore, the original alignment across the B coupling was deemed to
be proper and acceptable. Since the forcing frequency of the axial casing vibration was at running speed, the attack plan evolved to do everything possible to
minimize the synchronous 1X running speed vibration. Although there was no
580
Chapter-11
evidence of abnormal axial vibration of the HP-IP rotor, it was clear that radial
vibration of the LP turbine could be improved. Hence, when all else fails, go back
to the basics, and try to eliminate the 1X driving force. This includes any eccentricity, bow, or mass unbalance in the rotor system.
Initially, the B coupling between the LP turbine and the generator was
disassembled. The #3 bearing was removed, and a sling (or horizontal swing)
check was performed at the #3 bearing location. An overhead scale was used to
establish the proper vertical load on the free shaft, and dial indicator readings
were acquired at 45 increments. The documented runout during initial assembly was 6.0 Mils, and the current maximum runout was found to be in excess of
30 Mils. In addition, these readings were not repeatable. At this point, it was
clear that other problems existed, and the #1 and #2 bearings were opened for
inspection. The turbine #1 bearing was in good condition, but IP bearing #2 had
a section gouged out of the babbitt. The loose babbitt in #2 bearing probably contributed to the erratic sling checks. There was no doubt that this bearing would
have experienced an early failure if the loose babbitt had not been detected.
Nevertheless, the damaged #2 bearing was scraped, cleaned, and reassembled. This bearing repair resulted in consistent swing checks at the #3 bearing.
The two turbine rotors were realigned at the A coupling, and the final coupled
sling check runout at the #3 bearing was nominally 5.0 Mils. The #3 bearing was
then re-installed, and the B coupling was reassembled to factory tolerances.
The 8.0 Mil bow on the IP rotor was still present, but there was no realistic
opportunity to correct this bow in the available time frame.
The assembled turbine generator set was then restarted to obtain a fresh
set of baseline data. The next step was to trim balance the LP turbine rotor running between bearings #2 and #3. This was initially hampered by a midnight
startup combined with an additional bow on the HP-IP rotor. Problems like this
happen when long hours are combined with production pressures to get the
machinery on-line. In most situations, it is best for the key personnel to get some
sleep, and try running the machine during the daylight hours. In this specific
case, the controlled restart on the following morning proved to be successful, and
a new set of baseline or initial reference data was obtained at speed and no load.
Although the temptation exists to perform several runs followed by simultaneous shots on multiple balance planes, it is usually wise to begin with something more realistic like a single plane balance correction. For this machine, the
largest radial vibration amplitudes occurred at the #3 bearing, and it makes
sense to perform the first correction at this location. As noted in Fig. 11-27, a balance plane exists on the outer face of the exhaust wheel adjacent to the #3 bearing. A diagram of this balance plane with additional physical details is presented
in Fig. 11-30. This location is a view from the governor end of the HP turbine,
and it is immediately apparent that the 0 reference point is located at the
Keyphasor probe. This transducer is positioned 15 below the horizontal centerline on the left side of the machine. Since the Y-Axis probe is 45 above this
centerline, the total angle between transducers is 60 as noted. The X-Axis proximity probe is another 90 against rotation, and it is shown at 150 on Fig. 11-30.
Two Plane Balance
581
90
Y-Axis
Probe
120
60
150
X-Axis
Probe
CCW Rotation
30
180
10
15
20
Radius (Inches)
0
210
BB
BBBBBBBBMBMBM
MMMM
Keyphasor
330
300
Fig. 1130 Weight Configuration And Angular

Reference Defined on
Wheel Balance Plane #3
Existing 12 Weights in
Inner Groove = 605 Grams
240
270
7 New Weights Installed in

Outer Groove = 369 Grams
Inner Groove Radius = 13.04, and One Weight = 4.4

Outer Groove Radius = 14.92, and One Weight = 3.8
As discussed throughout this text, the vibration phase measurements are

always referenced to the physical location of the vibration probes. Ideally, all
phase measurements, weight angles, plots, and computed vectors are referenced
to the same radial probe position. From a practical standpoint this is not always
possible, and a slight modification in the thought process is necessary. Specifically, if we consider a single plane unbalance condition as described by equation
(11-2), and if the vector sensitivity from a calibration run is incorporated from
equation (11-4), the following result is obtained.
W
WA
W
U = S A = --------------- A = ---------------- = --------------------------B A
B A
( B A) 1
Division of the B vector by the A vector in the denominator of the above
expression requires a subtraction of angles (i.e., -). Clearly, wherever the reference point is located, the subtraction of vibration vector angles negates any reference to a specific or fixed angular location. Hence, the angle of the calculated
unbalance U in the above equation is totally dependent on the angular reference
used for the calibration weight W. If the shaft keyway is used as the 0 reference
point for the calibration weight (as per Fig. 11-30), then the calculated unbalance
will be referenced to the same angular reference point. This obvious simplification makes the field work a lot easier, since it can be correctly stated that:
keyway is 0, and all angles are counter-rotation from the keyway
582
Chapter-11
This makes life much easier for everyone on the jobsite, and it provides a
common and easily understandable angular reference scheme. Furthermore, if
youre looking from the turbine to the generator, or the generator to the turbine,
the angular location will come out properly (since it is stated as counter-rotation). It is also a good idea to use a bright colored metal marking pen to identify
the balancing weight angles on the wheel or shaft. Some balance planes have to
be accessed through the condenser inlet piping. This is a hot and miserable environment to work in, and it is easy to lose your sense of direction and perspective.
Thus, any type of pre-planned or identified angular layout will be extraordinarily
useful in getting the weights installed at the correct angles.
Getting back to the trim balance of the LP turbine, the axial face of the LP
wheel adjacent to #3 bearing was equipped with two trapezoidal shaped grooves
for similarly shaped balance weights. As shown in Fig. 11-30, the inner groove
has a 13.04 inch radius, and it contained 12 weights totaling 605 grams. These
weights were installed during the low speed field balance by OEM personnel.
The circumferential length of the inner groove was 81.93 inches (= 2 x x 13.04).
Since the balance weights were approximately one inch long, each weight covered an arc of about 4.4 (= 360/82). The outer balancing groove, with a radius of
14.92 inches was used during this field balance. The total length of this outer
groove was 93.75 inches (= 2 x x 14.92). For one inch long balance weights, each
weight would cover an arc of about 3.8 (= 360/94).
The angle of the initial calibration weight was based upon the 1X vectors at
the #3 bearing. Specifically, the center orbit on Fig. 11-28 showed a Y-Axis vector
of 4.58 Mils,p-p at an angle of 292. Since this turbine runs above the first critical, the phase angle would be representative of the high spot, and this would be
the location for adding weight. Since the Y probe is 60 counter from this physical keyway reference, the actual weight installation angle should be 352
(=292+60). At the same time, the X-Axis vector was 2.55 Mils,p-p at 138. Since
the X probe is 150 away from the keyway, the weight installation angle should
be 288 (=138+150). Thus, the Y probe calls for a weight addition at 352, and
the X probe wants weight added at 288. Both transducers are calling for a
weight in the lower left-hand quadrant of Fig. 11-30. Clearly, the angular difference is due to the ellipticity of the shaft orbit. For purposes of simplicity during
this initial installation, the weights were mounted at 0 to straddle the keyway.
The amount of weight to add was determined by applying a form of equation (11-64) to this 22,000 pound turbine rotor in the following manner:
2
291
291
Cal. Weight = Rotor Weight --------------- = 22000 ------------ = 144 Oz-In.
RPM
3600
As previously stated, the radius for the outer balance groove was 14.92
inches, and that would require a weight addition of 9.65 ounces (=144/14.92).
Since one ounce weighs 28.35 grams, the initial weight addition should be in
vicinity of 274 grams (=9.65 x 28.35). The available balance weights were about
55 grams each, and this 274 gram shot would need at least five weights. This
Two Plane Balance
583
amount of weight would also cover a 19 arc (=5 x 3.8). Obviously, the larger the
arc, the more offsetting the weights become. Hence, a vector summation of five
weights at 3.8 increments would yield an effective weight that must be lighter
than the simple arithmetic weight sum. To adjust for this weight spread, a total
of six weights were used. This provided a first calibration run using 310 grams
installed in this outer groove, straddling the 0 keyway.
At full speed of 3,600 RPM the 1X vibration levels at the #3 bearing were
reduced by 0.7 Mils,p-p, and the synchronous amplitudes at #2 bearing increased
by approximately 0.4 Mils,p-p. Fortunately, the 1X vectors at HP bearing #1, plus
both generator bearings were not appreciably influenced by this initial weight at
balance plane #3. Due to the vibration increase at bearing #2, it was evident that
a simultaneous weight correction would be necessary at the opposite end of the
LP turbine (close to #2 bearing).
The second calibration run was performed by adding 160 grams on the LP
shaft balance plane adjacent to the A coupling. At this location, two circumferential balance weight grooves are cut into the shaft. The forward groove by the
A coupling contained 9 weights totaling 448 grams. These weights were also
installed during the low speed field balance by OEM personnel. The aft groove
closest to the LP wheel was used for the installation of the 160 grams at 295
clockwise from the keyway. The previous six weights of 310 grams at 0 mounted
on the opposite end of the LP rotor (plane #3) remained in place. The vibration
response with these weights resulted in decreased vibration levels at both #2 and
#3 bearings. In addition, the vectors at the #1 bearing increased slightly, and the
shaft vibration at generator bearings #4 and #5 remained fairly constant.
Based on these three runs, a series of two plane balance calculations were
performed. The four balance sensitivity vectors (S11, S12, S21, and S22) were computed with equations (11-26) through (11-29). The mass unbalance calculations
were performed with two plane equations (11-20) and (11-21). The total weight
add results of these balance calculations are presented in Table 11-7. It was easily agreed upon that the LP rotor required a static shot at nominally 270 at both
ends of the rotor. The magnitude of the correction was subjected to additional
debate. It is apparent that large weight corrections are difficult to perform due
the self-canceling nature of weights distributed over a large portion of the circumference. Furthermore, it was demonstrated the additional weights placed in
Table 117 Weight Additions Based Upon Two Plane Balance Calculations Plus Vector
Average Weight Additions, And Summary of Final Balance Weights Installed
Balance Weight Origin
Shaft Balance
Plane #2
Wheel Balance
Plane #3
Based on Y-Axis Probes
1,118. Grams @ 245
648. Grams @ 278
Based on X-Axis Probes
1,289. Grams @ 270
826. Grams @ 260
Vector Average Correction
1,175. Grams @ 259
728. Grams @ 268
535. Grams @ 270
369. Grams @ 270
Final Weights Installed
584
Chapter-11
the LP rotor around 270 would have a detrimental influence upon the HP-IP
rotor, and the vibration at #1 bearing. General experience with this class of
machinery has shown that weight corrections in the order of 40% to 50% of the
calculated values are appropriately conservative. Hence, the final weights shown
in Table 11-7 were physically installed in this LP turbine rotor. It should be mentioned that this weight correction was essentially a static shot that was consistent with the weights previously installed during the low speed balance of this
rotor. This type of correction was not unusual, since the low speed balance often
underestimates the required weight due to lower sensitivity of the rotor at 250 to
300 RPM versus the actual machine at 3,600 RPM. Overall, the correction
weight angles were reasonable, and the weight magnitudes made good sense.
On the balance plane #2 adjacent to the A coupling and the #2 bearing, the
installed four calibration weights of 160 grams were supplemented by nine additional sliding weights with a mass of 375 grams. These thirteen weights with a
combined mass of 535 Grams were positioned around 270. On balance correction plane #3 (axial face of the aft LP wheel) adjacent to the #3 bearing, the
installed six calibration weights of 310 grams were supplemented by one additional sliding weight with a mass of 59 grams. The seven weights with a combined mass of 369 Grams were positioned around 270 as shown in Fig. 11-30.
Additional calculations were performed to predict the vibration response at
LP turbine bearings #2 and #3. The results of these calculations are discussed in
this chapter under the Response Prediction heading. In addition, the two plane
balance calculations were extended to include results between the #1 and #2
bearings, plus the #1 and #3 bearings. Although these computations may not be
completely linear, they do provide some indication of the potential vibration
severity on the HP-IP rotor due to weight corrections on the LP rotor. These calculations reinforced the fact that aggressive weight additions on the LP turbine
could adversely influence the vibratory behavior on the HP rotor #1 bearing.
After the final weight corrections were executed, the machinery train was
restarted. Examination of this startup data revealed acceptable vibration amplitudes at all measurement locations. The train was allowed to heat soak, and load
was gradually applied. The final runout compensated shaft orbits at each bearing are shown in Fig. 11-31 at a constant load of 115 megawatts.
This data is directly comparable to the initial orbits shown in Fig. 11-28.
Fig. 1131 Final Runout Compensated Orbits At Each Train Bearing At 115 Megawatts
Two Plane Balance
585
Within the final orbits presented in Fig. 11-31, it is clear that the HP turbine
bearing #1 was not adversely influenced by the balance weights. The vibration at
the IP bearing #2 was substantially reduced, and a significant improvement was
noted at the LP turbine bearing #3. There was also a slight reduction in vertical
response on the generator bearings. It is peculiar to note that a reverse precession now appears on bearings #3, #4, and #5. Unfortunately, the analysis and
explanation of this behavior does fall under the category of a whole different
story.
The initial versus the final 1X vibration vectors are presented in a tabular
format in Table 11-8 for bearings #2 and #3. One should never lose track of the
Table 118 Comparison Of Initial Versus Final 1X Vibration Vectors At Full Load
Initial Condition
112 Megawatts
Final Condition
115 Megawatts
IP Turbine Brg #2 Shaft #2Y
1.45 Mils,p-p @ 198
0.49 Mils,p-p @ 228
IP Turbine Brg #2 Shaft #2X
2.11 Mils,p-p @ 64
0.44 Mils,p-p @ 79
LP Turbine Brg #3 Shaft #3Y
4.58 Mils,p-p @ 292
3.06 Mils,p-p @ 316
LP Turbine Brg #3 Shaft #3X
2.55 Mils,p-p @ 138
1.05 Mils,p-p @ 11
LP Turbine Brg #3 Casing Vert.
3.58 Mils,p-p @ 224
0.73 Mils,p-p @ 231
LP Turbine Brg #3 Casing Axial
9.18 Mils,p-p @ 226
0.96 Mils,p-p @ 193
Vibration Transducer
original project objective. The main objective of this engineering project was to
reduce the axial casing vibration on the LP turbine to acceptable levels. At the
conclusion of field balancing activities, the vertical and axial casing vibration
amplitudes at the #3 bearing were significantly attenuated. Further reductions
occurred as the unit was loaded and allowed to heat soak. In Table 11-8, it is
clear that following the various corrections and field trim balancing of the LP
rotor, the 1X casing axial amplitudes were reduced to nominally 1.0 Mil,p-p. In
addition, transmitted vibration to the structure, the turbine deck, and the control room were all greatly reduced. It is speculated that the source of the axial
casing vibration on the LP turbine was due to an axial wobble of the large last
stage turbine wheels. This wobble was logically induced by excessive radial rotor
deflections that are dependent on the rotor mode shape at operating speed.
Although this hypothesis is difficult to prove with the available information, it
certainly does satisfy the observed machinery behavior.
From another perspective, the final shaft and casing vibration amplitudes
were very acceptable across the entire load envelope. Transient startup behavior
was quite satisfactory, and the turbine generator set remained in constant operation. Other problems, such as excessive hydrogen leakage on the generator seals
and stability problems with the electronic speed control system, still had to be
resolved, but the two coupled steam turbines were in excellent condition.
586
Chapter-11
WEIGHT SEQUENCE VARIATION

The previous discussion addressed a two plane balance based upon a rigid
set of rules regarding the installation and removal of calibration weights. The
developed equation array requires the installation of a calibration weight W1 at
the first balance plane, a data acquisition run on the machine, followed by a
shutdown, and removal of the W1 calibration weight. Sequentially, the second
calibration weight W2 is added to the second balance plane, and another data run
is made on the machine. Based on the initial vibration response data, plus the
two calibration runs, equations (11-20) and (11-21) may be used to compute the
balance corrections.
In actual field balancing situations, these rules are often modified due to
other influences and considerations (e.g. drilled balance holes). Typically, there
are two realistic variations in the sequence of the calibration weights installed
for the two calibration runs. First, consider the case where a calibration weight
W1 is mounted at balance plane 1 for the first run. This weight is allowed to
remain in position, and another weight W2 is attached at balance plane 2 for the
second calibration run. In this condition, the vibration response during the second calibration run is dependent on the combined effect of both weights W1 and
W2. Under this condition, the four balance sensitivity vectors must be computed
in accordance with the differential vectors for each specific weight change. Thus,
the balance sensitivity vectors for this condition are properly defined by the following expressions (11-22) through (11-25):
S 11
W1
= ----------------------
B 11 A 1
(11-22)
S 21
W1
= ----------------------
B 21 A 2
(11-23)
S 12
W2
= -------------------------
B 12 B 11
(11-24)
S 22
W2
= -------------------------
B 22 B 21
(11-25)
These resultant balance sensitivity vectors are then used to compute the
Weight Sequence Variation
587
mass unbalance in each plane as previously specified in equations (11-20) and

(11-21). At this point the two calibration weights may be removed, and the final
balance correction weights may be installed.
The other viable option is to leave the calibration weights in place (particularly with drilled holes), and perform a final correction that is vectorially equal
to the calculated mass unbalance minus the calibration weight (added or
removed) at each balance plane. The same final weight correction can also be
achieved by running a trim balance calculation using the last set of vibration
response vectors as the residual unbalance vectors. The results of this computational approach will provide the final incremental weight change vectors. Either
approach is acceptable, and the actual field situation dictates the appropriate
procedure.
The second potential variation on the basic procedure case consists of a calibration weight W2 mounted at balance plane 2 for the first run. This weight is
allowed to remain in place, and a new weight W1 is attached at balance plane 1
for the second calibration run. In this condition, the vibration response is again
dependent on the combined influence of both weights W1 and W2. The four balance sensitivity vectors for this condition are calculated in accordance with the
following four equations (11-26) through (11-29):
S 11
W1
= -------------------------
B 11 B 12
(11-26)
S 21
W1
= -------------------------
B 21 B 22
(11-27)
W2
S 12 = ----------------------
B 12 A 1
(11-28)
W2
= ----------------------
B 22 A 2
(11-29)
S 22
Following the computation of these four balance sensitivity vectors, the balancing equations (11-20) and (11-21) are used to solve for the effective rotor
unbalance at each of the two correction planes. The comments previously provided on the final addition (or removal) of weights to correct for the unbalance
vectors are also appropriate to this calibration weight sequence.
588
Chapter-11
Case History 39: Three Bearing Turbine Generator at 3,600 RPM

This case history addresses a 3,600 RPM turbine generator set. This is a 7.5
megawatt (10,000 HP) unit that was commissioned in the 1950s. Two identical
trains were installed, and both units display similar characteristics, and virtually identical mechanical difficulties. Although some unique features are associated with each train, they both share a common historical trait of synchronous
vibration problems. In recent years, the majority of the problems have been
traced to generator unbalance, plus high eccentricity of the solid coupling
between the turbine and generator. Although successful field balance corrections
have been performed on both units, the logic behind some of the balance corrections had not been completely understood.
The machinery arrangement sketch presented in Fig. 11-32 describes the
general arrangement of one unit. It is noted that this is a three bearing train,
and the generator inboard must be supported by the turbine exhaust bearing.
The 14 stage turbine driver is an extraction unit, with a surface condenser
directly below the exhaust flange. The synchronous generator has collector rings
mounted at the outboard end and a separate exciter. For many years this train
included a direct shaft-driven exciter. However, this was replaced with a separate solid state unit to improve train reliability. The entire machinery unit is
mounted on a mezzanine deck, and the bearing supports are reasonably compliant. This is particularly true at the generator outboard bearing, and various
hydraulic and screw jacks have been applied over the years to increase the vertical support stiffness at this location.
The machine is equipped with X-Y proximity probes at each journal bearInlet Ext
Exhaust
Normal
Thrust
Solid Coupling
Collector Rings
K
Extraction
Steam
Turbine
1ADa
1ADb
1VD
1HD
2VD
45 45
CW
Generator
13,800 Volts
3 - 60 Hz
2HD
3VD
45 45
2VA
2HA
45 45
3HD
45 45
3VA
3HA
45 45

Viewed from the
Turbine Governor
Fig. 1132 Machinery And Vibration Transducer Arrangement On Turbine Generator Set
589
ing. As noted in Fig. 11-32, the probes are oriented at 45 from the true vertical
centerline. Two axial probes are mounted at the governor end of the turbine, and
a once-per-rev Keyphasor is installed at the generator outboard. To provide
vibratory information on the flexible bearing housings, X-Y casing accelerometers are mounted on the #2 and #3 bearing caps. All of the transducers are connected to a permanent monitoring system. The proximity probes read out in
displacement units of Mils,p-p, and the accelerometers are integrated to IPS,o-p.
The data presented in this case history will include another level of integration
to convert the casing velocity to Mils,p-p of casing displacement.
Rotor configuration for this machinery train is illustrated in Fig. 11-33.
Overall length of the coupled rotors is 25 feet and 5 inches (305 inches). The turbine rotor weighs approximately 9,380 pounds, and the generator rotating
assembly weighs 12,080 pounds for a combined rotor weight of 21,460 pounds.
The turbine rotor contains three internal radial balance rings, and the generator
was constructed with two axial balance rings. The physical locations of the generator balance planes are noted on the rotor drawing.
Fig. 1133 Rotor And Proximity Probe Configuration On Turbine Generator Set
The calculated undamped first critical speed for this rotor system is 1,210
RPM. This is a translatory mode that exhibits the largest amplitudes in the
vicinity of the coupling. Generally, the first critical is not a problem for these T/G
Sets. The major problem usually occurs at the second or third critical speeds. The
calculated undamped mode shapes for both of these resonances are shown in Fig.
11-34. Vibration measurement planes at the #2 and the #3 bearings at both criticals are separated by a nodal point. Hence, a phase reversal would be expected
between bearings as the machine passes through these critical speeds. Naturally,
the system mass unbalance distribution will dictate the severity of the response
through each mode, and the damping would control the phase shift. These characteristics are directly attributable to the mode shapes through the second and
third criticals, and this complexity has often confused the field balancing logic.
Due to these complex modes, and the history of balance problems, the operating company elected to send out both the turbine and the generator rotors for
high speed balancing whenever a major overhaul was performed. T/G Set #1 was
subjected to a major overhaul, and the high speed shop balance of both rotors
proved to be effective. Following the correction of several control problems, this
unit was restarted and placed on line with minimal problems, and low vibration
levels. A year later, a similar overhaul was performed on #2 unit. The turbine
590
Chapter-11
Fig. 1134 Calculated

Undamped Shaft Mode
Shapes For Turbine Generator Set - Second and
Third Critical Speeds
rotor was refurbished with some rows of new buckets, and it was successfully
shop balanced at 3,600 RPM. Similarly, the generator rotor was subjected to various repairs at another shop, and it was also shop balanced at the full operating
speed of 3,600 RPM. This work was performed with a temporary shaft stub end
that was bolted to the coupling half, and used as the inboard journal.
Both rotors were reinstalled in their respective cases. The machine was
realigned, and the unit prepared for operation. The initial startup revealed high
vibration amplitudes at the first critical speed of 1,250 RPM. Shaft and casing
vibration levels were unacceptable as the unit entered the second critical, and
this run was terminated at 1,470 RPM with shaft vibration amplitudes
approaching 25 Mils,p-p at the generator outboard #3 bearing.
Following shutdown, it was determined that the governor speed ramp was
set at a low rate. This contributed to the high vibration due to extended operating time at the criticals. The ramp rate was increased, and during the next run
the full speed of 3,600 RPM was achieved. During this startup, the highest vibration amplitude of 15.6 Mils,p-p occurred at the #3 bearing at the second critical
speed of 1,650 RPM. At full speed, the shaft vibration data at the #2 and #3 bearings are shown in Figs. 11-35 and 11-36. Based upon the high vibration amplitudes encountered at the #3 bearing, a physical bearing inspection at this
location was considered necessary. The disassembled #3 bearing revealed babbitt
damage, and expanded clearances. It is reasonable to conclude that the majority
of this damage occurred during the initial aborted run. This damaged bearing
was replaced, and it was checked for proper clearances.
A review of the initial vibration data revealed acceptable amplitudes at the
governor end of the turbine. However, the response at the turbine exhaust #2

Initial Shaft Vibration At #2 Bearing
591

Initial Shaft Vibration At #3 Bearing
bearing, and the generator outboard #3 bearing were unacceptable. In all cases,
the vibratory characteristics were dominated by rotational speed 1X motion, and
the largest amplitudes appeared on the generator. Based on the available data,
and previous experience with this machinery, a balance correction for the generator outboard was computed. At this time a 216 gram, heavy metal weight was
installed at 96 at the generator outboard. The physical location of this weight,
plus all other balance weights, are shown on the mechanical documentation diagrams presented in Figs. 11-37 and 11-38.
The T/G set was restarted, and full speed data with the 216 gram weight at
the generator outboard revealed a significant improvement in vibration amplitudes. Next, a 265 gram, heavy metal weight was installed at 63 at the generator coupling end. The unit was restarted, and full speed data revealed a further
reduction in shaft and casing vibration amplitudes. Based upon the installation
of these two weights, a complete set of two plane balance calculations were performed. Table 11-9 summarizes the computations based upon shaft proximity
probes. A two plane balance was performed between the two vertical probes, and
a separate two plane balance was performed between the horizontal probes
mounted on each end of the generator. In addition, the casing accelerometers
were integrated to displacement, and duplicate calculations based upon the casing motion were performed. The summarized results of these calculations are
shown in Table 11-10 for the vertical and horizontal accelerometers. The casing
data uses a zero slow roll vector, and all vibratory and weight parameters are
consistent with the shaft calculations presented in Table 11-9.
Due to the positive improvement obtained from the original two weights,
there was a reluctance to remove these weights. Hence, these weights remained
592
Chapter-11
Table 119 Generator Two Plane Balance Calculations Based On Shaft Proximity Probes
Vertical Shaft Probes

Probe #2VD
Probe #3VD
Mag.
@ Angle
Mag.
Horizontal Shaft Probes

Probe #2HD
Probe #3HD
@ Angle
Input
Mag.
@ Angle
Mag.
@ Angle
Input
0.95
4.55
1.50
5.02
265
Calculated
A1c=
4.97
B11c=
2.25
B12c=
5.22
C11=
3.89
C12=
1.26
W1e=
265
S11= 68.12
S12= 171.4
Output
U1=
334
WA1=
334
E1=
A1=
B11=
B12=
W1=
@
@
@
@
@
212
323
344
309
63
E2=
A2=
B21=
B22=
W2=
1.55
8.70
2.12
7.36
216
@
@
@
@
@
68
103
106
111
96
@
@
@
@
333
2
319
116
248
63
307
208
A2c=
B21c=
B22c=
C21=
C22=
W2e=
S21=
S22=
7.48
1.31
6.32
5.25
1.74
216
50.48
124.1
@
@
@
@
110
153
121
293
247
96
130
209
259
79
U2=
WA2=
310
310
@
@
@
@
@
@
@
@
@
@
259
79
0.90
2.52
0.77
3.07
265
Calculated
A1c=
2.90
B11c=
1.48
B12c=
3.30
C11=
2.42
C12=
0.70
W1e=
265
S11= 109.5
S12= 308.6
Output
U1=
307
WA1=
307
E1=
A1=
B11=
B12=
W1=
@
@
@
@
@
316
62
81
53
63
E2=
A2=
B21=
B22=
W2=
1.60
5.90
2.08
4.88
216
@
@
@
@
@
152
196
167
183
96
@
@
@
@
79
111
69
224
19
63
199
77
A2c=
B21c=
B22c=
C21=
C22=
W2e=
S21=
S22=
4.88
0.68
3.60
2.94
1.59
216
90.14
135.9
@
@
@
@
209
205
196
14
60
96
49
36
264
84
U2=
WA2=
210
210
@
@
@
@
@
@
@
@
@
@
231
51
in place, and the balance sensitivity vectors for each of the four data sets were
computed in accordance with equations (11-26) through (11-29). The two plane
balance calculations were then performed with equations (11-20) and (11-21). In
both cases, the calculations were executed on a Microsoft Excel spreadsheet.
For purposes of explanation, the slow roll vectors in Table 11-9 are identiTable 1110 Generator Two Plane Balance Calculations Based On Casing Accelerometers
Vertical Casing Probes

Probe #2VA>D
Probe #3VA>D
Mag.
@ Angle
Mag.
@ Angle
Input
0.00
2.33
1.04
2.55
265
Calculated
A1c=
2.33
B11c=
1.04
B12c=
2.55
C11=
1.81
C12=
0.56
W1e=
265
S11= 146.4
S12= 385.7
Output
U1=
364
WA1=
364
E1=
A1=
B11=
B12=
W1=
Horizontal Casing Probes

Probe #2HA>D
Probe #3HA>D
Mag.
@ Angle
Mag.
@ Angle
Input
0
322
346
310
63
E2=
A2=
B21=
B22=
W2=
0.00
6.52
0.84
2.79
216
@
@
@
A2c=
B21c=
B22c=
C21=
C22=
W2e=
S21=
S22=
6.52
0.84
2.79
2.11
3.89
216
125.6
55.5
@
@
322
346
310
110
249
63
313
207
107
123
92
260
298
96
163
158
@
@
263
83
U2=
WA2=
203
203
@
@
271
91
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
0
107
123
92
96
0.00
1.07
0.23
1.14
265
Calculated
A1c=
1.07
B11c=
0.23
B12c=
1.14
C11=
0.93
C12=
0.15
W1e=
265
S11= 285.0
S12= 1,440.
Output
U1=
328
WA1=
328
E1=
A1=
B11=
B12=
W1=
0
327
342
320
63
E2=
A2=
B21=
B22=
W2=
0.00
5.93
0.82
2.63
216
@
@
@
A2c=
B21c=
B22c=
C21=
C22=
W2e=
S21=
S22=
5.93
0.82
2.63
1.86
3.32
216
142.5
65.1
@
@
327
342
320
135
261
63
288
195
115
137
120
293
291
96
130
165
@
@
248
68
U2=
WA2=
236
236
@
@
278
98
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
0
115
137
120
96
593
fied by E1 and E2. The initial vibration at each plane is specified by A1 and A2.
The measured vibration response with calibration weights installed are identified by the four vectors B11, B12, B21, and B22. All eight of these vector quantities
carry the units of Mils,p-p at Degrees. The two calibration weights W1 and W2
have engineering units of Grams at Degrees. These vector quantities provide the
input portion of the spreadsheet. In the calculated section of the spreadsheet the
slow roll runout is subtracted from each of the balancing speed vibration vectors
to provide visibility of the actual shaft motion. Next, a series of four new vectors
identified as C11, C12, C21, and C22 are presented. These are the differential
vibration vectors that represent the change in vibration due to the installation of
the calibration weights. If these differential vectors are very small, the associated balance sensitivity vector will be quite large. This is indicative of a condition where the calibration weight was undersized, and it was insufficient to
produce a measurable response. The other possibility is that the installed calibration weight at a specific balance plane has minimal effect upon a particular
measurement plane. In either case, the diagnostician must have visibility of the
magnitude of these differential C vectors.
The four balance sensitivity vectors S11, S12, S21, and S22 are listed at the
bottom of calculated data sections on Table 11-9. These vectors have engineering
units of Grams per Mil,p-p at Degrees. The smaller the magnitude of this number,
the more sensitive the location will be to weight addition. Conversely, a large
sensitivity vector magnitude reveals an insensitive combination as discussed in
the previous paragraph. Finally, the output section of the spreadsheet summarized the calculated unbalance at each plane by U1 and U2. For situations where
weight will be added to the machine, the unbalance vectors are adjusted by 180,
and WA1 and WA2 are used to identify the two weight add vectors.
At this point it is meaningful to summarize the results of the four sets of
balancing calculations. The magnitude and location of the calculated mass
unbalance was extracted from Table 11-9 for the proximity probes, and the companion Table 11-10 for the casing accelerometers. These calculated mass unbalance results are presented in the following Table 11-11:
A good comparison exists between the vertical versus the horizontal calculations. In addition, the shaft and casing computations are in general agreement.
From this summation, it may be concluded that the coupling end unbalance is
Table 1111 Summary Of Calculated Generator Unbalance Vectors
Calculated Unbalance At
Coupling End
Calculated Unbalance At
Outboard End
Shaft Vertical
334 Grams @ 259
310 Grams @ 259
Shaft Horizontal
307 Grams @ 264
210 Grams @ 231
Casing Vertical
364 Grams @ 263
203 Grams @ 271
Casing Horizontal
328 Grams @ 248
236 Grams @ 278
Transducers
594
Chapter-11
approximately 340 grams at an angle of nominally 260. Similarly, the effective

mass unbalance at the outboard plane will have a magnitude of approximately
240 grams at a nominal angle of 260.
The unbalance should be correctable by a weight addition of 340 grams at
80 at the coupling plane, plus 240 grams at 80 at the outboard end. However,
the inboard coupling end balance plane already contains the calibration weight
of 265 grams at 63. The outboard plane has a 216 gram calibration weight at an
angle of 96. Rather than remove these existing calibration weights, and install
new correction weights, it makes sense to supplement the existing weights with
smaller additional weights to equal the desired correction vectors. To be perfectly
clear on this point, the weight vectors are summarized in Table 11-12. The first
row in this summary specifies the desired correction weight vectors. The second
row of this table presents the magnitude and angular location of the existing calibration weights in both planes. The vector difference between the desired and
existing weights is listed as the additional weight vector in the bottom row of
Table 11-12. From this vector subtraction, an additional weight of 116 grams at
122 should be installed at the coupling end plane, and 68 grams should be
mounted at 19 on the generator outboard correction plane.
Table 1112 Balance Weight Corrections Based Upon The Vector Difference Between
Desired Correction Weights And Existing Calibration Weights Installed In The Generator
Weight
Coupling Correction
Outboard Correction
Desired Weight Correction
340 Grams @ 80
240 Grams @ 80
Existing Calibration Weights
265 Grams @ 63
216 Grams @ 96
Additional Weight (Vector Difference)
116 Grams @ 122
68 Grams @ 19
This weight add determination may also be performed by running another

set of balancing calculations. In this approach, the last set of vibration response
vectors (with both calibration weights installed) are combined with the previously calculated set of four balance sensitivity vectors. The resultant balance calculation are generally referred to as a trim balance. That is, the balance
calculations are applied to the last set of vibration response measurements, and
a final trim balance for the unit is selected. In some instances, the computed trim
corrections are insignificant, and the balancing exercise is thereby completed. In
the case of this generator, the resultant trim corrections were meaningful. For
comparative purposes, Table 11-13 summarizes the vector weight difference from
Table 11-12, combined with the average correction calculated from the final set
of vibration response vectors. These two sets of potential weight corrections are
listed with the actual set of final balance weights that were physically installed
at both ends of the this generator.
As previously stated in this chapter, field balancing is often composed of a
series of compromise decisions. As shown on Figs. 11-37 and 11-38, the generator
only contains eleven axial balance holes at each end of the rotor. Each threaded
595
Table 1113 Summary Of Potential And Actual Balance Weight Corrections

Weight
Coupling Correction
Outboard Correction
Additional Weight (Vector Difference)
116 Grams @ 122
68 Grams @ 19
Additional Weight (Trim Calculation)
108 Grams @ 102
90 Grams @ 15
94 Grams @ 96
85 Grams @ 31
Actual Weight Installed
hole is one inch in diameter, and one and a quarter inches deep. Hence, there are
definite physical limits to the location and the maximum size of the balance
weights. For this type of weight correction, it makes sense to have a stick of allthread available to use for the fabrication of balance weights. Once weight magnitudes are selected, the equivalent length may be determined, and that length
cut off of the section of all-thread. The balance weight should have the threads
dressed, and a screw driver slot milled at one end. Before installation, the balance weight should be accurately weighed to verify that the final weight is of the
correct mass.
With respect to the current generator balance problem, it was clear from
the first two rows of Table 11-13 that the coupling end plane requires another
weight of approximately 110 grams at an average angle of 110. The outboard
balance plane needs about 80 grams in the vicinity of 20. These potential vector
weight additions were compared against the available balance weights, plus the
empty balance holes. Following this reconciliation, a 94 gram weight was
installed at 96 on the coupling end, and an 85 gram weight was installed at 31
on outboard balance plane. These corrections were considered to be acceptably
close to the calculated weight requirements, and they were physically achievable
on the actual machine.
The validity of these final weight corrections was tested during the next
startup. The synchronous 1X vibration amplitudes were again reduced during
the startup through the three critical speeds, plus the normal operating speed
condition at 3,600 RPM. In the initial startup, transient vibration amplitudes
exceeded 20.0 Mils,p-p. After the final balance correction, the maximum vibration
was 3.6 Mils,p-p through the sensitive second critical speed region. At full operating speed of 3,600 RPM the final vibration amplitudes were below 1.0 Mil,p-p at
all measurement locations. This included the runout compensated shaft vibration amplitudes, plus casing displacement. Furthermore, this low vibratory
behavior was retained as the unit reached full load and heat soak.
The last issue to be addressed is the final documentation of the balancing
project. Certainly the detailed spreadsheets of the balancing calculations are
important. Equally significant are the mechanical descriptions of the weights
installed (or removed), plus the details of the balance correction planes. As an
example of this type of information, Figs. 11-37 and 11-38 were produced.
In retrospect it should be noted that the two plane balance correction on
this generator ended up as a simple static shot. This is self-evident by comparing
the effective weight add vectors in Figs. 11-37 and 11-38. The effective weight
596
Chapter-11
Vertical Centerline
K Probe
le
Ho
#3
26
35
Horizontal
Centerline
Hole #2 @ 227
2
16
@
11
10.5"
Radius
#
ole
#9
Hole #10 @
96
129
le
19
Ho
Ho
63
e#
1
@ 31
#8 @
Calibration Weight
of 265 Grams
92
#6
Hole
#4
5 @ 32
le
Ho
Hole
#7
Horizontal
Probes
2HD & 2HA
Hole
Hole #
Vertical
Probes
2VD & 2VA
Empty Holes
1" Diameter
8 Thds/Inch
1.25" Deep
Trim Weight
of 94 Grams
C W R ot at io n
View from the Turbine Governor End

Correction Weights Vectorially = 348 Grams at 71
Fig. 1137 Balance Correction Plane At Inboard Coupling End Of Generator
Vertical Centerline
K Probe
#3
26
Horizontal
Centerline
Hole #2 @ 227
Ho
@
11
le #
2
16
129
Hole #10 @
96
19
4
Ho
le
#1
@
#9
le
63
Ho
le
Ho
@ 31
#8 @
292
8
35
Calibration Weight
of 216 Grams
4@
Hole
325
#6
Hole
#7
le #
Trim Weight
of 85 Grams
Horizontal
Probes
3HD & 3HA
Ho
ol
5@
Hole #
Vertical
Probes
3VD & 3VA
10.5"
Radius
Empty Holes
1" Diameter
8 Thds/Inch
1.25" Deep
C W R ot at io n
View from the Turbine Governor End

Correction Weights Vectorially = 263 Grams at 79
Fig. 1138 Balance Correction Plane At Outboard End Of Generator
597
angle at the coupling end was 71, and the outboard correction occurred at 79.
This static correction might seem to be incorrect when viewed against the apparent pivotal response displayed on the orbits in Figs. 11-35 and 11-36. However,
the undamped mode shapes in Fig. 11-34 did reveal the presence of a nodal point
between the vibration probes at the #2 turbine exhaust, and the #3 generator
outboard bearing. Hence, the documented vibratory behavior, the analytical
mode shapes, and the effect of the balance corrections are in unison.
Case History 40: Balancing A 36,330 RPM Pinion Assembly
The following example considers a high speed pinion that is driven by a
cryogenic expander turbine at a normal speed of 36,330 RPM. As shown in Fig.
11-39, the pinion mates with a low speed bull gear that is coupled to an induction
generator that runs at 3,622 RPM. This machinery train extracts energy from
the gas stream, and converts this energy into electrical power. This energy
extraction also lowers the temperature of the gas passing through the expander.
Gas Inlet
2Y
2X
3Y
45 45
Cold Exhaust
3X
4X
45 45
Ratio =
10.0313:1
Expander
Turbine
Pinion
32 Teeth
33 Pounds
1 Stage
1,290 HP @
36,330 RPM
1Y
1X
45 45
CCW
4Aa & b
Khs
Gear
Box
Kls
Induction Generator
Bull Gear
321Teeth
740 Pounds
5Aa & b
Rotation and
Probe
Orientation Viewed
From The Expander
Outboard
4Y
45 45
5Y
Rated 1,500 KW @ 3,622 RPM

4,160 Volts - 3 - 60 Hz
6Y
30
60
5X
30
60 6X
7Y
7X
45 45
8Y
8X
45 45
CW
Fig. 1139 Machinery & Transducer Arrangement On Expander Turbine Driven Generator
From an instrumentation standpoint, the train is well-equipped with X-Y

proximity probes at each journal bearing. In addition, dual thrust probes are
mounted at the outboard end of the bull gear, and the pinion. To provide synchronous tracking, high and low speed Keyphasor probes are installed at the blind
end of both gear elements. This transducer array was easily accommodated at
the outboard end of the bull gear. However, the outboard, or blind end of the pinion, became crowded with five proximity probes installed at the end of a two-inch
diameter shaft. To provide a proper installation of probes, with adequate spacing
between transducers to eliminate any cross talk between probes, the OEM
included a four-inch diameter stub end for the pinion shaft. This instrumentation hub is shown at the right end of the pinion drawing in Fig. 11-40.
This drawing also describes the general pinion configuration. It is noted
that the assembly is a 32 tooth, double helical gear. The overall pinion length is
598
Chapter-11
26.72"
16.50"
4.50"
Double Helical Gear

32 Pinion Teeth
Weight = 33 Pounds
Probes
Fig. 1140 Rotor Configuration On Gear Box High

Speed Pinion
Coupling
Balance
Plane
Probes
Journals
3.10" Long
2.00"
Blind End
Balance
Plane
26.72 inches, and it runs with an up mesh. Total weight of the pinion assembly
was approximately 33 pounds, and the rotor is supported on tilt pad bearings
with 13.40 inches between centers. Physically, the pinion was in good mechanical
condition. The assembly was slow speed balanced on a flexible pedestal balance
machine to residual unbalance levels below 0.1 gram inches. Dial indicator measurements on the pinion revealed a straight assembly with mechanical runout
varying between 0.10 and 0.15 Mils, T.I.R.
Analytical calculations revealed the presence of two critical speeds that the
pinion must transcend during startup. As shown in Fig. 11-41, the first critical
speed occurs at approximately 18,510 RPM. This is an overhung mode that is
primarily associated with the coupling hub. The calculated second critical speed
appears at 23,380 RPM, as shown in Fig. 11-42. This mode is likewise an overhung mode, and it is primarily driven by the four-inch diameter instrument hub
on the outboard end of the pinion. It should be mentioned that the cantilevered
masses at the pinion ends constitute 40% of the total weight. The coupling half
accounts for 23% of the assembly weight, and the instrument hub provides 17%
of the total assembly weight. Hence, the analytical mode shapes describing a pivotal action at each end plane are considered to be realistic. In addition, the
sequence of critical speeds makes sense with respect to the weight distribution,
plus the greater overhang of the coupling hub.
Fig. 1141 Undamped Pinion Mode

Shape At 1st Critical Speed = 18,510 RPM
Fig. 1142 Undamped Pinion Mode

Shape At 2nd Critical Speed =23,380 RPM
599
The radial vibration measurement planes are located sufficiently close to

the planes of maximum motion to provide good representative data of pinion
vibration. Furthermore, the only available balance correction planes are at the
pinion outboard ends. There is no opportunity to add weights to this assembly.
Thus, any pinion balance corrections must be performed by drilling holes into
the coupling hub, or by drilling holes into the blind end instrumentation hub.
Initial condition of the pinion assembly is presented in Fig. 11-43 that displays Bode plots of the runout compensated Y-Axis probes at each end of the
rotor. From this data, it is apparent that the pinion exhibits a flexible shaft
response with a resonance speed range between 16,000 and 28,000 RPM. This
range agrees with the previously calculated undamped critical speeds. At full
operating speed, the amplitude and phase appear within a reasonably stable plateau region. Thus, any changes in running speed will have minimal effect upon
the synchronous 1X vibration vector amplitudes or phase angles. Hence, it was
decided to use 36,000 RPM as a suitable balancing speed for this pinion.
It is important to develop an overview of the balance state. From the previous information, it is apparent that the pinion exhibits a flexible shaft response,
with the majority of the 1X vibration occurring at the blind outboard end. The
pinion is straight, with no evidence of secondary resonances, or gyroscopic
effects. Hence, the characteristics are indicative of mass unbalance that can be
addressed by weight corrections at the modally effective end planes.
The initial runout compensated 1X shaft orbits and time base plots for both
ends of the pinion are shown in Fig. 11-44. The coupling end vectors indicate a
weight add angle of 288 with respect to the vertical probe, and 214 from the
horizontal probe. Adding 90 to the horizontal probe angle yields 304, which is
compatible with the vertical angle of 288. Since the pinion is operating above
the critical speed, the indicated probe angles are representative of the high spot,
which is about 180 from the unbalance. Subtracting 180 from the runout com-
Fig. 1143 Bode Plot Of Pinion Vertical

Probes Before Balancing
Fig. 1144 Orbit and Time Plots Of Pinion

Synchronous Vibration Before Balancing
600
Chapter-11
pensated angles yields an unbalance location (weight removal) of 108 to 124.

The outboard end shows a similar response through the resonance, but a
higher vibration at full operating speed. The vertical outboard probe shows a
phase of 24, with 286 exhibited by the horizontal probe. Again, adding 90 to
the horizontal angle yields 16 high spot indication with respect to the vertical
probe position. Thus, the vertical and horizontal probes are identifying high
spots of 24 and 16, respectively. The unbalance location would be 180 removed
from the high spot, and the weight removal angle would be 204 to 196.
As previously mentioned, the weight changes had to be drilled holes, and
each weight change was a final correction. Although the outboard end displayed
the highest vibration levels, the coupling end was eminently more accessible.
Hence, the first correction was made on the coupling end. In accordance with the
previous discussion of phase angles, the coupling correction was made by drilling
an 1/8 diameter hole at 120. The selected depth was 5/32, which resulted in a
weight removal of 0.22 grams. At the 2.0 inch radius, and the balance speed of
36,000 RPM, this was equivalent to a centrifugal force of 36 pounds. Since the
pinion weighed only 33 pounds, this might seem like an excessive correction.
On slower speed machinery, it is common to use initial weights that produce a centrifugal force equal to 5 to 15% of the rotor weight. However, on high
speed pinions this traditional rule of thumb does not work very well, and extra
weight is normally required. It is presumed that the additional weight requirements are due to the higher stiffness of most pinion assemblies, plus the large
gear contact forces. As always, if the machine has a variable speed driver, the initial weight correction may be more aggressive, and the unit carefully observed
during a controlled startup. Conversely, if a fixed speed driver (e.g. motor) is
employed, then the initial weight should be channeled towards a more conservative direction. Fortunately, this unit was equipped with a variable speed driver,
and the 0.22 gram correction was made at 120 on the coupling.
The results of this initial correction are presented in Table 11-14 that summarizes the runout compensated vibration response vectors, and the computed
single plane mass unbalance calculations from equations (11-4) and (11-6):
Table 1114 Initial Single Plane Balance Correction On Pinion Coupling Hub
Transducer
Shaft Vibration
Calculated Unbalance
Coupling - Probe 3Y
0.16 Mils,p-p @ 270
0.29 Grams @ 114
Coupling - Probe 3X
0.04 Mils,p-p @ 10
0.19 Grams @ 123
Outboard - Probe 4Y
1.43 Mils,p-p @ 6
0.74 Grams @ 42
Outboard - Probe 4X
1.53 Mils,p-p @ 275
1.08 Grams @ 227
This initial hole drilled at the hub was quite effective in reducing the coupling end vibration vectors. In fact, the calculated initial unbalance vectors from
both coupling probes (0.29 Grams @ 114, and 0.19 Grams @ 123) are quite close
601
to the selected correction of 0.22 Grams @ 120. Hence, this initial correction was
adequate to properly balance the coupling end of the pinion.
At the pinion outboard, vibration vectors appeared to be unaffected by the
coupling end weight correction. However, the following Table 11-15 summarizes
the vibration vector changes due to the coupling end weight removal:
Table 1115 Differential Vibration Vectors Due To Correction On Pinion Coupling Hub
Transducer
Vibration Change
Coupling - Probe 3Y
0.45 Mils,p-p @ 115
Coupling - Probe 3X
0.31 Mils,p-p @ 31
Outboard - Probe 4Y
0.47 Mils,p-p @ 280
Outboard - Probe 4X
0.33 Mils,p-p @ 173
From this viewpoint, it is clear that the 0.22 gram correction at the coupling produced a net amplitude response at the outboard bearing that matched
the change at the coupling end. This leads to the conclusion that cross effects
between balance planes cannot be ignored. The next obvious step would be to
make a correction at the outboard instrumentation hub, and then compute a two
plane balance solution.
As per the previous discussion, the correction angle for weight removal of
the initial outboard unbalance should be between 196 and 204. The first correction at the coupling end rolled these initial angles slightly, and the new outboard
high spot was 6 on the 4Y probe, and 275 with respect to the 4X probe (from
Table 11-14). Adding 90 to the 4X probe angle yields a 5 high spot that matches
the 6 high spot indicated by the 4Y probe. Hence, the weight at the outboard
should be installed at 5 to 6, or removed at 185. This angular location was
combined with a 1/8 diameter hole, 5/32 deep, for a total weight removed of
0.22 grams at the outboard instrumentation hub. The results of this second balance correction are summarized in Table 11-16.
Table 1116 Results Of Balance Correction On Pinion Outboard Instrumentation Hub
Transducer
Shaft Vibration
Vibration Change
Coupling - Probe 3Y
0.19 Mils,p-p @ 284
0.05 Mils,p-p @ 342
Coupling - Probe 3X
0.02 Mils,p-p @ 0
0.03 Mils,p-p @ 196
Outboard - Probe 4Y
1.04 Mils,p-p @ 14
0.43 Mils,p-p @ 166
Outboard - Probe 4X
1.15 Mils,p-p @ 283
0.43 Mils,p-p @ 74
Following this run with the correction at the outboard blind end plane, it is
noted that a good response was evident on the outboard, and minimal change
602
Chapter-11
was noted on the coupling end. Now the vectors from both runs are combined
into a two plane balance calculation, using equations (11-22) through (11-25) for
determination of the balance sensitivity vectors. Equations (11-20) and (11-21)
are used for the pinion unbalance calculations. These calculations are based
upon the original vibration vectors, plus the required trim from the last set of
vibration vectors. The spreadsheet describing the complete array of vectors, and
intermediate results for both sets of probes are presented in Table 11-17.
Table 1117 Pinion Two Plane Balance Calculations Based On Shaft Proximity Probes
Vertical Probe Calculations

Probe #3Y
Probe #4Y
Mag.
@ Angle
Mag.
@ Angle
Input
0.27
0.84
0.39
0.43
-0.220
Calculated
A1c=
0.61
B11c=
0.16
B12c=
0.19
C11=
0.45
C12=
0.05
W1e=
0.22
S11= 0.489
S12= 4.400
Output
U1=
0.34
WA1=
0.34
Trim
A1=
0.43
A1c=
0.19
U1=
0.13
WA1=
0.13
E1=
A1=
B11=
B12=
W1=
Horizontal Probe Calculations

Probe #3X
Probe #4X
Mag.
@ Angle
Mag.
@ Angle
Input
324
299
304
308
120
E2=
A2=
B21=
B22=
W2=
0.10
1.37
1.34
0.94
-0.220
@
@
@
A2c=
B21c=
B22c=
C21=
C22=
W2e=
S21=
S22=
1.47
1.43
1.04
0.47
0.43
0.22
0.468
0.512
@
@
288
270
284
115
342
300
185
23
24
6
14
280
166
5
20
199
@
@
127
307
U2=
WA2=
0.80
0.80
@
@
195
15
308
284
138
318
A2=
A2c=
U2=
WA2=
0.94
1.04
0.58
0.58
12
14
199
19
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
208
24
4
12
185
0.24
0.52
0.21
0.23
-0.220
Calculated
A1c=
0.28
B11c=
0.04
B12c=
0.02
C11=
0.32
C12=
0.03
W1e=
0.22
S11= 0.688
S12= 7.333
Output
U1=
0.26
WA1=
0.26
Trim
A1=
0.23
A1c=
0.02
U1=
0.04
WA1=
0.04
E1=
A1=
B11=
B12=
W1=
229
221
236
232
120
E2=
A2=
B21=
B22=
W2=
0.08
1.56
1.46
1.07
-0.220
@
@
@
A2c=
B21c=
B22c=
C21=
C22=
W2e=
S21=
S22=
1.64
1.53
1.15
0.33
0.43
0.22
0.667
0.512
@
@
214
10
360
31
196
300
269
169
286
275
283
173
74
5
127
291
@
@
123
303
U2=
WA2=
0.80
0.80
@
@
204
24
232
360
143
323
A2=
A2c=
U2=
WA2=
1.07
1.15
0.59
0.59
282
283
211
31
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
@
116
286
274
282
185
The trim calculations shown in Table 11-17 are identical to the unbalance
calculations from the standpoint of equation structure. That is, the same equations (11-20) and (11-21) are used for both sets of unbalance calculations. However, the unbalance values U1 and U2 shown in the output section of Table 11-17
are based upon the original vibration vectors (A1 and A2). Under the trim section, the U1 and U2 unbalance vectors are based upon the final vibration amplitudes recorded after drilling the hole at the outboard hub (i.e., B12 and B22).
Thus, the ending point for one problem becomes the starting point for the next
set of balance calculations.
It should also be noted that holes were drilled for the two calibration
weights. This was accommodated by the spreadsheet by entering a negative
weight (i.e., -0.220 grams) into the Input section of the balance spreadsheet. To
maintain consistency throughout the balance calculations, this was converted to
603
a positive weight by adding 180 to the location of the calibration weights, and
listing this as an equivalent weight W1e and W2e in the calculation portion of the
spreadsheet. Thus, -0.22 grams at 120 was converted to +0.22 grams at 300,
and the outboard end -0.22 grams at 185 was converted to +0.22 grams at 5.
For direct comparative purposes, the computed initial unbalance and trim
vectors are extracted from the spreadsheet in Table 11-17, and are summarized
in Table 11-18:
Table 1118 Summary Of Two Plane Calculations For Original Plus Trim Unbalance Weights
Transducer
Original Unbalance
Residual Trim
Vertical - Coupling 3Y
0.34 Grams @ 127
0.13 Grams @ 138
Vertical - Outboard 4Y
0.80 Grams @ 195
0.58 Grams @ 199
Horizontal - Coupling 3X
0.26 Grams @ 123
0.04 Grams @ 143
Horizontal - Outboard 4X
0.80 Grams @ 204
0.59 Grams @ 211
From this data, the vertical Y-Axis, and the horizontal X-Axis probes are
providing consistent results both in terms of the initial pinion unbalance, and
the required trim correction. The coupling end changes are minimal, but then
the coupling end response to the outboard weight change was very small. In the
general assessment, the two plane balance calculations called for an additional
outboard weight removal of 0.58 grams @ 199 based upon the response from the
vertical probes, and a comparable weight removal of 0.59 grams @ 211 based
upon the horizontal probes.
The actual action taken at the outboard balance correction plane was to
drill another 1/8" diameter hole, 3/8" of an inch deep, for a total weight removal
of 0.54 grams. This hole was drilled at an angle of 205. At this time, a correction
was not made at the coupling end of the pinion due to the low residual trim corrections computed for this location. At the balancing speed of 36,000 RPM, the
resultant runout compensated vibration vectors, and the vectorial vibration
change for all four radial probes due to the outboard weight removal are summarized in Table 11-19 as follows:
Table 1119 Results Of Second Balance Correction On Pinion Outboard Hub
Transducer
Shaft Vibration
Vibration Change
Coupling - Probe 3Y
0.37 Mils,p-p @ 301
0.20 Mils,p-p @ 314
Coupling - Probe 3X
0.23 Mils,p-p @ 235
0.24 Mils,p-p @ 232
Outboard - Probe 4Y
0.35 Mils,p-p @ 102
1.08 Mils,p-p @ 174
Outboard - Probe 4X
0.30 Mils,p-p @ 328
0.96 Mils,p-p @ 60
604
Chapter-11
From Table 11-19 it is clear that the runout compensated shaft vibration
vectors are well under control. The weight change at the outboard end also produced a measurable response on the coupling plane. Overall, the outboard plane
was significantly improved, but the coupling end was slightly degraded. Rerunning the two plane calculations resulted in the final set of residual trim corrections, as summarized in Table 11-20:
Table 1120 Calculated Final Two Plane Trim Correction on Pinion
Transducer
Residual Trim
Vertical - Coupling 3Y
0.18 Grams @ 127
Vertical - Outboard 4Y
0.03 Grams @ 173
Horizontal - Coupling 3X
0.17 Grams @ 143
Horizontal - Outboard 4X
0.12 Grams @ 25
These calculations indicate a minor correction at the outboard end, and a

slight correction at the coupling side. Due to the time required to disassemble
the outboard housing, it was decided to make a single correction at the coupling
plane. The current runout compensated vectors called for a high spot from the
coupling Y-Axis probe of 301, and 235 from the X probe. Again, adding 90 to
the X-Axis probe response angle yields 325 which is compatible with the Y-Axis
probe angle. The weight removal location would be 180 away; thus, the two coupling probes show a weight removal angle of 121, and 145, for the Y and X
probes respectively. These angles correspond directly to the results obtained
from the two plane calculations summarized in Table 11-20.
Fig. 1145 Orbit And Time Plots Of Pinion

Synchronous Vibration After Balancing
Fig. 1146 Bode Plot Of Pinion Startup

After Two Plane Balancing
605
Now, recall that the first correction was a hole drilled at 120. Based on the
positive results of this initial correction, it was decided to extend the depth of the
120 hole to 5/16. This extra hole depth resulted in an additional 0.24 grams
removed at the coupling. The results of this last correction were quite positive, as
shown in Fig. 11-45 that displays the orbit, and time base plots obtained at
36,450 RPM. These plots are certainly indicative of very acceptable shaft relative
motion at full speed. For direct comparison purposes, the initial and final full
speed vectors are summarized in Table 11-21:
Table 1121 Comparison Of Initial Versus Final Runout Compensated Shaft Vibration
Transducer
Initial Vibration
Final Vibration
Coupling - Probe 3Y
0.61 Mils,p-p @ 288
0.04 Mils,p-p @ 60
Coupling - Probe 3X
0.28 Mils,p-p @ 214
0.06 Mils,p-p @ 352
Outboard - Probe 4Y
1.47 Mils,p-p @ 24
0.19 Mils,p-p @ 114
Outboard - Probe 4X
1.64 Mils,p-p @ 286
0.19 Mils,p-p @ 338
The balance corrections provided a significant improvement in the rotational speed vectors at full operating speed. Runout compensated vectors after
the four balance runs are quite acceptable, and the balancing activities were concluded. The total field time required for execution of this 36,000 RPM pinion balance was approximately 5 hours from initial data acquisition to final startup.
The improvement in 1X response is also apparent in the Bode plot shown in Fig.
11-46. Note that the response across the entire speed domain is attenuated. In
particular, the peak amplitudes exhibited during transition through the critical
speeds have been decreased from maximum levels of 3.2 Mils,p-p to current peaks
of 0.7 Mils,p-p. Also note that the definition between the first and second critical
speeds has improved. This is probably indicative of reduced modal coupling due
to the improved balance state.
The pinion thrust probes responded to improvements in the pinion balance.
For example, Table 11-22 compares the 1X axial vectors before and after balancing: The thrust probes measure position and vibration of the outboard hub. It is
reasonable for radial motion to manifest as an axial wobble (e.g., as on mode
shape plots). Hence, as the balance state is improved, the axial motion should
decrease. This anticipated behavior is supported by the vectors in Table 11-22.
Table 1122 Comparison Of Initial Versus Final Pinion Axial Shaft Vibration
Transducer
Initial Vibration
Final Vibration
Axial - Probe 4Aa
0.46 Mils,p-p @ 22
0.08 Mils,p-p @ 190
Axial - Probe 4Ab
0.47 Mils,p-p @ 214
0.10 Mils,p-p @ 328
606
Chapter-11
THREE PLANE BALANCE

As previously mentioned, there are unbalance arrays that consist of couple
and static combinations that may require unusual weight corrections. These
types of dynamic unbalance problems may be solved in two planes, or they may
be more conveniently addressed in three correction planes. For instance, consider
Fig. 11-47 describing a three mass rotor kit. This is the same device used for discussion of the two plane unbalance, with the addition of a third disk. Again, X-Y
proximity probes are mounted inboard of the bearings, and their location allows
proper detection of synchronous motion. The two outboard masses are positioned
at the quarter points of the rotor to enhance the pivotal mode response. The new
mass was placed at the midspan to drive the translational (first) mode.
Fig. 1147 Three Mass

Rotor Kit
R
C W ot
Outboard
End
0
45
0315
45
315
0
270
90
45
315
90
270
225
135 Bearing
90
270
225
Journal
180
135
225
135
180
180
Coupling End
The undamped mode shape at the translational first critical speed is

depicted in Fig. 11-48 at the critical speed of 4,200 RPM. From this translational
mode shape it is clear that the midspan mass is the most modally effective location for correcting vibration response amplitudes through the first critical speed.
Certainly the two outboard masses will provide a positive contribution in correcting the translational mode, but the major influence occurs at the midspan.
The calculated undamped mode shape at the pivotal or second critical speed
is displayed in Fig. 11-49. This mode occurs at a nominal speed of 8,000 RPM,
Fig. 1148 Undamped Mode Shape Of

Three Mass Rotor Kit At 4,200 RPM
Fig. 1149 Undamped Mode Shape Of

Three Mass Rotor Kit At 8,000 RPM
Three Plane Balance
607
and is characterized by the midspan nodal point, plus the out-of-phase behavior
across the nodal point. It is clear that the midspan mass will be ineffective for
pivotal mode corrections, since it is very close to a shaft nodal point. Thus, a couple correction between the outboard planes would be required to control the
vibratory characteristics at the pivotal resonance.
The initial behavior of this three mass rotor kit is shown in the Bode plot in
Fig. 11-50. Within this transient data, the previously mentioned first critical at
4,200 RPM, and the pivotal second mode at 8,000 RPM are clearly visible. In
addition, two lower frequency resonances at 2,000 and 3,000 RPM are also
apparent. Further analysis reveals that the 2,000 RPM response is a structural
support resonance for the rotor kit. It was not highly visible in either the one or
two plane examples, due to the lower rotor masses and lower excitation forces in
this speed domain. The polar plots in Figs. 11-51 and 11-52 show this 2,000 RPM
resonance as a small inside loop occurring just above the plot origin (i.e., just
above slow roll speed).
The 3,000 RPM peak is diagnosed as a horizontal translational resonance
that is coupled to the vertical translational resonance at 4,200 RPM. This is the
classic split critical condition where a machine passes through a soft horizontal
resonance, followed by a harder vertical resonance. The polar plots in Figs. 11-51
and 11-52 exhibit identical behavior through this speed range. That is, the rotor
translates through one critical (horizontal), and then it translates through the
vertical critical. Additional verification was provided by the horizontal proximity
probes (not shown). These orthogonal transducers confirmed the split critical
hypothesis by displaying much higher amplitudes at 3,000 versus 4,200 RPM.
The casual observer might identify these resonances as four different shaft
critical speeds. This would result in substantial levels of grief and confusion if
one attempted to modally correct four criticals on a rotor system that really contained only a split first, combined with a pivotal second mode. In all cases, it
should be recognized that a proper understanding of the motion characteristics

Vertical Probes On Three
Mass Rotor Kit Before Balancing
608
Chapter-11
of the mechanical system will allow the most rapid and cost-effective field balance solution.
The initial vertical polar plots that are equivalent to the Bode plot in Fig.
11-50 are presented in Figs. 11-51 and 11-52. The double loop for the split first
critical is clearly evident on both the coupling and the outboard plots. Constructing a directional vector from the plot origin to the end of the first mode response
results in a nominal 93 angle at the coupling end, combined with a 118 angle at
the outboard. These 1st mode midspan correction vector directions are shown on
these plots with the heavy vector lines.

Three Mass Rotor Kit Before Balancing

Three Mass Rotor Kit Before Balancing
The polar plots also display vector directions for weight additions to correct
the pivotal response at 8,000 RPM. Again, these vectors were determined by
evaluating the rotor response specifically associated with the pivotal resonance.
The weight add locations were determined by drawing a vector from the start of
the pivotal resonance loop, to the end of the loop. Specifically, the coupling end
plot displays a desired directional correction in the vicinity of 143 (start to end
of 2nd mode). Simultaneously, the outboard end plot requires a directional correction in the vicinity of 322 (start to end of 2nd mode loop). A 179 difference
exists between these directions which helps substantiate the couple nature of the
pivotal mode and the associated unbalance.
Not surprisingly, these weight add directions are almost identical to the two
plane example previously discussed. In fact, this should be the case, since the
same disks and the same balance weight distributions were used for the two
plane and three plane balancing examples. It is meaningful to note that the addition of the heavy midspan mass had minimal influence upon the pivotal mode.
Thus, the physical measurements confirm and reinforce the theoretical calculations and the validity of the presented mode shapes.
Three Plane Balance
609
As before, it was determined that a couple weight set could be installed to

correct the pivotal second mode. At the inboard coupling end of the rotor, a 0.5
gram weight was placed at 135 (143 desired from the polar plot). At the outboard mass, another 0.5 gram weight was also installed. At this location, the
weight was placed at the 315 hole (322 desired from the polar plot).
The midspan weight should be located between 93 and 118 as previously
discussed. The heavy midspan wheel already contained weight at the 112 hole,
and the 90 hole was empty. Hence, a 1.2 gram weight was installed at the midspan disk at 90. A larger balance weight was used at the midspan, due to the
significantly heavier weight of the new disk installed in the middle of the rotor.
The results of this three plane correction for two modes are displayed in
Figs. 11-53 and 11-54. Note that the initial data required a 20 Mil,p-p plot radius
to contain the high vibration amplitudes. However, following the three plane correction, vibration levels were substantially reduced, and both plots were easily
accommodated by a 2.5 Mil,p-p radius. The residual first mode and the residual
unbalance for the second mode are quite evident on these two polar plots.

Three Mass Rotor Kit After Balancing

Three Mass Rotor Kit After Balancing
Overall, the polar plotting techniques provided the necessary visibility to

allow a three plane balance correction based upon two measurement planes. It is
also quite clear that the final vibration vector from one mode (e.g. the first mode),
becomes the initial runout vector for the next mode (e.g., the second mode).
Hence, even uncoupled modes may be adversely influenced by the residual
unbalance from an earlier mode.
The equations and techniques used for the single plane and the two plane
balance may also be applied to more complicated mechanical systems. For example, a three plane unbalance problem on a linear mechanical system may be
described by the following equations that address the vibration response mea-
610
Chapter-11
sured at three independent locations due to unbalance at three separate planes:
where:
U1 U2 U3
A 1 = --------- + --------- + ---------
S 11 S 12 S 13
(11-30)
U1 U2 U3
A 2 = --------- + --------- + ---------
S 21 S 22 S 23
(11-31)
U1 U2 U3
A 3 = --------- + --------- + ---------
S 31 S 32 S 33
(11-32)

S 32 = Sensitivity Vector at Bearing 3 to Weight at Plane 2(Grams/Mil,p-p at Degrees)
Three Plane Balance
611
Addition of a calibration weight W1 at plane 1 results in the following set of

three new vector response equations:
where:
B 11
U1 + W1 U2 U3
= ---------------------- + --------- + ---------
S 11 S 12 S 13
(11-33)
B 21
U1 + W1 U2 U3
= ---------------------- + --------- + ---------
S 21 S 22 S 23
(11-34)
B 31
U1 + W1 U2 U3
= ---------------------- + --------- + ---------
S 31 S 32 S 33
(11-35)

Removal of calibration weight W1 at balance plane 1; plus the addition of a

calibration weight W2 at plane 2 produces the next set of vector equations:
U1 U2 + W2 U3
B 12 = --------- + ---------------------- + ---------
S 11 S 12 S 13
(11-36)
B 22
U1 U2 + W2 U3
= --------- + ---------------------- + ---------
S 21 S 22 S 23
(11-37)
B 32
U1 U2 + W2 U3
= --------- + ---------------------- + ---------
S 31 S 32 S 33
(11-38)
612
Chapter-11
where:

Removal of calibration weight W2 at balance plane 2, followed by the addition of calibration weight W3 at plane 3 produces the final set of vector equations:
where:
B 13
U1 U2 U3 + W3
= --------- + --------- + ----------------------
S 11 S 12 S 13
(11-39)
B 23
U1 U2 U3 + W3
= --------- + --------- + ----------------------
S 21 S 22 S 23
(11-40)
B 33
U1 U2 U3 + W3
= --------- + --------- + ----------------------
S 31 S 32 S 33
(11-41)

In each of these equations, the first subscript defines the measurement

plane, and the second subscript describes the correction or balance plane. Hence,
this twelve equation array contains fifteen known vector quantities, i.e., the
twelve measured vibration response vectors, A1 through B33, plus the three calibration weight vectors W1 through W3. The calculation procedure initially solves
for the nine unknown balance sensitivity vectors, and finally the three unknown
mass unbalance vectors, U1, U2, and U3.
As derived during the two plane balance, the balance sensitivity vectors are
determined by dividing the appropriate calibration weight vector by the differential vibration amplitude vectors. Rather than deriving each individual expression, it can be shown that the balance sensitivity vectors for a three plane
balance are computed with the following expressions:
Three Plane Balance
613
S 11
W1
= ----------------------
B 11 A 1
(11-42)
S 12
W2
= ----------------------
B 12 A 1
(11-43)
S 13
W3
= ----------------------
B 13 A 1
(11-44)
S 21
W1
= ----------------------
B 21 A 2
(11-45)
S 22
W2
---------------------=
B 22 A 2
(11-46)
S 23
W3
= ----------------------
B 23 A 2
(11-47)
W1
S 31 = ----------------------
B 31 A 3
(11-48)
W2
= ----------------------
B 32 A 3
(11-49)
W3
S 33 = ----------------------
B 33 A 3
(11-50)
S 32
614
Chapter-11
Equations (11-42) through (11-50) may be derived from equation (11-30)

through (11-41), or they may be expanded directly from equation (11-17). In
either case, the differential vibration vectors should always be examined to
insure that a realistic change was imposed by the calibration weight. If the
installed calibration weight does not produce a significant change in the vibration vector(s), the balance calculations may end up based upon noise level data.
Hence, it is mandatory to review the vectors to verify that the calibration
weight(s) have produced an acceptable and measurable change in the vibration
response vector(s). This fundamental concept is just as important in a single
plane balance as it is in a three plane problem.
Once the vibration data has been acquired, and the validity of the vectors
checked the nine sensitivity vectors in equations (11-42) through (11-50) may
be calculated. These vectors are then combined with the initial vibration
response vectors to compute the mass unbalance at each correction plane. One
solution of the three initial expressions (11-30) to (11-32) is presented in the following equations (11-51), (11-52), and (11-53).
1
1
1
1
1
1
A 1 ------------------ ------------------ + A 2 ------------------ ------------------ + A 3 ------------------ ------------------
S S
S S
S S
22 33 S 23 S 32
13 32 S 12 S 33
12 23 S 13 S 22
U 1 = ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- (11-51)
1
1
1
1
1
1
--------------------------- --------------------------- --------------------------- + --------------------------- + --------------------------- ---------------------------
S S S
S S S
S S S
S S S
S S S
S S S
11 22 33
12 21 33
11 23 32
13 21 32
12 23 31
13 22 31
1
1
1
1
1
1
A 1 ------------------ ------------------ + A 2 ------------------ ------------------ + A 3 ------------------ ------------------
S S
S 13 S 31 S 11 S 33
S 11 S 23 S 13 S 21
21 33 S 23 S 31
U 2 = ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- (11-52)
1
1
1
1
1
1
+
+
------------------------------------------------------------------------------------------------------------------------------------------------------
S S S
S S S
S S S
S S S
S S S
S S S
12 21 33
11 22 33
11 23 32
13 21 32
12 23 31
13 31 22
1
1
1
1
1
1
A 1 ------------------ ------------------ + A 2 ------------------ ------------------ + A 3 ------------------ ------------------
S S
S S
S S
31 22 S 21 S 32
11 32 S 12 S 31
12 21 S 11 S 22
U 3 = ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- (11-53)
1
1
1
1
1
1
--------------------------- --------------------------- + --------------------------- --------------------------- --------------------------- + ---------------------------
S S S
12 21 33 S 11 S 22 S 33 S 11 S 23 S 32 S 13 S 21 S 32 S 12 S 23 S 31 S 13 S 22 S 31
These equations may be structured in different ways, and still obtain the
correct results. A manual solution to the above three plane equations is prohibitive due to the complexity of the manipulations, plus the time required for performing these calculations. Once again, these equations have been programmed
on a Microsoft Excel spreadsheet, and a representative example is presented in
Table 11-23. The vibration data set was obtained from a three mass rotor, with
probes mounted close to each disk. In a manner similar to the other balancing
examples in this chapter, this spreadsheet describes the entire array of traditional calculations. This includes runout subtraction, vector changes, sensitivity
vectors, plus the unbalance calculations. In addition, the supplemental calculations of response prediction and trim calculations are included.
Three Plane Balance
615
Table 1123 Three Plane Balance Calculations Based On Shaft Proximity Probes
Vertical Probe Calculations

Probe #1V
Mag.
@ Angle
Probe #2V
Mag.
@ Angle
Probe #3V
Mag.
@ Angle
Input
E1=
A1=
B11=
B12=
B13=
W1=
0.35
2.22
0.56
0.96
3.14
4.780
@
@
@
@
@
@
126
264
113
287
224
270
E2=
A2=
B21=
B22=
B23=
W2=
0.44
2.61
1.05
2.39
2.16
5.860
@
@
@
@
@
@
218
246
224
248
260
248
E3=
A3=
B31=
B32=
B33=
W3=
0.22
2.17
3.14
2.66
0.22
4.780
@
@
@
@
@
@
67
161
209
116
171
180
Calculated
A1c=
B11c=
B12c=
B13c=
C11=
C12=
C13=
W1e=
S11=
S12=
S13=
2.49
0.23
1.30
3.21
2.72
1.39
2.03
4.78
1.757
4.216
2.355
4.01
4.01
4.20
2.39
1.37
1.38
0.09
0.31
0.40
0.06
0.15
0.15
@
@
@
@
@
@
@
@
@
@
269
93
292
230
90
68
179
270
180
180
1
A2c=
B21c=
B22c=
B23c=
C21=
C22=
C23=
W2e=
S21=
S22=
S23=
2.23
0.61
2.02
1.86
1.68
0.24
0.73
5.86
2.845
24.42
6.548
5.77
5.77
5.80
1.48
0.23
0.50
0.12
0.54
0.45
0.07
0.44
0.44
@
@
@
@
@
@
@
@
@
@
251
228
254
269
80
45
21
248
190
203
159
A3c=
B31c=
B32c=
B33c=
C31=
C32=
C33=
W3e=
S31=
S32=
S33=
2.20
3.32
2.52
0.35
2.33
1.90
1.95
4.78
2.052
3.084
2.451
3.33
3.33
3.25
2.05
1.88
1.33
0.31
0.50
0.38
0.18
0.13
0.13
@
@
@
@
@
@
@
@
@
@
167
211
120
209
253
62
340
180
17
186
200
Output
U1=
WA1=
114
294
U2=
WA2=
40
220
U3=
WA3=
45
225
Predict
Z1=
C11z=
C12z=
C13z=
A1c=
A1=
@
@
@
@
@
270
90
45
224
247
140
Z2=
C21z=
C22z=
C23z=
A2c=
A2=
@
@
@
@
@
225
80
22
66
259
226
Z3=
C31z=
C32z=
C33z=
A3c=
A3=
@
@
@
@
@
225
253
39
25
106
90
Trim
A1=
A1c=
U1=
WA1=
@
@
@
122
96
136
316
A2=
A2c=
U2=
WA2=
@
@
@
227
304
290
110
A3=
A3c=
U3=
WA3=
@
@
@
85
107
117
297
It is apparent that these three plane calculations are quite complicated,

and they offer many opportunities for computational errors during the execution
of the vector manipulations. The casual form offered by the single plane balance
calculations is missing, and the equation simplification possible with the two
plane balance is also unattainable. It is also clear that balance calculations for
additional planes (e.g., 4 or more planes) can become extraordinarily complicated. Due to this calculation complexity, it is generally advisable to perform
these calculations with computer assistance. Since circa 1980, several dedicated
programs have been available for these types of balancing calculations. Generally, these multiplane balance programs are configured as matrix solutions.
616
Chapter-11
The matrix solutions can be easily configured as a single plane or a multiple

plane problem. The number of balance planes is typically equal to the number of
independent vibration measurement planes n. Generally, matrix solutions can
accommodate 10 or 20 planes. Although measurement accuracy and true influence of each correction plane upon each measurement plane becomes questionable with more than three planes, the vector calculations can be significantly
improved by using a matrix of the following configuration:
U1
S 11 S 12 S 13 S 1n
A1
U2
S 21 S 22 S 23 S 2n
A2
U3
Un
S 31 S 32 S 33 S 3n
S n1 S n2 S n3 S nn
A3
(11-54)
An
The form of this matrix is the same as the original equation (11-6). Basically, identical statements are made in both expressions. That is, mass unbalance is equal to vibration times the balance sensitivity vectors for each plane.
Again, solution of single plane problems are quite easy, and two or three plane
problems may be conveniently handled on a spreadsheet. When balance calculations are performed for four or more planes, the spreadsheet complexity becomes
significant, and matrix solutions become the only reasonable approach. In any of
these cases, the fundamental concepts of linearity apply, and the machinery
diagnostician must be meticulous in the acquisition of transient and steady state
vibration data. Furthermore, the mode shapes must be understood, and the most
effective balance planes fully utilized.
STATIC COUPLE CORRECTIONS

The previously discussed three plane balance requires three balance correction planes in addition to three separate and independent vibration measurement planes. When dealing with a single machine case, there are very few
machines that will allow meaningful shaft vibration measurements at more than
two locations. Even on machines that may have three or more accessible rotor
balance planes, the techniques described in the last section cannot be applied
due to a lack of suitable vibration measurement planes.
One solution to this dilemma resides in a minor restructure of the logic
associated with the weight corrections. Specifically, two balance planes may be
defined and associated with a weight couple, and a third plane may be used for a
static correction. Other combinations of balancing weights may also be used, but
at this point simply consider a situation of two measurement planes combined
with three correction planes. Presume that the machine contains a static unbal-
Static Couple Corrections
617
ance U, combined with a couple or moment unbalance M. The previous general

two plane equations (11-7) and (11-8) may be modified to represent this particular situation as follows:
U M
A 1 = --------- + -----------
S 1u S 1m
(11-55)
U M
A 2 = --------- + -----------
S 2u S 2m
(11-56)
In these two expressions, the sensitivity vectors are associated with the
static unbalance and the moment unbalance, The calibration run for the static
unbalance U is performed exactly the same as previously described. The calibration run for the moment unbalance M is executed with a coupled pair of weights.
That is, two identical weights are installed on the rotor at two different balance
planes, and the weights are mounted 180 apart. This combined angular and
axial separation of the pair of weights produces a balance couple, or moment
upon the rotor.
Generally, one of the couple weights is used as the reference for the moment
calculations. It is recognized that the other couple weight is the same size (and
probably mounted at the same radius from the rotor centerline). Also, the second
couple weight is always located 180 away from the reference couple weight. The
sketch in Fig. 11-55 is representative of typical weight placement on a rotor. The
static correction at the rotor midspan is located at approximately 330. The couple correction at the coupling end balance plane is shown at nominally 300, with
an angle of 120 for the second couple weight at the outboard end balance plane.
The static-couple balance calculations are performed exactly the same as a
standard two plane balance. The static calibration and correction weights are
applied at the same balance plane. Typically, this would be close to the midspan
Static Unbalance
Correction Weight
Fig. 1155 Combined

Static And Couple Unbalance On A Three Mass
Rotor Kit
R
C W ot
Couple Unbalance
Correction Weight
Outboard
End
0
45
0315
45
315
0
270
90
45
315
90
270
225
135
90
270
225
180
135
225
135
180
180
Couple Unbalance
Correction Weight
Coupling End
618
Chapter-11
of the rotor. The couple or moment calibration and correction weights are applied
at two separate balance planes that are often close to the ends of the rotor. The
final correction moment weights are scaled from the calibration moment weight
in terms of magnitude and angular location.
Using this approach, it is possible to perform a three plane balance using
two measurement planes. It is also meaningful to recognize that the midspan
(static) correction weight is the balance weight that will provide the greatest
influence upon the first (translational) critical speed. Similarly, the outboard
(couple) correction weights are the balance weights that will provide the greatest
influence upon the second (pivotal) critical speed. Hence, as with all balancing
projects, the actual deflected shape of the rotor must be considered in conjunction with any applied weights.
MULTIPLE SPEED CALCULATIONS

In all of the previously discussed cases, the balance sensitivity vectors
describe the rotor response to unbalance at specified correction planes, at one
operating speed. For balancing flexible rotors that operate above several critical
speeds, it may be necessary to identify several distinct balance correction speeds.
For example, consider a rotor that operates above the second critical speed. If difficulty is encountered in balancing this rotating assembly, then balance data
unique to each critical speed may be acquired to allow additional visibility of
modal response. For instance, the following five speeds may be identified:
Slow Roll
Below First Critical Speed
Between First and Second Critical Speed
Above Second Critical
At Full Operating Speed
Obviously the first speed point is used to identify shaft runout characteristics at slow roll speeds (e.g., 500 RPM). As previously discussed, these runout
vectors are vectorially subtracted from each of the balancing speed vibration vectors in order to determine the actual rotor vibration response vectors. In this
example, vibration data would be acquired at each speed, during the initial run,
and during each of the calibration weight runs. This data would then be used to
calculate unbalance correction vectors at each of the four speed points. If this
was a two plane balance, then the resultant information would consist of two
plane weight corrections for each speed point.
If the static-couple technique discussed in the last section is applied, then
the resultant data would consist of corrections for three balance planes. In either
case, the final array of information could be visually examined, and a reasonable
weight set selected. Sometimes one or more points may be violated in order to
achieve acceptable vibration levels at operating speed. In many instances the
modes are uncoupled. Thus, a midspan correction would not influence the second
Response Prediction
619
critical, and a couple correction at the end planes would have minimal effect
upon the first critical.
Another way of evaluating this multiple speed data would be to apply a
Least Squares Balancing calculation in the manner defined by E.J. Gunter and
A. P. Palazzolo9 at the University of Virginia. Their technique provides the lowest overall response across the entire set of variable speed data points, but it
does not necessarily guarantee the lowest possible response at any specific operating speed. Again, human judgment must be applied to select the most reasonable set of balance correction weights, to achieve the lowest possible vibration
amplitudes, in the time available for field balancing of the rotor.
RESPONSE PREDICTION
Evaluation of the suitability of a particular set of balance weights can be
achieved by installing the weights into the machine, and running the machine
up to normal operating speed. If the weight selection is correct, then the synchronous vibration amplitudes will decrease. However, if the weight selection is
incorrect due to any number of reasons, then the results can vary from embarrassing to catastrophic. In addition, operating and fueling costs for large units
generally prohibit unnecessary runs on the machinery.
Hence, it is often desirable to evaluate the influence of balance weights
upon a rotor before installation of the weights and running the machine. Since
the mechanical systems under consideration are presumed to be linear, the most
direct way to perform this evaluation consists of using the calculated balance
sensitivity vectors to estimate the vibration response. For instance, consider a
single plane balance that is performed in accordance with equations (11-4) and
(11-6). Assume that a correction weight Z is to be installed on the rotor. Although
the mass unbalance vector U has been computed, the correction weight and
angle often vary from the ideal or calculated unbalance. The predicted vibration
change due to the installation of this correction weight is calculated by a version
of equation (11-2) as follows:
Z
A z = ---S
where:
Az
= Predicted Vibration Vector Due To Weight Z (Mils,p-p at Degrees)
= Sensitivity Vector To Unbalance (Grams/Mils,p-p at Degrees)
= Weight Vector To Be Installed On Machine (Grams at Degrees)
(11-57)
The predicted change in rotor vibration must now be summed with the ini9 Alan B. Palazzolo and Edgar J. Gunter, Multimass Flexible Rotor Balancing By The Least
Squares Error Method, Rotor Dynamics Course Notes, Part IV, (University of Virginia, Charlottesville, Virginia, 1981), pp 92-123.
620
Chapter-11
tial runout compensated vibration to determine the new anticipated vibration

vector as shown in equation (11-58):
A C- New = A z + A c
where:
Ac
(11-58)
= Previous Runout Compensated Shaft Vibration (Mils,p-p at Degrees)
A C- New = Predicted Runout Compensated Shaft Vibration Vector (Mils,p-p at Degrees)
The predicted runout compensated vibration response vector AC-New may

now be summed with the shaft runout vector E to obtain the estimated direct
synchronous vibration vector as presented in the expression (11-59):
A C = A C- New + E
(11-59)
This procedure for calculating the effect of a balance weight, equation (1157), combining this weight effect with the initial runout compensated shaft
vibration to determine the new shaft vibration vector, equation (11-58), plus the
addition of the slow roll runout, equation (11-59), will prove to be quite accurate
for a linear mechanical system. Experience on a variety of machines has shown
that the vector magnitude is generally within 0.5 Mils,p-p, and the predicted
angle is within 30 of the final measured value. If this type of correlation does
not exist, then consideration should be given to bearing configuration, system
nonlinearities, shaft preloads, thermal effects, fluidic forces, bearing instability,
or other mechanisms that could influence the synchronous response of the rotor.
To demonstrate some typical results of these response prediction calculations, consider the information shown in Table 11-24. This data was extracted
from case history 38, which describes a two plane balance of a large double flow
low pressure turbine rotor. The predicted shaft vibration vectors are shown in
column 2 of this table, and the actual measured values with the balance weights
installed are listed in the center column 3. The differential vibration vector
amplitudes are presented in column 4, and the difference between the measured
and the predicted phase angles are listed in column 5.
Table 1124 Comparison Of Predicted Versus Measured Vibration Response Vectors
Shaft
Vibration
Transducer
Predicted
1X Vibration
Response Vectors
Measured
1X Vibration
Response Vector
Differential
Vector
Amplitude
Differential
Phase
Angle
Brg #2 - #2Y
0.48 Mils,p-p @ 226
0.95 Mils,p-p @ 208
0.51 Mils,p-p
18
Brg #2 - #2X
0.76 Mils,p-p @ 55
0.91 Mils,p-p @ 57
0.15 Mils,p-p
Brg #3 - #3Y
1.50 Mils,p-p @ 225
1.47 Mils,p-p @ 255
0.77 Mils,p-p
30
Brg #3 - #3X
2.16 Mils,p-p @ 121
2.31 Mils,p-p @ 129
0.35 Mils,p-p
Response Prediction
621
Note that the predictions from the two X-Axis probes are much closer than
the results from the Y-Axis transducers. Since this machine rotates counterclockwise in elliptical bearings, the shaft orbits will tend to be elliptical. Often, the
major orbit axis will be closer to the measurement direction of the X probes (45
to the right of vertical). Simultaneously, the minor orbit axis will tend to be
aligned closer to the measurement direction of the Y probes (45 left of vertical).
For this specific mechanical configuration, a balance weight change will generally be more apparent on the X probes rather than the Y probes. This basic characteristic of the orbit shape combined with the location of the X-Y proximity
probes often yield greater accuracy for the X probe balance sensitivity vectors. In
addition, the response measured by the #3Y probe is also influenced by the
reverse precession on the generator rotor that is briefly discussed in case history
38. This additional influence contributes to the inaccuracy of the Y probe measurements for this specific example. Overall, the behavior documented in Table
11-24 is quite typical. Again, it must be recognized that the predict calculations
are fully dependent on the machine geometry, external influences, shaft preloads, and the measurement accuracy of the input data.
This response prediction technique is applicable to any situation where the
balance sensitivity vectors are known, or can be calculated from available data.
This procedure can be extended to two, three, or multiplane balance problems. In
each case, the weight added (or removed) at each plane is vectorially multiplied
by the appropriate sensitivity vectors, and the individual vibration response vectors summed at each measurement plane. An example of a three plane predict is
shown in the spreadsheet in Table 11-23. In this case, three different weights
provided three different vibration response vectors that had to be vectorially
summed at each bearing. This predicted weight response was then summed with
the last shaft vibration vector to determine the predicted runout compensated
shaft vibration. For direct comparison with field instrumentation readings, the
slow roll runout may be vectorially added to each runout compensated result to
determine the final anticipated vibration vector at each measurement plane.
This tool provides a realistic means for evaluating the effect of different
weight combinations on a rotor. For example, the results due to the application of
various correction weights to a series of multiple speed calculations can be quantified and compared. Although these predict calculations may not be totally accurate due to minor nonlinearities in the mechanical system, they will generally
discriminate between an acceptable and an unacceptable weight correction.
Thus, reasonable weight correction decisions may be based upon documented
predict calculations rather than gut feel speculation.
622
Chapter-11
TRIM CALCULATIONS
Following the installation of balance correction weights on a rotor, the
machine may still exhibit some residual vibration. In many cases this may be
due to an inability to match the required balance corrections with field weights.
In other situations, nonlinearities in the mechanical system, the influence of
external forces, or a variation in the balance sensitivity vectors may be responsible. On machines with fluid film bearings, it has been shown that journal centerline position within the bearing changes with the applied loads. Thus, as rotor
balance is improved, the unbalance force at each bearing will be reduced, and
changes in eccentricity position, oil film thickness, and bearing stiffness and
damping coefficients will occur. These changes in bearing characteristics may
appear as alterations of the balance sensitivity vectors. Hence, the balance sensitivity vectors that were established under one balance condition have changed in
magnitude or angle as the balance state changes.
Regardless of the source of the variations, there is often a need to perform a
final trim balance calculation, combined with a final weight adjustment. The execution of this trim balance calculation is virtually identical to the original balance calculations. For example, a two plane balance solution was previously
presented in equations (11-20) and (11-21). In order to perform a trim calculation, the initial vibration vectors (A1c and A2c) are simply replaced by the current
vibration response vectors. The resulting calculation will identify the final trim
balance weights. This approach reuses the same balance sensitivity vectors.
Although some variations to the S vectors have probably occurred, this final trim
calculation will generally bring the machinery vibration amplitudes into a desirable and acceptable range.
Specific examples of trim balance calculations are presented and discussed
as part of the turbine generator case history 39. In addition, the high speed pinion balancing case history 40 also applies this concept of trim computations. This
is a highly useful technique that may be employed in conjunction with the
response prediction calculations, and the multiple speed calculations. On large
machines that incur significant startup and shutdown costs, the machinery diagnostician should use all of the available computational tools at his or her disposal to provide a mechanically suitable and cost-effective field balance in the
minimum number of runs.
The diagnostician should also be fully aware that multiple or repetitive
trim calculations are indicative of abnormal machinery behavior. If the machine
calls for a new trim shot during successive startups, this suggests that some type
of mechanical malfunction may be active. In this situation, the diagnostician
should start looking for other malfunctions, such as a loose impeller, entrained
fluids, loose thrust collar, progressive bearing damage, or a cracked shaft.
Balancing Force Calculations
623
BALANCING FORCE CALCULATIONS

It is always important to consider the physical forces applied to a rotor by
the addition or removal of balance weights. It can be shown that the centrifugal
force (mr2) from a balance weight may be calculated by:
2
RPM
Force pounds = M grams R inches ---------------
4, 000
where:
(11-60)
F = Centrifugal Force (Pounds)

M = Mass of Correction Weight (Grams)
R = Correction Weight Radius (Inches)
RPM = Rotational Speed (Revolutions / Minute)
In this expression, the units for a balance weight correction in grams are
combined with the balance weight radius measured in inches, and the machine
speed in revolutions per minute. The resultant centrifugal force is a radial force
that the balance weight applies to the rotor. This lateral force occurs at the location of the balance weight correction plane.
This same expression for centrifugal force may be expressed in other units.
For instance, if the balance weight is measured in ounces, and the weight radius
remains in inches, the common unbalance units of ounce-inches may be used. To
calculate the centrifugal force F in pounds at any speed, the following equation
(11-61) may be applied:
2
Force pounds
where: Wounces
RPM
= W ounces R inches ---------------
750
(11-61)
= Weight of Correction Weight (Ounces)
For large machines, the balance weight may be expressed in pounds, with
the weight radius remaining in inches. For this condition, the unbalance units of
pound-inches may be used to calculate the centrifugal force F in pounds, at any
speed with equation (11-62):
2
Force pounds
RPM
= W pounds R inches ---------------
188
(11-62)
where:Wpounds = Weight of Correction Weight (Pounds)
The magnitude of this centrifugal force is a significant consideration during

field balancing. Specifically, if the applied force is very small, such as with a
small calibration weight, the machine response will be minimal, and the resultant vibration change vectors and the associated balance sensitivity vectors will
be understated. In this condition, the validity of the balancing calculations will
probably be compromised.
624
Chapter-11
At the other extreme, if the centrifugal force from the calibration or balance
weight is large, then the rotor may be damaged due to excessive radial forces. In
essence, the addition of a small weight will be useless, and a large weight may be
dangerous. Hence, it is necessary to evaluate the appropriateness of the weight
selection. A reasonable method to quantify the severity of the balance or correction weight centrifugal force is to compare this force against the rotor weight.
For example, if a 100 gram weight is installed at a 10 inch radius, on a
machine that runs at 5,000 RPM; the centrifugal force is computed from equation 39 to be 1,562 Pounds. If this balance weight is placed on a 10,000 pound
rotor, the centrifugal force is only 16% of the total rotor weight. However, if this
same weight is mounted on a 500 pound rotor; the resultant centrifugal force is
more than three times the total rotor weight at 312%. This type of centrifugal
force to rotor weight ratio is normally considered to be excessive.
In most field balancing situations, it is customary to install initial calibration weights that produce centrifugal forces in the vicinity of 5% to 15% of the
rotor weight. The conservative 5% value is generally applied to machines that
rapidly accelerate up to speed. Units such as electric motors, or high speed
expanders fall under this category. If the previous equation (11-61) is equated to
5% of the rotor weight, the initial calibration weight (in ounce-inches) may be
computed with the following equation (11-63):
2
5% Cal. Weight Ounce
Inches
168
= Rotor Weight pounds ---------------
RPM
(11-63)
The more aggressive 15% force value is generally reserved for machines
such as steam turbines, or machines driven by turbines, that may be started up
slowly, and easily tripped if the weight is incorrect. Again, if the previous equation (11-61) is equated to 15% of the rotor weight in pounds, the initial calibration weight (in ounce-inches) may be computed with equation (11-64):
2
15% Cal. Weight Ounce
Inches
291
= Rotor Weight pounds ---------------
RPM
(11-64)
Forces from balance correction weights are often in the range of 10% to 50%
of rotor weight. Certainly larger weights may be successfully installed, but good
engineering judgment and caution must be applied. The diagnostician must also
be aware of the fact that large eccentric rotor elements, or shaft bows, will produce substantial radial forces on a rotor assembly. For example, if equation (1162) is modified to consider a rotor element with a weight measured in pounds,
and an eccentricity of the mass center measured in Mils, the resultant force at a
given speed may be determined by the following equation (11-65):
2
RPM
Force pounds = Element pounds Eccentricity mils ---------------
5, 930
(11-65)
Balancing Force Calculations
625
If a single disk or wheel is properly balanced, and that element is installed

eccentrically, equation (11-65) will identify the effective force based upon the element weight, eccentricity, and speed. Similarly, if a balanced rotor has a significant shaft bow, the midspan eccentricity will be indicative of the displacement of
the mass centerline. Thus, the rotor weight, bow radius, and speed may be used
to determine the effective force upon the rotating assembly.
It must always be recognized that field trim balancing of a rotor does have
limitations. If the required balance correction weights are excessive, then there
may be internal damage to the rotor, and disassembly of the machine and inspection of the rotor may be warranted. In other situations, if the balance weights
cannot be placed at modally effective locations, then the calculated correction
weights may be abnormally large, and they may generate excessive radial forces.
To assist in defining reasonable calibration or correction weights placed at modally effective locations, Table 11-25 is offered for consideration:
Table 1125 Typical Balance Sensitivity Magnitudes For Various Rotor Configurations That
Have Been Field Balanced With Modally Effective Weights At Available Balance Planes
Rotor Weight
(Pounds)
Rotor Speed
(RPM)
Sensitivity
(Gram-Inches/Mil)
High Speed Pinion
35
36,000
1.0 to 4.0
Turbo-Expander
900
14,000
20 to 100
Two Pole Motor
3,000
3,600
50 to 250
Two Pole Generator
10,000
3,600
500 to 2,000
Steam Turbine
18,000
3,600
3,500 to 15,000
Single Shaft Gas Turbine
21,000
5,200
150 to 600a
Induced & Forced Draft Fans
6,000
1,200
1,000 to 5,000
Machine Type
aBased
on weights located on external coupling hubs.
Table 11-25 represents a general overview of balance sensitivity magnitudes obtained during field balancing on various types of machines. The installed
balance weights are located at modally effective planes, or at the only available
locations (e.g., generator end planes). This table is useful for a quick sanity check
on the magnitude of a calibration weight. For instance, if a two pole motor is to
be balanced, a reasonable sensitivity value from Table 11-25 would be 120 GramInches per Mil. For a 4.0 inch weight correction radius, and a desired 1.0 Mil,p-p
response, the calibration weight should be approximately 30 grams (=1 x 120 / 4).
In the overview, it must also be acknowledged that field balancing is generally not a cure for a damaged or distressed rotor. Field balancing is ideally suited
to provide a final trim to a rotor running in the actual operating mechanical configuration (case, bearings, seals, and coupling). Field balancing is definitely not a
Magic Elixir to be casually applied whenever vibration levels increase.
626
Chapter-11
BALANCE WEIGHT SPLITTING
Fig. 1156 Balance

Weight Splitting Between
Two Adjacent Holes
Weig
ht
@ Ho
le "B"
Machines such as large industrial fans are typically balanced by welding

weights directly to the wheel. Other rotors, such as high speed pinions, are often
balanced by drilling holes in the vertical faces on each side of the gears. For these
types of balance corrections, the weight additions or removals can be performed
at any required angular location. However, there are many machines, such as
steam turbines and generators, that are equipped with radial or axial balance
rings that contain a series of circumferential drilled and tapped balance holes.
For machines equipped with balance rings, the angular spacing between
holes may vary from 5 on large diameter rotors, to a coarse set of holes at 30 on
small rotors. It is rare for a calculated balance correction weight to be coincident
with one of the existing holes. Hence, it is often necessary to split a required balance correction weight into two or more holes. The fundamental requirement for
this weight split is for the vector summation of all the installed weights to be
equal to the desired balance correction weight vector. For an example of this type
of situation, consider the vector diagram in Fig. 11-56.
t
igh
U"
="
e
dW
e
sir
De
ht @
Weig
Hole
"A"
In this sketch, the desired correction weight is identified by U acting at an

angle . The weight at the first balance hole is specified as A, and the balance
hole angle is . The weight at the second hole is identified as B at an angle of .
All angles are measured against rotation, and all angles are referenced to the
location of the vibration probe. Thus, balance hole A at is the first hole from the
probe, and balance hole B at is the second hole from the probe. The specific
relationship between these three weight vectors can be stated as:
U = A+B
(11-66)
This vector equation may be expanded into a summation of vertical and

horizontal forces as shown in the following two expressions:
U sin = A sin + B sin
U cos = A cos + B cos
This pair of expressions may now be solved for the weights A and B with
the following equations (11-67) and (11-68):
Balance Weight Splitting
627
cos sin sin cos

A = U ----------------------------------------------------------------------
cos sin sin cos
(11-67)
sin cos cos sin

B = U -------------------------------------------------------------------------
cos sin sin cos
(11-68)
As an example of vector hole splitting, assume that a balance calculation

concludes with a required correction of 50 grams at an angle of 50. Further
assume that the available balance holes are located at 30 and 60. The required
weight for the first hole at 30 may be calculated using equation (11-67).
cos 50 sin 60 sin 50 cos 60
0.643 0.866 0.766 0.500
A = 50 ------------------------------------------------------------------------------------------- = 50 ------------------------------------------------------------------------
0.866 0.866 0.500 0.500
0.557 0.383
0.174
A = 50 --------------------------------- = 50 ------------- = 17.4 Grams
0.750
0.250
0.500
Similarly, the weight required at the 60 hole is determined by (11-68):
sin 50 cos 30 cos 50 sin 30
0.766 0.866 0.643 0.500
B = 50 ------------------------------------------------------------------------------------------- = 50 ------------------------------------------------------------------------
0.866 0.866 0.500 0.500
0.342
0.663 0.321
B = 50 --------------------------------- = 50 ------------- = 34.2 Grams
0.750
0.250
0.500
Hence, this required 50 gram balance weight at 50 may be exactly duplicated by installing:
Hole A = 17.4 Grams 30
Hole B = 34.2 Grams 60
The validity of these calculations may be verified by performing a vector
summation of the results. In this case, the sum of the above two vectors are equal
to 50 grams at 50. This concept may be expanded to several holes by simply
extending the equation array. Obviously, the effectiveness of hole splitting diminishes as the angular spread increases. When it is necessary to install a large
amount of weight in a balance ring, there may not be sufficient holes available to
accommodate the required weight. In these situations the diagnostician should
consider the use of heavy weights fabricated from tungsten based alloys. These
weights are nearly twice as heavy as equivalent steel weights, and they may provide the necessary mass when a large correction is required.
628
Chapter-11
WEIGHT REMOVAL
As previously mentioned, some machines require weight removal to implement a proper rotor balance. This weight may be removed by grinding acceptable
surfaces, or by drilling holes into the rotor. Although grinding operations are
common for many shop balancing procedures, they are difficult to quantify during field balancing. Any type of field grinding should always be approached with
great care. The two main areas of concern are excessive weight removal that
exceeds the desired balance change, and the potential for weakening the
machine parts subjected to the grinding wheel. Major mechanical failures have
been traced to excessive grinding on critical machine surfaces.
Fortunately, weight removal by drilling holes is much more controllable and
definable. For instance, Table 11-26 is offered for reference purposes. This tabular summary documents the weight associated with standard diameter holes
(inches) drilled to various depths (inches). These weights are based upon a 118
drill bit tip, and steel density of 0.283 Pounds per Inch3.
Table 1126 Hole Weights In Grams For 118 Drill Bit In Steel
Diam.
1/8"
3/16"
1/4"
5/16"
3/8"
7/16"
1/2"
9/16"
5/8"
11/16"
3/4"
1"
Depth
3/32"
1/8"
5/32"
3/16"
7/32"
1/4"
9/32"
5/16"
11/32"
3/8"
13/32"
7/16"
15/32"
1/2"
17/32"
9/16"
19/32"
5/8"
21/32"
11/16"
23/32"
3/4"
25/32"
13/16"
27/32"
7/8"
29/32"
15/16"
31/32"
1"
0.108
0.158
0.207
0.256
0.305
0.354
0.404
0.453
0.502
0.551
0.601
0.650
0.699
0.748
0.798
0.847
0.896
0.945
0.995
1.044
1.093
1.142
1.192
1.241
1.290
1.339
1.389
1.438
1.487
1.536
0.199
0.310
0.421
0.532
0.642
0.753
0.864
0.975
1.086
1.196
1.307
1.418
1.529
1.640
1.750
1.861
1.972
2.083
2.193
2.304
2.415
2.526
2.637
2.747
2.858
2.969
3.080
3.191
3.301
3.412
0.275
0.472
0.669
0.866
1.063
1.260
1.457
1.654
1.851
2.048
2.245
2.442
2.639
2.836
3.033
3.230
3.427
3.624
3.821
4.018
4.215
4.412
4.608
4.805
5.002
5.199
5.396
5.593
5.790
5.987
0.615
0.922
1.230
1.538
1.846
2.153
2.461
2.769
3.077
3.384
3.692
4.000
4.308
4.615
4.923
5.231
5.539
5.846
6.154
6.462
6.770
7.077
7.385
7.693
8.001
8.308
8.616
8.924
9.232
0.708
1.151
1.594
2.037
2.480
2.923
3.367
3.810
4.253
4.696
5.139
5.582
6.026
6.469
6.912
7.355
7.798
8.241
8.685
9.128
9.571
10.01
10.46
10.90
11.34
11.79
12.23
12.67
13.12
1.325
1.928
2.531
3.134
3.737
4.341
4.944
5.547
6.150
6.753
7.357
7.960
8.563
9.166
9.769
10.37
10.98
11.58
12.18
12.79
13.39
13.99
14.59
15.20
15.80
16.40
17.01
17.61
1.414
2.202
2.990
3.778
4.566
5.354
6.142
6.929
7.717
8.505
9.293
10.08
10.87
11.66
12.44
13.23
14.02
14.81
15.60
16.38
17.17
17.96
18.75
19.54
20.32
21.11
21.90
22.69
2.388
3.385
4.382
5.379
6.376
7.374
8.371
9.368
10.36
11.36
12.36
13.36
14.35
15.35
16.35
17.34
18.34
19.34
20.34
21.33
22.33
23.33
24.32
25.32
26.32
27.32
28.31
3.686
4.917
6.148
7.379
8.610
9.841
11.07
12.30
13.53
14.77
16.00
17.23
18.46
19.69
20.92
22.15
23.38
24.61
25.84
27.08
28.31
29.54
30.77
32.00
33.23
34.46
3.863
5.353
6.842
8.332
9.821
11.31
12.80
14.29
15.78
17.27
18.76
20.25
21.74
23.23
24.72
26.21
27.70
29.19
30.67
32.16
33.65
35.14
36.63
38.12
39.61
41.10
7.433
9.206
10.98
12.75
14.52
16.30
18.07
19.84
21.61
23.39
25.16
26.93
28.70
30.48
32.25
34.02
35.80
37.57
39.34
41.11
42.89
44.66
46.43
48.20
11.32
14.47
17.62
20.77
23.92
27.07
30.22
33.38
36.53
39.68
42.83
45.98
49.13
52.28
55.44
58.59
61.74
64.89
68.04
71.19
74.34
77.50
80.65
Shop Balancing
629
The tabulated weights may be scaled up or down to accommodate materials

of different density by using the following equation (11-69):
Material Density
Weight Removed grams = Table Value ----------------------------------------------------
0.283
(11-69)
Densities for many common metals are presented in Appendix B of this

text. For instance, if a 1/2 diameter hole is drilled to a depth of 5/8 in steel, the
weight removed from Table 11-26 is found to be 13.23 grams. If the drilled material was aluminum, the density from Appendix B is equal to 0.0975 Pounds per
Inch3, and the amount of weight removed is computed by:
0.0975
Aluminum Weight Removed grams = 13.23 Grams ---------------- = 4.56 Grams
0.283
If the material subjected to drilling was heavier than steel, for example
brass with a density of 0.308 Pounds per Inch3, the weight removed by drilling
the same sized hole may be determined by:
0.308
Brass Weight Removed grams = 13.23 Grams ------------- = 14.40 Grams
0.283
Hence, using the combination of Table 11-26 and the actual material density, the diagnostician may accurately estimate the weight removed by drilling
holes in any material. In addition, different hole depths may be extrapolated
from Table 11-26. For instance, if a hole depth of 1.25 inches was selected, the
weight for a 1.0 may be combined with a 1/4 hole to approximate the weight of
metal removed. Note, there will be some variation with the actual metal removed
due to the inclusion of two drill tip cone weights from the table.
SHOP BALANCING
The vast majority of the rotating elements for process machines are shop
balanced with slow speed balancing machines. As discussed in previous portions
of this text, this is a generally acceptable procedure if all rotating components
are dimensionally correct, and all components are properly balanced before and
during assembly. In addition, rotor flexibility and the number of balance resonances (critical speeds) to be transcended, and the associated shaft mode shapes
at each resonance should be considered. The general characteristics of these slow
speed shop balancing machines will be discussed in this section.
There are also two families of high speed balancing machines that are operational in some shops. This includes the vertical high speed balance pits used for
small expander and compressor impellers. These machines may run in excess of
60,000 RPM, and they are generally employed for balancing individual components. Vertical balancing machines in this category are typically used by OEMs
630
Chapter-11
for component balancing, and they are generally not used for assembled rotors.
A few large horizontal machines are in existence that can accommodate turbomachinery rotors up to 40,000 pounds in weight, and they can operate at
speeds in excess of 20,000 RPM. This second category of high speed balancing
machines are installed in vacuum chambers to minimize aerodynamic (windage)
forces from turbine blades and compressor impeller vanes. These installations
include complete lube oil supply systems, sophisticated speed control systems,
vacuum pumps, variable stiffness pedestals, plus extensive vibration data acquisition and processing instrumentation. These facilities are expensive to build
and maintain, but when a rotor demands a high speed shop balance, no other
type of balancing machine is acceptable. Most of these machines are quite
unique, and many units approach the complexity of a formal shop test stand.
Due to the many intricacies of these units, a detailed examination of high speed
shop balancing machines exceeds the current scope of this text.
The discussion contained herein is directed at the common types of shop
balancing machines. These are slow speed balancing machines that operate
between 200 and 1,000 RPM. In these units, the rotor to be balanced is placed
between a pair of rollers, and the rotor is driven by a direct belt drive mechanism. Vibration pickups are mounted by each set of rollers (bearings), and an
optical trigger probe is employed to measure rotor speed and phase. Machines in
this category are further subdivided as soft bearing or hard bearing machines.
A rendition of a typical soft bearing shop balancing machine is presented
in Fig. 11-52. Soft bearing balancing machines are designed with rollers that are
mounted on either flexible supports or sliding carriages that react to the unbalance forces. The horizontal motion of the carriage assembly is measured with a
velocity or a displacement transducer as shown in the Fig. 11-57. An electronic
data collection system is used to record the vibration response of the calibration
weight, and calculate the weight correction. Machines of this type are very flexible, and the horizontal stiffness is quite low to allow maximum measurable
motion. As such, the first natural frequency for the balancing machine is very
low, and the rotors are balanced at speeds above the first critical of the balance
machine. Due to this soft horizontal stiffness, an unbalanced rotor plus the supComponent
to be
Balanced
Resultant Motion
Due To Unbalance
Sliding Base
Rigid Base
Fig. 1157 Soft Bearing

Shop Balancing Machine
Vibration
Transducer
Shop Balancing
631
port rollers vibrate freely in the horizontal direction. From an analogy standpoint, a soft bearing balancing machine is like a velocity coil. That is, it operates
above the fundamental natural frequency of the spring-mass-damper system.
Soft bearing machines generally readout in displacement, and must be calibrated by placing known weights at defined angles at each of the selected balance correction planes. In many ways this is similar to field balancing with
calibration weights. However, in the case of the shop soft bearing machine, the
detailed balancing calculations are performed by the electronics box attached to
the measurement instrumentation. It should be noted that rotors of different
weights with the same amount of unbalance will vibrate differently on a soft
bearing machine. This is due to the influence of the rotor weight upon the resonance, and the resultant behavior of the rotor above the balancing machine resonant speed. These machines may also be somewhat dangerous in the hands of an
inexperienced operator. If a rotor contains a large unbalance (original or induced
by the operator), the horizontal forces may be sufficient to drive the spinning
rotor out of the balance machine.
The more common type of shop balancing unit is known as a hard bearing
balancing machine. An example of this machine is presented in Fig. 11-58. This
type of balancing machine consists of rigid rollers that are mounted on vertical
support pedestals. This type of balancing machine operates well below the natural frequency of the combined rotor and support system. Using a vibration transducer analogy, a soft bearing machine runs above the resonance like a velocity
pickup and a hard bearing balancing machine operates below the system resonance like an accelerometer. In many cases, a soft bearing machine with flexible
pedestals may physically look like a hard bearing machine with rigid supports.
To be safe, the diagnostician should always inquire about the specific characteristics of any machine used for balancing machinery rotors.
Hard bearing balancing machines come in a variety of sizes, and they can
handle a wide range of rotor weights and configurations. The mechanical motion
Component
to be
Balanced
Resultant Motion
Due To Unbalance
Vibration
Transducer
Fig. 1158 Hard Bearing

Shop Balancing Machine
Rigid Base
Tuned
Vertical
Supports
632
Chapter-11
produced by a rotor on a hard bearing machine is generally proportional to the

centrifugal force in the rotor to be balanced. This is an advantage over the soft
bearing units. Specifically, on a hard bearing machine, the vibration signal is not
influenced by rotor weight, or the bearing and pedestal mass. Hence, a hard
bearing machine typically displays a constant relationship between measured
vibration and residual rotor unbalance.
Hard bearing machines are also equipped with various types of electronic
packages and built-in computers to process the measurements and display vector
outputs of unbalance in units of grams, ounces, gram-inches, or ounce-inches,
plus the appropriate phase angle. Some machines incorporate a digital display,
and other units have a desirable polar coordinate display screen for each balance
plane. Modern units are also combined with various types of printers for hard
copy documentation, plus digital storage of the rotor balance response characteristics. This is quite important for production balancing of many similar rotors.
Overall, the hard bearing machine is easier and faster to operate. However, certain rotating machinery may require a soft bearing machine because of the operating conditions and/or sensitivity of the unit.
Before balancing is attempted on stacked rotors, the individual rotor elements, such as the shaft, balance piston, thrust collar, and all of the impellers,
should be component balanced. The shaft runout must always be checked and
any discrepancies addressed. In addition, the journals must be measured and
verified as both round and straight. Failure to start with a straight shaft and
proper journals will compromise the entire balance project by doing nothing but
balancing the rotor bow and/or eccentricity. On large shafts or solid turbine
rotors it is often necessary to roll the shafts for several hours in the balancing
machine before attempting any balance corrections. This also applies on built-up
rotors such as large gas turbines with precision axial fits between the wheels.
On bare shafts with keyways, it is mandatory that all keyways are filled
with half keys. On rotating elements such as impellers, thrust collars, and balance pistons, a precision shop mandrel must be used to component balance each
of these elements. The mandrel itself must be a perfectly straight shaft that is
balanced to the lowest possible levels of residual unbalance. The mandrel must
not contribute to any unbalance or eccentricity of the machine element to be balanced. A light interference fit should be used between the rotor element and the
mandrel. There can be no looseness or wobble of the component to be balanced on
the mandrel. In all cases, accurate mandrels will help insure a properly balanced
rotor assembly, and sloppy mandrels will result in unnecessary grief.
After all components are balanced to within the specified tolerances, it is
time to assemble the rotor. In some cases, the components are stacked on the
rotor, and the assembled rotor is subjected to a final trim balance. On more sensitive rotors, the balancing continues concurrently with the rotor assembly. That
is, the rotor assembly begins by mounting the middle one or two wheels with the
proper keys, and interference fit to the shaft. After these initial wheel(s) are
mounted, the shaft is rechecked for runout, and the wheels are checked for radial
runout plus axial wobble. These readings become part of the job documentation,
and they must be retained for comparison as the rotor balance and assembly
Shop Balancing
633
progresses. If runouts are acceptable, then the balance of the shaft plus the initial wheel(s) are checked. At this stage, only minor corrections should be necessary. If the balance machine calls for major weight corrections, then something is
wrong, and the only realistic option is to stop and find the origin of the problem.
Following the mounting of the initial wheels, the rotor is built-up or stacked
by adding wheels in pairs. One wheel is added from each end of the rotor, and
any necessary two plane balance corrections are performed on the recently added
wheels. This procedure is continued until the entire rotor assembly is completed,
and all runouts and critical radial and axial dimensions are within specification.
All rotating components should be mounted and indexed to some repeatable
point of reference. This insures that all of the rotor components are reassembled
to the same reference point during each rebuild of the rotor.
Some rotors can only be stacked from one direction. On these units the
wheels are added one at a time, and typically a single plane balance would be
performed on the last installed wheel. On solid or integral rotors such as steam
turbines, and segmented rotors such as gas turbines, the ability to stack balance
does not exist. On these rotor configurations, the trim corrections performed on
the shop balancing machine should only be performed after a suitable slow roll
period to insure that any shaft bows are minimal.
When balancing an assembled rotor with multiple elements, weight
removal or addition must be carefully evaluated. When a static type unbalance is
found, weight corrections should be made at the center planes to minimize
deflection through the translational critical. If a couple unbalance is encountered, weight corrections should be made to the modally effective outboard balance planes. Preferably, weight changes should be made by removing material
from a mechanically sound location. In most cases, it is advisable to clay the
rotor before any permanent weight changes are made to the rotor. It is highly
desirable to refrain from using field balancing holes for shop balancing. When a
rotor leaves the shop balancing machine, the accessible field balancing holes
should either be totally empty, or totally filled with full-length balance weights.
Either approach will allow the machinery diagnostician the ability to fully utilize
the effectiveness of the field balancing planes. Finally, when balance weights are
welded to a rotor, they should be located were they will not be a personnel hazard, and they must not affect rotor performance or provide any potential interference with any stationary parts of the assembled machine.
The issue of acceptable balance tolerances is an integral part of any discussion of shop balancing. The various standards and specifications that are applied
to this work are somewhat overwhelming. Numerous standards organizations
and various types of OEMs have different views of what constitutes an acceptable residual unbalance. These variations are understandable, since machines
such as gear boxes tolerate higher levels of unbalance due to the fact that the
transmitted gear forces are considerably larger than any mass unbalance forces.
In another example, centrifugal pump rotors are usually restrained by impeller
wear rings, and the pumped fluid provides significant damping to the rotor.
Hence, many pump rotors operate successfully with higher residual unbalance
levels. These existing standards vary from prescribing a particular fraction of
634
Chapter-11
the acceleration of gravity (e.g., 0.1 Gs) to some combination of rotor weight and
speed. Many industries have adapted the API standard of 4W/N as the acceptable shop balance tolerance. This expression is defined as follows:
4W
U oz in = --------------N
(11-70)
where: Uoz-in = Maximum Residual Unbalance Per Plane (Ounce-Inches)

W = Journal Static Weight (Pounds)
N = Maximum Continuous Machine Speed (Revolutions / Minute)
For instance, if a 1,500 pound rotor has a maximum speed of 5,500 RPM,
the acceptable residual unbalance per journal would be calculated as follows:
4W
4 ( 1, 500 2 )
4 750
U oz in = --------------- = ------------------------------------ = ------------------ = 0.545 Ounce-Inches
5, 500
N
5, 500
This level of residual unbalance is generally achievable with the typical
shop balancing machines. In fact, some end users have reduced this tolerance to
2W/N for critical machines, or during component balancing to reduce the accumulated residual unbalance. Equation (11-70) requires a knowledge of the static
rotor weight per journal. If the rotor is symmetrical, the total rotor weight may
be divided in half to determine the weight per journal. For a non-uniform rotor,
the journal weights must be measured or calculated as in case history 3. This is
particularly important on rotors such as large steam turbines that have more
weight supported by the exhaust bearing versus the governor end bearing. The
maximum continuous operating speed of the machine must be used to calculate
the balance tolerance in (11-70). Please note, this is not the balancing machine
speed, it is the maximum continuous speed of the rotor installed in the field. For
those that prefer balance units of grams, equation (11-70) may be restated in
equation (11-71) where the residual unbalance is in gram inches:
113.4 W
U gm in = ------------------------N
(11-71)
The journal static weight W is still expressed in pounds, and the maximum
speed N is still stated in revolutions per minute. The only difference between
equations (11-70) and (11-71) is that the conversion factor of 28.35 grams per
ounce has been included in equation (11-71).
Once a rotor has met the balance specifications, it is desirable to perform a
residual unbalance check to verify that the rotor is balanced correctly at each
plane. This not only proves that the rotor is balanced, but it will verify that the
machine is working properly. A residual unbalance check is accomplished by taking a known amount of weight at a known radius, and a known angle, and verifying that the balance machine properly tracks the weight. For example, if the
4W/N balance tolerance is 0.50 ounce-inches, this is equivalent to 14.18 graminches. For a 10 inch correction radius, then the tolerable unbalance would be 1.4
grams. After the component (or rotor) has been balanced, assume that the bal-
Shop Balancing
635
ancing machine displays a 1.2 gram unbalance at 90, and a radius of 10 inches.
This residual unbalance can be verified by placing a 10 gram weight sequentially
at 12 locations (30 apart) at the 10 inch radius. The balance machine reading
(grams at degrees) are recorded at each position, and the twelve measured vectors are plotted in a polar coordinate format. Fig. 11-59 displays the results
obtained by adding 10 grams at 0, and then moving this weight at 30 increments. Note that the measured balance machine angles are close to 30 increments, but some deviations appear. This is due to slight differences in actual
placement of the clay test weight. If a rotor was equipped with 12 evenly spaced
balance holes, the angular variation would be minimal, and adjacent readings
would generally be separated by 30.
0
30
330
60
300
90
270
0
5
Grams
10
240
Fig. 1159 Twelve Point

Residual Unbalance
Check On Shop Balance
15
120
210
150
180
If the final polar plot does not reveal a reasonably circular pattern, then the
offending point(s) should be rerun. The residual unbalance is simply the center
of the data circle. To determine the residual unbalance, take the highest amplitude reading, subtract the lowest reading, and divide the difference by two as in
the following equation (11-72):
Maximum Minimum
Residual Unbalance = ------------------------------------------------------------------
2
(11-72)
With respect to the data shown in Fig. 11-59, the residual unbalance is easily determined from equation (11-72) as follows:
Max Min
12.4 10.0
Residual Unbalance = ------------------------------- = --------------------------- = 1.2 Grams
2
2
This residual unbalance should be equal to the final display on the balancing machine. If these residuals are equal, the balancing machine is functioning
properly. If the 10.0 inch balance weight radius is included, the following applies
636
Chapter-11
1 Oz
Residual = 1.2 Grams 10.0 Inches = 12.0 Gm-In ----------------------- = 0.42 Ounce-Inches
28.35 Gm
If the balancing machine is set to read in gram-inches instead of grams,
then this final step is not required. In some cases it is also desirable to determine
if the balancing machine is too sensitive, or not sensitive enough for the component. Hence, the performance of a sensitivity test may be desirable, and the manufacturer of the balancing machine should provide specific test instructions.
There are certain types of machines, such as three bearing units, highly
flexible rotors, and/or rotors that operate above several critical speeds that cannot be successfully shop balanced. Machines of this type should be shop balanced
to the lowest possible levels, and final trim balance corrections performed in the
field when the rotor is installed in the casing. This approach is also necessary for
rotors that display unusual synchronous 1X variations as a function of temperature or load. If the machine will faithfully duplicate a particular 1X response
(e.g., hot to cold, and cold to hot), there is a good possibility that this type of
repetitive behavior can be attenuated with a properly sized and positioned balance weight correction.
As stated earlier in this chapter, mass unbalance corrections are very useful for extending the life span of a piece of mechanical equipment by reducing the
level of transmitted forces. However, like any corrective measure, shop or field
balancing should always be placed in the context of what are we really balancing
out, and where did it come from. In some cases, the application of rotor balancing
may mask a more serious problem (e.g., cracked shaft). In all cases, the machinery diagnostician should thoroughly review the situation, and strive to maintain
equilibrium between sound engineering judgment and economic responsibility.
BIBLIOGRAPHY
1. Bently, Donald E., Polar Plotting Applications for Rotating Machinery, Proceedings
of the Vibration Institute Machinery Vibrations IV Seminar, Cherry Hill, New Jersey
(November 1980).
2. Eisenmann, Robert C., Some realities of field balancing, Orbit, Vol. 18, No. 2 (June
1997), pp. 12-17.
3. Eshelman, R., Development of Methods and Equipment for Balancing Flexible
Rotors, Armour Research Foundation, Illinois Institute of Technology, Final Report
NOBS Contract 78753, Chicago, Illinois (May 1962).
4. Gunter, E.J., L.E. Barrett, and P.E. Allaire, Balancing of Multimass Flexible
Rotors, Proceedings of the Fifth Turbomachinery Symposium, Gas Turbine Laboratories, Texas A&M University, College Station, Texas (October 1976), pp. 133-147.
5. Jackson, Charles, Balance Rotors by Orbit Analysis, Hydrocarbon Processing, Vol.
50, No. 1 (January 1971).
6. Palazzolo, Alan B., and Edgar J. Gunter, Multimass Flexible Rotor Balancing By
The Least Squares Error Method, Rotor Dynamics Course Notes, Part IV, (University of Virginia, Charlottesville, Virginia, 1981), pp. 92-123.
7. Thearle, E.L.,Dynamic Balancing of Rotating Machinery in the Field, Transactions
of the American Society of Mechanical Engineers, Vol. 56 (1934), pp. 745-753.
Machinery Alignment
12
12
isalignment problems have plagued

machines ever since the need evolved to transmit torque from one mechanical
device to another. Early industrial machines did not have extensive alignment
problems due to low speeds, low horsepower, and compliant connections. However, as machinery sophistication developed over time, so have the requirements
for improved alignment. Modern machinery trains usually consist of primary
drivers directly coupled to the driven equipment. The induction motor driving
the process pump, the gas turbine driving the feed gas compressors, and the
steam turbines driving the generator all share a common characteristic. All of
these machinery trains require alignment of components, internals, and coupled
shafts for safe and reliable operation.
To meet the needs for improved machinery alignment, a number of technical approaches have been devised, and many good references are available to the
machinery diagnostician. One of the earliest techniques to employ electronic
instrumentation for this task was the 1968 paper by Charles Jackson1. More
recently, John Piotrowski2 authored the latest revision of his book that includes
the details of laser tools for improved alignment measurements and accuracy.
These techniques have been further refined with the use of portable computers
that interface directly with the measurement system. These devices allow consistent recording of the moves, plus automated calculations and graphing.
It should be self-evident that proper alignment is critical to the life of a
machine, and the consequences of misalignment can be seen through the train.
Coupling wear or failure, bearing failures, bent rotors or crankshafts, plus bearing housing damage are all common results of poor alignment. The extent of the
damage is directly related to the magnitude of the misalignment. For example, a
general purpose motor driven pump with a slight shaft misalignment might
experience premature seal failure, bearing damage, or coupling wear. In this
case, the marginal shaft alignment decreases the time between failures, and
increases the annual repair costs. However, a severe misalignment between this
pump and motor could be potentially destructive to the plant surroundings, and
1 Jackson, Charles. Shaft Alignment Using Proximity Probes, ASME Paper 68-PET-25, Dallas, Texas (September 1968).
2 Piotrowski, John, Shaft Alignment Handbook - 2nd ed., (New York: Marcel Dekker, Inc., 1995).
637
638
Chapter-12
the attendant personnel. In fact, there have been many documented cases of couplings that have traveled for several thousand feet after a failure. In this scenario, the primary concerns include personnel safety, plus extensive repair costs.
In the overview, alignment consists of three distinct categories that are
identified as shaft, bore, and position alignment. Shaft alignment is the most
common form of alignment performed on machinery. There are many procedures,
practices, and tools available to obtain a precision shaft alignment. Bore alignment addresses the position of the internal machine components with relation to
fixed items such as the main bearings. Bore alignment is used for tasks such as
locating diaphragms with respect to bearing centerlines. Position alignment is
primarily reserved for machine location or elevation. It is also commonly used to
measure and correct for thermal growth of the machinery. Within this chapter,
the fundamentals of machinery position, bore, and shaft alignment will be discussed, and descriptive case histories will be presented.
PRE-ALIGNMENT CONSIDERATIONS
Prior to embarking on any alignment project, the diagnostician must evaluate the machinery installation, and select the method, tools, and procedures to
be applied. Since each machinery installation differs is size, speed, power, location, and function, it is necessary to integrate all of the alignment variables in a
cohesive plan before commencing the actual work. The fundamental items to be
addressed are summarized as follows:
Machine arrangement, type, bearing configuration, and viewing position.

Coupling type, condition, runout, speed, and transmitted torque.
Potential thermal growth or shrinkage.
Potential pipe strain.
Condition of foundation, baseplate, sole plate, and anchor bolts.
Location and condition of leveling bolts and jack bolts.
Shim selection, and soft foot checks.
Obstructions to alignment work.
Machinery alignment offsets and tolerances.
Machine arrangement often dictates the alignment process and method.

On any train, it is necessary to identify the fixed plus the moveable machines.
The fixed equipment is the unit that will not be moved during the alignment
work. Conversely, the moveable machines will be moved to obtain the correct
alignment. For instance, in a typical pump-motor application, the pump remains
fixed, and the motor is the moveable machine. In a turbine driven compressor
train, the turbine remains fixed, and the compressor is the moveable unit. If a
gear box is included between the turbine and compressor, the gear box becomes
the fixed machine, and the turbine and the compressor both become moveable.
The criteria for determination of which machine is fixed or moveable is basically
a decision defining the moveability of the various machines. A motor typically
Pre-Alignment Considerations
639
has no external forces or pipe strain, and it may be easily moved. Turbines are
sensitive to external forces, and they usually become the fixed units. A gear box
in a machine train almost always becomes the fixed equipment. If the gear box is
moved, the alignment moves required farther down the machine train may not
be possible. Certainly there are exceptions, such as reciprocating compressors
that are driven through a gear. In this arrangement, the compressor becomes the
fixed equipment, and the gear box plus driver become the moveable units.
Machine types and specific bearing configurations can greatly influence the alignment process due to special requirements or considerations. Since
most alignment techniques are performed at zero speed, it is necessary to anticipate and accommodate the machinery behavior between zero speed and full operating conditions. For example, large machines with sleeve bearings will
experience a shift in shaft centerline position as the journal progresses from a
rest position to full running speed. The shaft will rise on the oil film in the direction of rotation, and this centerline change should be considered in the cold
alignment offset. As another example, tilting pad bearings can cause problems
during shaft alignment. Machines with load between pads (LBP) bearings retain
the shaft during alignment. However, machines equipped with load on pad (LOP)
bearings can cause problems if the shafts are turned in different directions during alignment sweeps. This may cause the bottom pad to pivot, and this could
cause the shaft to shift horizontally, which would corrupt the alignment data.
Special conditions exist for heavy rotors such as industrial gas turbines, or
long generator rotors. These types of assemblies will sag due to gravity when the
shaft is not turning. Once the machine is running, centrifugal and gyroscopic
forces will straighten the rotor. However, the static gravitational sag, or bow, will
cause the shaft ends to deflect upward outboard of the journal bearings. It is necessary to know this shaft deflection to set the machine at the proper at rest location. If the catenary curve describing the static rotor is not known, then it may
be computed with the analytical techniques described in chapter 5 of this text.
Overhung machines have a similar problem, due to the fact that the overhung wheel often pulls the rotor down at zero speed. In essence, the shaft pivots
across the wheel end bearing, and this action forces the coupling end journal to
the top of the available bearing clearance (or vice-versa depending on specific
rotor geometry). To compensate for this motion, it is common practice to push the
shaft back down into the bearing with a pair of rollers positioned on top of the
coupling end journal. Gear boxes may also provide difficult situations during the
alignment process. As shown in Figs. 10-1 and 10-2, the direction of the gear contact forces are opposite for each gear. One element tries to sit down in the bearing, and the other gear wants to climb to the top of its respective bearing. Again,
this actual running position should be taken into account during alignment.
Although radial offsets generally command most of the attention during
machinery alignment, the diagnostician must also consider the axial position of
the respective rotors. For instance, gear boxes with double helical elements will
typically have only one thrust bearing on the bull gear. The pinion will center
itself in the helix, and this axial running position must identified in order to set
the proper axial coupling spacing for the high speed pinion. Similarly, the mag-
640
Chapter-12
netic center for motors must be identified so that the motor rotor may be properly located in the running condition. This knowledge will allow the correct
spacing to be established between the motor and the driven coupling hub.
Although axial spacing is important on any machine, it is essential to maintain
the proper axial dimensions on units equipped with diaphragm couplings.
The viewing position of the machinery to be aligned is critical to any
alignment work. The fundamental concepts of up-down, left-right, and fore-aft
become totally meaningless unless a definitive observation or reference point is
used. This observation point or direction must be established at the start of the
alignment work, and that point is maintained throughout the entire project. Furthermore, the final alignment documentation must reflect this viewing direction.
It is also desirable to reference the compass points for additional clarification.
This viewing position often varies between centrifugal and reciprocating
machines. Hence, complete documentation of the machinery layout is the only
way to maintain historical continuity.
Coupling types will have a significant influence on alignment, and it is
important to understand how a particular coupling works. Many technical
papers have been published on the various coupling types. In addition, excellent
overviews on the entire topic are presented in the 1986 book by Jon Mancuso3
plus the 1994 text by Mike Calistrat4. As mentioned in the previous paragraph,
part of shaft alignment is the proper setting of coupling spacing, and/or shaft end
gap. This type of information, plus the associated tolerances, are generally specified on the certified coupling drawing. Coupling types will also govern the configuration of the dial indicator alignment bracket(s). Also, it is important to
determine how the shafts will be rotated to take the alignment readings. On
flange type gear couplings, the flange bolts work well to turn the shafts, but due
to their size they often require the use of larger brackets.
Furthermore, in planning the alignment job, consideration should be given
to the handling and intermediate storage of the coupling parts. For instance, a
gear coupling may have two dozen coupling bolts to be removed, and saved for
the reassembly. However, some diaphragm couplings with intricate spool piece
designs fall into the category of 1,000 bolt couplings. These units require a lot of
time to disassemble, and even more time to put back together. Coupling bolts are
usually body fit bolts that are matched and balanced. If someone happens to lose
a coupling bolt or nut, the entire project becomes unavoidably delayed.
The physical condition of the coupling can greatly affect the alignment. A
worn or damaged coupling may produce erratic readings, or have high runouts.
It is also easy to misdiagnose vibration data as misalignment when the real
problem is a damaged coupling. In all cases, the coupling should be thoroughly
inspected prior to beginning any alignment job. Check the gear teeth, shim
packs, grid members, bolts, or whatever components exist in the coupling assem3 Jon R. Mancuso, Couplings and Joints - Design, Selection, and Application, (New York: Marcel Dekker, Inc., 1986).
4 Michael M. Calistrat, Flexible Couplings, Their Design Selection and Use, (Houston: Caroline
Publishing, 1994).
641
bly. Elongated bolt holes, signs of excessive heat, or any evidence of damaged or
excessive wear should be sufficient information to have the coupling committed
to the dumpster. This simple act will save considerable time and effort during
alignment, plus it will improve overall machinery reliability.
Shaft or coupling hub runout can influence alignment readings, and it
can reflect the machine condition, or the presence of other mechanical problems.
Runouts should always be checked prior to the alignment process. Shaft runout
may reveal internal mechanical problems such as a cracked or distorted shaft. In
these situations, the problem should be resolved before continuing with the
alignment. If the shaft runout originates due to a poorly fabricated shaft or
rough surface finish, the surface may be cleaned up or polished. Coupling hub
runout can be indicative of problems such as a hub bored off center, tapered or
cocked hub, loose shaft fit, or any number of other reasons. Some of these problems must be corrected, and others may be dealt with in the alignment process.
The critical factor is that the shaft and coupling runouts must be checked,
recorded, and their origin specifically understood.
From equation (2-98) it was shown that torque is a function of both speed
and horsepower. To state it another way, the transmitted horsepower is the product of speed and torque. Naturally a constant must be applied to obtain the correct engineering units, but the fundamental relationship persists. That is,
increases in speed or torque will result in an increase of the transmitted horsepower. In many ways, this formulates part of a difficult engineering problem of
accommodating machinery power increases, while maintaining mechanical
integrity of the entire coupling assembly.
Truly, the coupling in many modern machines represents a separate rotor
system that is often isolated via flexible connections to the adjacent machines. As
such, the diagnostician must be aware of the potential for coupling problems
associated with mass unbalance, eccentricity, excessive clearances, a variety of
preloads, plus the possibility for resonant behavior on axially compliant couplings. In the majority of large or critical machines, the couplings are precision
elements that are designed to meet a specific set of criteria. Any attempt to
increase the speed or load rating of these units should be made with proper engineering study and surveillance. Although some operating facilities seem to
thrive on continually changing coupling types and configurations there is considerable wisdom in staying with a coupling design that has a proven track
record for continuous and reliable service.
Most machines experience temperature changes between the cold condition
during which the alignment is performed, and the loaded condition at full process rates. It is common knowledge that hot machines such as steam turbines
expand, and cold machines such as cryogenic pumps shrink. Typically, the cold
alignment condition is adjusted to compensate for the anticipated change in
physical dimensions. In some cases, the thermal growth or contraction is
measured as discussed later in this chapter. In other instances, particularly on
new installations, the amount of thermal growth must be estimated during
alignment. Fortunately, industrial metals have a reasonably uniform coefficient
of thermal expansion. This material constant may be combined with the length
642
Chapter-12
of the machine element and the differential temperature change, to determine

the dimensional change. The specific relationship is expressed as follows:
L = T L C
L
T
Where
=
=
L =
C =
(12-1)
Anticipated Thermal Growth or Contraction (Mils)

Average Change in Temperature (F)
Length or Height of Heat Affected Area (Inches)
Coefficient of Thermal Expansion (Mils / Inch-F)
For example, consider a gear box between an electric motor and a centrifugal compressor. The gear box centerline height is 47.5 inches, and the average
skin temperature is 125F. If the alignment was performed at an ambient temperature of 75F, the thermal growth may be estimated from equation (12-1). If
the gear box is a cast steel housing, the coefficient of thermal expansion may be
obtained from Table B-1 in the appendix of this text. For this material, the coefficient of thermal expansion is 0.0063 Mils/Inch-F. Combining this value with the
previously mentioned physical parameters, the thermal growth may be computed with equation (12-1) as follows:
L = T L C = { 125 75 F } 47.5 Inches 0.0063 Mils/InchF = 15.0 Mils
Thus, the anticipated gear box growth would be 15 Mils vertically between
ambient and operating temperature. Another way to obtain the same result
would be to use the chart presented in Fig. 12-1. In this plot, equation (12-1) was
Steel Thermal Growth (Mils)=0.0063 Height (Inches) x Differential Temp. (F)
100
Support Leg Height (Inches)
100"
90
Thermal Growth of Steel (Mils)
90"
80
80"
70
70"
60
60"
50
50"
40
40"
30
30"
20
20"
10
10"
0
0
20
40
60
80
100
120
140
160
180
Average Differential Temperature (F)

Fig. 121 Thermal Growth Of Cast Steel Machine Elements
200
643
solved for various combinations of differential temperatures (10 to 200F), combined with pedestal or support leg heights ranging from 10 to 100 inches. As
indicated, this chart is for cast steel, and it is sometimes easier to use than the
equation. For this current example, an average differential temperature of 50F
may be located on the bottom axis. A vertical line up to the centerline height of
approximately 47.5 inches, and a horizontal line to the left axis reveals a thermal growth of 15 Mils. Naturally this chart would change for other materials
with different coefficients of thermal expansion.
It should also be mentioned that the average differential temperature
should be taken literally. That is, at operating conditions, a temperature profile
will exist along the length of the support leg, or casing. A series of surface temperature measurements should be made along the length of the element, and a
realistic average operating temperature determined. Under normal alignment
conditions, the entire pedestal will be at the same temperature, which is probably equal to ambient temperature. Hence, it is not necessary to run a temperature profile under this condition. For clarity, the differential temperature T in
equation (12-1) should be more precisely defined as follows:
T = ( T avg-run ) ( T align )
(12-2)
where: Tavg-run = Average Pedestal Temperature During Normal Running Operation (F)
Talign = Ambient Pedestal Temperature During Machinery Alignment (F)
Please note that the average support temperature during normal operating
conditions Tavg-run is not necessarily the same as the temperature at the middle
of the support. Steam leaks or other materials that blow directly onto the pedestal may significantly influence the average pedestal temperature.
The same expressions may be used for the computation of the shrinkage of
machines that operate in cold service. For instance, the 316 SS supports on a
cryogenic expander are 24 inches tall, and the average support temperature during operation is -45F. This may be difficult to determine since the pedestals are
completely covered with ice under this condition. Obviously, the temperature
that must be used is the average metal temperature, and not the ice temperature. If the alignment was performed at an ambient temperature of +75F, the
differential temperature T may be determined from equation (12-2) as:
T = ( T avg-run ) ( T align ) = ( 45F ) ( 75F ) = 120F
As stated, the supports are 316 SS, and from Table B-1, the coefficient of
thermal expansion is 0.0092 Mils/Inch-F. If this value is combined with the differential temperature, and the pedestal height, the thermal shrinkage may be
calculated with equation (12-1) as follows:
L = T L C = { 120 F } 24 Inches 0.0092 Mils/Inch-F = 26.5 Mils
It is sometimes hard to believe that metal can shrink that much when it
gets cold. Nevertheless, it should be realized that the differential temperatures
for cryogenic services are often quite large, and this results in substantial con-
644
Chapter-12
traction of the machinery dimensions. Of course, this is the same problem that
must be addressed in starting up cryogenic equipment to insure that the stationary parts do not contract and seize the rotating elements.
Both hot and cold services have the potential for generating pipe strain on
the machinery. This is a three-dimensional problem that potentially combines
forces and moments on the machinery flanges. These piping loads often vary
from ambient temperature to normal operating conditions. Clearly, pipe strain
can make it impossible to achieve a precision alignment, and it can cause serious
damage to the associated equipment. In most facilities, the original piping systems are designed by the Engineering and Construction firm (E&C). These piping designs must fit in a specified space, and provide the best possible flow
characteristics, combined with acceptable pipe strain. This is a complex design
effort, and computer simulations are used to model the piping system dynamics.
For reciprocating machines, the problem is even more complex, since the acoustic
and pressure pulsations characteristics must also be evaluated and optimized.
The final piping configuration usually includes a variety of pipe hangers, supports, expansion loops, expansion joints, plus various spring loaded cans. In some
cases, dampers or snubbers are required to dissipate the piping system energy.
It should also be mentioned that spring cans are a common and effective
method to help relieve piping strain when they are operational. However, these
spring loaded cans are a piece of hardware that can malfunction. It is important
to periodically inspect these cans, and insure that the pipe is really supported by
the spring. In new plants, the factory installed shipping stops may not be
removed, and the effectiveness of the can is substantially reduced. In addition,
long-term corrosion or internal dirt buildup can cause the spring can to lock up,
and again minimize the effectiveness of the can.
If possible, machines should be aligned with the large bore piping detached.
Prior to attaching this process piping to the machine, each flange should be
checked for aligned vertical and horizontal flange position, parallel flange faces,
and proper bolt hole orientation. In general, angular face alignment, and concentric hole orientation provide greater concerns than minor vertical or horizontal
offsets. Tolerances for pipe alignment are typically available from the E&C, or
the end user engineering specifications. However, the real test of a successful
flange mating between the piping system and the machinery is obtained by monitoring the coupling position as the process flanges are bolted in place. This activity is monitored by mounting vertical and horizontal dial indicators on an
adjacent machine, and observing the shaft or coupling hub motion as the flanges
are connected. Any pipe strain that produces position changes at the coupling of
greater that 2.0 Mils should be corrected.
In all cases, the machinery foundation must be visually inspected for any
cracked, broken, or missing concrete or grout. The primary function of a foundation is to support the gravity and the dynamic loads imposed by the machinery
system. A weak or damaged foundation will not act as a rigid member, and unacceptable movement may occur. Furthermore, oil and chemical contamination can
weaken or corrode a foundation with minimal evidence of surface destruction. In
some cases it is desirable to check the concrete or grout integrity with a chipping
645
hammer, or perhaps a core sample is required in other situations.

Due to the superior compressive and tensile strength of epoxy grout, most
machinery installations use this material between the concrete foundations and
the baseplates. Epoxy grouts also have a greater resistance to water, oil, and
chemical damage. However, these materials are sensitive to stress risers, and
proper installation requires all sharp corners or edges to be radiused. There are
many excellent epoxy grouts available, and the machinery diagnostician should
carefully compare the physical and mechanical properties, plus the advantages
and limitations of each grout to select the proper material for each job.
Prior to any alignment job, it is highly desirable to inspect anchor bolts
for damage. Broken or loose bolts must be replaced, and proper torque values
should be maintained. Many times anchor bolts break above or at the surface of
the foundation. This is usually caused from corrosion, over torquing, misaligned
holes, or loose equipment. Anchor bolts also break below the surface, and this
may not be discovered until the bolt is tightened or the machine fails. Several
inspection techniques exist to test anchor bolts. Ultrasonic testing (UT) is the
most common form of inspection to determine the distance from the top of the
bolt to the first inclusion. This could be the bottom of the bolt, a bend, or a crack.
If the original bolt configuration is not known, the UT data could be useless.
Proper design and installation of machine baseplates or sole plates is
also imperative to a good running machine, and a proper alignment. Initially, the
diagnostician should inspect for corrosion or erosion that may weaken the baseplate, or even separate the plate from the grout. Hollow cavities or low spots are
locations for water or corrosive chemicals to collect and damage the baseplate. A
common practice of sounding a baseplate by tapping the surface with a hammer
can reveal voids or separations. The sound or chime produced will vary between
solid areas where the baseplate and grout are in direct contact versus spots
where voids or separations exist.
Machine or baseplate leveling bolts (sometimes called jack bolts) are
important considerations during initial installation of a baseplate. They allow
the baseplate to be precision leveled quickly and accurately. Often they are
installed on top of a steel plate that rests directly on the concrete foundation.
These bolts must be numerous and sufficiently strong to support the baseplate.
They should be sealed prior to grouting, and they should be removed after the
grout has set. Retained leveling bolts can be detrimental to the installation. It is
possible to distort or separate a baseplate from the grout if leveling bolts are still
in place during final anchor bolt torquing.
In new installations, the baseplates or sole plates must be checked for levelness and flatness. In most plants, the only tool available to check a baseplate is a
precision machinist level. Readings can be used for coplanar comparisons as well
as slope measurements. In most applications a perfectly level installation is
desired, but not mandatory. However, the flatness of the baseplate or sole plate is
critical. If the machine mounting surfaces are not perfectly flat, the internal bore
alignment, as well as the shaft alignment, is in jeopardy. When addressing flatness concerns beyond a simple coplanar comparison, it is often necessary to
employ optical or laser measurements.
646
Chapter-12
Unlike the leveling bolts used to set baseplates, jack bolts are used to
move machines around on a fully installed baseplate. Standards for horizontal,
vertical, and axial jack bolts should be rigidly mounted on the baseplate to allow
quick and easy machine movements. Temporary screws or clamps work well with
smaller machines, but they become ineffective on large, heavy machines. After
the alignment is complete, the jackbolts may be removed from the machine. In
most plants, the jack bolts are allowed to remain in-place for the next alignment
job. In these situations, the jack bolts should be backed off at least 0.25 inches
from the machine foot. This will allow the machine to grow, and not be restrained
by a close fitting jack bolt. The concept that the jack bolts keep the machine in
line during operation is incorrect. In actuality, a tight jack bolt can provide a detrimental restraining force on the machine.
Shims should always be thoroughly inspected on any alignment job. Corroded, damaged, or separated shims must be replaced. Although some individuals tend to reuse old shims for convenience, the replacement of questionable
shims will save a great deal of time during the alignment process. There are
many companies that manufacture and sell precut shims in various configurations. These shims are usually sold in sets that range in thickness from 1.5 to
250 Mils. Shim stock material can also be purchased in rolls, and cut to match
the dimensions of the machine foot. Laminated (or glued together) shims are also
available in standard fractional thickness with 1.5 to 3.0 Mil thick laminations.
The selection of proper shim materials is a very important consideration.
Carbon steel shims should not be used due the effective thickness growth as carbon steel begins to rust. Stainless steel or brass shims should be used whenever
possible. Many times carbon steel shims are used on custom fit applications
where the surface grinders are equipped with magnetic tables. Obviously, stainless steel or brass cannot be retained by a magnetic table, and only carbon steel
shims can be ground. Nevertheless, stainless steel is the preferred shim stock
material, even though it is more expensive and harder to cut than brass. In
applications where brass will not be chemically attacked or subjected to severe
environments, brass shims may be used. Brass is often used for large shims due
to their ease of use and general availability.
During initial installation of a machinery train, the individual machines
should be set on very thick shims. In small applications, a shim thickness of
0.125 inches is acceptable. For larger machines, the initial shim thickness should
be at least 0.375 to 0.500 inches to provide an adequate range for vertical alignment moves. On very hot machines, the initial shim thickness may be even
larger. Rust, paint, dirt and other foreign objects can cause the machine to move
unpredictably, or allow the machine to move while in service. To insure a good
alignment, the machine foot, sole plates, and shims must all be clean and dry.
The number of shims installed under a machine foot should typically be
limited to 5 shims. If more than 5 shims are required to align a piece of equipment, thinner shims should be swapped out for thicker ones. Contamination and
shim springiness become problems when excessive shims are used. The size of
the shim should be large enough to support the machine, and allow the hold
down bolts to pull down evenly. Precut shims are sufficient for smaller general
647
purpose equipment, but large machines may require custom fit shims. When
installing shims, the shim area should be a minimum of 75% of the machine foot.
An example of an improperly supported machine was presented in case history
33, which describes the problems encountered on a 3,000 HP induction motor.
In many machines, one or more support feet may be higher or lower than
the others, or one or more support feet may be bent in relation to the others. This
condition is commonly referred to as soft foot, and it can cause many problems
during alignment and operation. Soft foot may be checked with a feeler gauge, or
a dial indicator. Ideally, this is checked before starting alignment by removing all
shims from under each machine foot, and setting the machine flat on the baseplate. Feeler gauges should then be used to measure any clearance between each
machine support foot and the baseplate. Clearances of less than 2.0 Mils are generally acceptable. Next, tighten down all hold down bolts, and then loosen each
foot consecutively, and measure any clearance with the feeler gauges. This simple test is then repeated for each of the remaining machine feet. If clearances
exceed 2.0 Mils in either feeler gauge check, the soft foot must either be corrected
by machining or shimming.
Once shims have been installed, the feeler gauge method may become difficult. The second way to check soft foot is by using a dial indicator. To check the
soft foot with this method tighten down all hold down bolts, and position a magnetic base and dial indicator to measure deflection of the machine foot with
respect to the base. The indicator should be located so that it will not interfere
with loosing or tightening of the foot bolts. After the indicator is zeroed, loosen
the foot bolt, and read the total movement shown by the dial indicator. If the
indicator reads less than 2.0 or 3.0 Mils the foot is acceptable. Repeat the process
for each of the machine feet and shim as required.
During the alignment process it is necessary to set up measurement equipment, rotate shafts, tighten and loosen bolts, etc. Physical obstructions may
seriously hamper this work. In position alignment, equipment placement is very
important for accessibility to all measurements points. Walls, beams, piping, and
machine components can all interfere with the line of sight. In any type of optical
alignment, the placement of the tripod, and the targets may also interfere with
the tightening of bolts, or the installation of other machine components. During
shaft alignment, coupling guards, oil piping, or typical bearing instrumentation
may interfere with the indicator brackets. These items can damage equipment or
greatly lengthen the alignment process. In all cases, the diagnostician should
think through the alignment process, and attempt to eliminate or work around as
many of the potential physical obstructions as possible.
Once the mechanical items are resolved, the proper cold alignment offsets
must be included to allow the machines to move into an aligned position during
normal operation. The available techniques for measuring the thermal growth
(or contraction) are discussed in the last part of this chapter. After these alignment offsets have been established, allowable target values should be identified.
These target values may be listed as acceptable variations to centerline offsets,
indicator readings, or elevation readings.
In a perfect world, the final alignment numbers would be identical to the
648
Chapter-12
Distance Between Coupling Pivot Points (Inches)
50
40
30
20
M
00
,
20
RP
00
RP
0
0,
0
,0
RP
0
,0
RP
0
,0
M
M
M
RP
RP
RP
00
00
0
0
3,
2,
10
M
8
00
RP
0
1,
6
5
4
3
2
1
1
10
20
30
40
50
Reverse Indicator Reading Alignment Tolerance (Offset in Mils)

Fig. 122 Typical Tolerance Chart Relating Coupling Span, Speed, And Allowable Offset
desired offsets. However, this is seldom attained, and alignment tolerances

are employed to define an acceptable alignment range. For instance, Fig. 12-2
displays a typical plot of coupling length versus alignment tolerance for various
machine speeds. From this diagram it is clear that longer couplings have greater
misalignment tolerances. Furthermore, low speed machines are more tolerant to
misalignment than high speed units. Both of these conclusions make good practical sense, and no one will question the need for better alignment on higher speed
machines with shorter couplings. The actual alignment tolerances will vary in
accordance with the type of coupling, the machinery configuration, the applicable
engineering specifications, the sophistication and accuracy of the measurements,
plus the experience level of the people performing the alignment work. In virtually all cases, the more experience and knowledge contained within the alignment crew, the better the final alignment job.
Clearly, there are a variety of tools and techniques available to perform the
position, bore, and shaft alignment tasks. Excellent references are available from
the vendors of optical and laser alignment equipment. Due to the variations
between lasers, and the continuing evolution of these products, it is difficult to
supply up-to-date references in this text. However, the diagnostician should have
no problem finding current information at any trade show. The final selection of
the best approach for machinery alignment generally comes down to the experience of the individuals performing the work, plus the available hardware, and
the specific tasks to be performed. In order to provide the diagnostician with an
overview of the options, the following discussions are offered for consideration.
Optical Position Alignment
649
OPTICAL POSITION ALIGNMENT

Position alignment is a three-dimensional problem of locating machinery
with respect to stationary bench marks. This includes the initial surveying work
to establish the proper location for a foundation, the accurate positioning of a
baseplate, as well as optimization of the casing position. This type of alignment
work is necessary during new construction, and it is also required during various
types of machinery maintenance. Optical tooling is the most common type of
measurement equipment used for position alignment. However, laser tooling is
also used, and that approach will be discussed in the next section. Although
much of the position alignment work is performed by land surveyors, the tools
and concepts discussed in this text are directed at machinery alignment.
Optical alignment tooling has been used for many years, and it has been
successfully adapted to position, bore, and shaft alignment. Obviously, special
tooling is often required to perform a specific task on a particular machine. Fortunately, a wide variety of tools are readily available for almost every use. The
primary instrument for optical alignment is the optical telescope. This device
is a telescope that consists of a tube containing an objective lens, focusing lens,
reticle with cross lines or similar pattern, plus an eyepiece. The focusing lens is a
moveable element located between the objective lens and the reticle. By moving
this lens, images can be accurately focused on the reticle. The objective lens
inverts the image, and the reticle is inverted as well. The eyepiece inverts and
magnifies the image so the viewed image is correct to the human eye.
Fig. 123 Tilting Level

K&E Paragon Series
An example of an optical telescope is the Keuffel & Esser (K&E) tilting

level shown in Fig. 12-3. This device is used to establish a single plane in space
perpendicular to gravity (i.e., level). This instrument is ideal for setting elevations and checking levels and flatness. Several companies provide optical levels,
and the specifications vary between vendors, but the overall accuracies are similar. The tilting level is a telescope mounted on a four screw leveling system that
is attached to a tripod. The four leveling screws in the tilt axis are coincident
650
Chapter-12
with the azimuth axis, and this eliminates errors during the leveling process.
This telescope is equipped with an optical micrometer accurate to thousandths of
an inch up to 100 feet. The magnification automatically varies from 20x at a near
distance to 30x at infinity. Two positioning or setup levels are integral with this
instrument. The first device is a circular vial level that has a sensitivity of 10
minutes per 2 millimeters of movement. The second device is a coincidence type
level with a sensitivity of 20 seconds per 2 millimeters of movement. Combined
with an optical magnification, this allows leveling to within 1 second of arc.
More sophistication and capability in the optical telescope is obtained with
a jig transit, such as the K&E unit depicted in Fig. 12-4. This is a versatile
instrument that consists of a telescope mounted on a base that allows rotation
around both the elevation and azimuth axes. Rotation around the elevation axis
allows the user to view a vertical plane, and establish plumb lines. The azimuth
center is hollow, and the telescope line of sight is centered above this yoke to
allow the capability of viewing vertically downward. Like the tilting level, a jig
transit is mounted on a four screw leveling system to the tripod plate. The four
leveling screws in the tilt axis are coincident with the azimuth axis, and this
eliminates errors during the leveling process. Elevation leveling is accomplished
Fig. 124 Jig Transit

K&E Paragon Series
with two speed tangent screws to allow precision leveling. For setup, the jig transit is equipped with one circular vial level and one coincidence type level, much
like a tilting level. However, on a jig transit, the circular vial level is installed on
the transit base, and the coincidence level is mounted on the telescope.
Jig transits incorporate a variety of special features. Many units have a
mirror on the telescope axle to be used for collimation activities. Jig transit
651
squares have a hollow telescope axle with a semi-reflective mirror on one end.
The telescope square has a telescope installed in the hollow horizontal axle. This
cross axis telescope may be fitted with an auto-collimation unit. This enables an
optical reference to be kept from a mirror, while using the line of sight telescope.
In all cases, the optical telescope must be properly supported on a rigid
stand or tripod. Many commercial types of supports are available. This includes
rigid stationary units, plus a variety of portable stands with adjustable legs. A
quality field stand must be able to be located on loose and uneven surfaces, and
the entire assembly should have mechanical locks to prevent movement during
alignment readings. In some instances, a portable stand cannot be used due to
physical limitations such as an unstable deck or insufficient space for a tripod. In
these cases, a customized stand or support plate may be purchased or fabricated
to suit the particular configuration.
The position of any machine element in space may be quantified by mounting precision optical scales on the object. Differential optical measurements
between these scales and fixed bench marks will allow an accurate determination of the three-dimensional location of a machine element. If differential optical measurements are made between locations, the absolute (bench mark)
reference does not exist, but the user still retains the ability to check for level or
flatness of the machine surface. Obviously, there are many ways to use an optical
telescope in conjunction with a precision optical scale.
Fig. 125 Portion of Ten

Inch Long K&E Optical Bar
Scale With 20 Divisions
Per Inch
In order to extract the required accuracy from the optical measurements,

the scales must be accurately engraved, and easy to read. For example, Fig. 12-5
depicts the end of a 10 inch long K&E optical scale. The graduated lines are
black to provide good visibility against the Wyteface background. In this diagram, the graduated optical scale where the center of each pair of lines to the
next is 0.050 0.001 inches. The scales are numbered in inches, and appropriate
subdivisions. Other K&E scales are provided with four different patterns to
allow for various sight distances. These typical optical scale separations and
their associated sight distances are summarized in Table 12-1.
To accommodate various physical installations, K&E provides optical scales
in standard lengths of 3, 10, 20, and 40 inches. The scales are held in position on
the machinery with a variety of methods. For instance, Starrett produces a
weighted scale holder that is useful on nonmagnetic surfaces. A variety of other
supports are available that use magnetic bases. In some cases a scale mounting
base may be attached with epoxy cement. On critical machinery applications
where precise repetition of the mounting point is mandatory, a dowel hole and
surface spot-face are simultaneously produced with a special drill bit. A stainless
steel dowel pin is then inserted into the hole, and the target holder is mounted
652
Chapter-12
Table 121 Optical Scale Line Separations Versus Sight Distance

Scale Separation
Sight Distance
4.0 Mils
Up To 7 Feet
10.0 Mils
7 to 20 Feet
25.0 Mils
20 to 50 Feet
60.0 Mils
50 to 130 Feet
on the dowel pin. This type of installation requires more setup time for the first
set of readings, but repeatability is excellent, and future comparative reading
can be acquired much faster.
Following selection of the optical telescope, support stand, and the appropriate scales or targets, the instrument must be properly setup. Detailed instructions are provided in the instruction manuals for each device. In general terms,
the setup begins by focusing the image for the individual viewer. If the eyepiece
is not properly focused parallax may occur, and this causes the image and the
reticle to appear in two separate planes. The instrument must be leveled with
the four screw leveling system mentioned earlier. The circular vial level is the
primary indicator for this task. For telescopes used for leveling, a peg adjustment may be used to check the instrument calibration, and identify any necessary corrections. For situations where the transit must reference two points of
interest, the instrument must be bucked-in to both points. The two points of
interest may be bench marks that are established by the observer, or specific
locations on a machine. In other situations, readings have to be taken on a vertical face to check square and plumb. The jig transit may be used to establish
vertical plumb lines, or check an entire surface for plumb.
Some telescopes are equipped with an auto-collimation eyepiece (semitransparent mirror and a light). When this type of telescope is pointed at a mirror, the observer can see the reticle, and its reflected image. By moving the transit until the cross lines from reticle and the reflected image converge, the
instrument will be collimated. This adjustment of the line of sight is equivalent
to the observation of parallel light rays that are focused on the reticle. If an autocollimation eyepiece is not available, auto-reflection may be used. The procedure consists of mounting a printed target on the front end of the telescope, and
adjusting the telescope or mirror until the reticle and the reflection of the target
coincide. Auto-reflection is not as accurate as auto-collimation.
In position alignment, it is often necessary to measure the squareness of
an object. This implies that one plane is perpendicular to another. This can be
accomplished by auto-collimation from a second instrument to the axle mirror of
the primary unit. In another method, a pentaprism can be mounted on the front
of the telescope, and rotated to sweep a perpendicular plane. Alternately, a jig
transit telescope square may be set with the cross axis telescope. The cross axis
telescope can be auto-collimated or auto-reflected. Finally, the telescope can be
set to individual bench marks and bucked-in, as previously mentioned.
653
There are certainly many setup steps required for the proper use of optical
equipment. Although it may not be immediately apparent, optical readings are
often based upon small changes measured over long distances. For instance,
Table 12-2 is a conversion chart for changing angular arc readings measured in
seconds into arc lengths in Mils. From this table, it is noted that a 60 second arc
(= 1 minute = 0.017 degrees), has an arc length of 349 Mils for a 100 foot span.
Thus, a very small change in an angle will represent a significant change in linear position of an object. For this reason, optical alignment measurements must
Table 122 Conversion Of Seconds Of An Arc To Arc Length
Equivalent Arc Length in Mils
Arc Length
(Seconds)
10 Foot
Span
20 Foot
Span
30 Foot
Span
40 Foot
Span
50 Foot
Span
100 Foot
Span
0.6
1.2
1.7
2.3
2.9
5.8
1.2
2.3
3.5
4.6
5.8
11.6
2.9
5.8
8.7
11.6
14.5
29.1
10
5.8
11.6
17.4
23.3
29.1
58.2
20
11.6
23.3
34.9
46.5
58.2
116.4
60
34.9
69.8
104.7
139.6
174.5
349.1
be carefully performed to provide meaningful results. Optical measurements

should always be rechecked and sometimes rechecked. On critical machinery, it
is highly desirable to have two people obtain independent measurements to verify the accuracy of the data.
As applied to position alignment, optical measurements are primarily
directed towards obtaining accurate optical scale readings in three dimensions.
These readings may be combined with suitable bench marks to accurately determine the physical position of a point in space. This is necessary for tasks such as
locating foundations, positioning baseplates, or positioning machinery cases. The
numbers can also be used to reference flatness or level of an object. This type of
measurement is employed to insure that support feet on a baseplate are properly
positioned, and correctly machined. For an improved perspective on position
alignment, case history 41 provides an overview on a major position alignment
project dealing with a high pressure reciprocating compressor.
654
Chapter-12
Case History 41: Hyper Compressor Position Alignment

The machine under consideration represents one of four high pressure ethylene compressors in a low density polyethylene plant. High stage discharge
pressure for this machine varies between 28,000 and 35,000 Psi (depending on
product). A general arrangement of this reciprocating compressor is described in
the plan view of Fig. 12-6. This compressor is driven by a 3,500 HP synchronous
motor. This 30 pole rotor is mounted directly on the compressor crankshaft, and
it runs at a synchronous speed of 240 RPM. As shown in Fig. 12-6, the crankshaft
is supported on five main bearings, with the #1 bearing at the oil pump (outboard) end of the crankcase, and the #5 bearing supporting the motor rotor. The
compressor is a four throw unit, with auxiliary crossheads that drive two cylinders on each throw (eight cylinders total). Each auxiliary crosshead is mounted
on a separate concrete pedestal, and the motor stator is on separate sole plate.
Visual inspection revealed that the auxiliary crosshead and the motor foundations were in good condition. However, the main foundation under the crankcase had horizontal cracks at 5 to 7 feet down from the grout cap. Oil and
chemicals had impregnated the foundation cracks, which showed the foundation
winking during operation. In essence, the top of the main foundation was severed from the support base. Several foundation repairs had been tried over the
years, none of which were successful. The machine was continually plagued with
numerous mechanical problems. For instance, main bearing failures occurred at
a rate of one failure every two years. Since there are four identical compressors
with the same problems, the average failure repetition rate was one failure every
six months. This is certainly an unacceptable main bearing failure rate, and the
separated foundation was considered to be one of the obvious culprits.
Main Crosshead
Synchronous Motor
3,500 HP - 30 Pole - 240 RPM
4,600 Volts - 3 - 60 Hz
Auxiliary Crosshead
Bearing #5
Bearing #4
Cylinder 2
Cylinder 1
3rd Throw
Cylinder 7
Cylinder 8
Bearing #3
Bearing #2
Crank Case
4 Throws
2nd Throw
Plungers
Cylinder 5
1st Throw
Cylinder 6
4th Throw
Auxiliary
Crosshead
with Slipper
Cylinder 3
Cylinder 4
Bearing #1
Stationary Tie Rods
Reciprocating Drive Rods

Oil Pump
Fig. 126 Machinery Arrangement Plan For Hyper Reciprocating Compressor
655
Clearly, the main foundation under the crankcase required extensive

repairs, and that task dictated the removal of the compressor crankcase, and
main crossheads. To properly repair this foundation, the existing concrete must
be chipped out to an elevation below the cracked and oil soaked sections. Fortunately, the motor and auxiliary crosshead pedestals do not require any repairs,
and they remained intact. From a timing standpoint, the machinery train was
scheduled for a major overhaul, and all machine elements were to be removed for
complete refurbishing by the OEM. Not surprisingly, when the compressor was
shutdown for this overhaul, it was discovered that the main crankcase #2 bearing had failed. This failure was consistent with previous bearing history.
On high pressure mechanical equipment of this type, it is customary to reference or position the various machine elements with respect to the common
crankcase and associated crankshaft. During the type of extensive overhaul, it is
imperative to remove the crankcase, rebuild the foundation, and reinstall the
crankcase in exactly the same location. If the crankcase in not reinstalled in the
same place, the motor, auxiliary crossheads, piping, and other appurtenances
will not line up properly. This can cause undue difficulty, and seriously delay the
completion of the maintenance turnaround on this unit.
Since the main foundation will be removed, the compressor anchor bolts
must also be replaced. Prior to pouring a new foundation, new anchor bolts will
have to be positioned to match the compressor footing plan, plus they must be at
the correct elevation. Obviously, this type of repair project requires accurate
position alignment measurements prior to crankcase removal, plus additional
position alignment during reassembly.
The fundamental task during disassembly of this unit was to document the
actual positions of the machine elements with respect to a defined set of threedimensional bench marks. Permanent X-Y-Z bench marks did not exist, and it
was necessary to establish these fixed reference points during disassembly. Since
the compressor elements are normally referenced to the crankshaft, the crankshaft bore centerline was considered as the primary control parameter. Hence, it
is necessary to document the crankshaft bore to the new bench marks.
For purposes of consistent definition, the Y-axis will be defined as horizontal, and collinear with the crankshaft axial centerline (i.e., fore and aft). Horizontal motion (i.e., left and right) will be considered as the X-Axis. Changes in
elevation (i.e., up and down) will be defined as the Z-axis. This X-Y-Z coordinate
system will have an origin at the crankcase outboard looking towards the motor.
The readings are further identified with a traditional plus or minus () direction.
Note that this convention is different from the viewing position on centrifugal
machinery, and it is also different from the specialized plunger labeling applied
in case history 35. The current type of coordinate system is commonly employed
on reciprocating machinery alignment.
The installed Z-axis bench mark consisted of a 1/4 inch bolt on a structural
steel supporting beam. This bolt was assigned an arbitrary elevation of 100
inches. To determine crank shaft centerline elevation, slipper guide targets were
set and centered (0.25 Mils) in the #1 and #5 bearing bores. Using the precision
ground outer diameter of the target, a 40 inch optical scale was mounted on the
656
Chapter-12
top of each target, and read with a tilting level. The scale reading was corrected
by 1.125 inches to determine the centerline elevation of both end bearings. In
addition, the elevations at eight locations on the crankcase top gasket surface
were obtained with the precision tilting level.
The Y-axis centerline was referenced from the motor face of the crankcase.
A jig transit was bucked-in to this face with a 3.200 inch offset. The transit was
then sighted on the motor sole plates on each side of the crankcase, and reference
lines were scribed on the sole plates to establish the bench mark. Scale readings
were then taken from the two front anchor bolts to establish the Y-axis control
for the crankcase and the anchor bolts. This was followed by elevation readings
of the anchor bolts and flanges. For reinstallation purposes, elevation readings
were also obtained from each of the dog house (main crosshead) anchor bolts and
flanges.
Establishment of the X-axis bench marks was more difficult. Initially, an
alignment telescope was positioned in front of the #5 bearing on cone mounts,
and this telescope was bucked-in to the bore of bearings #1 and #5. An aluminum
plate was attached to the concrete wall behind the crankcase, and lines were
scribed on the plate to reference the crankshaft centerline. This bench mark provided a reference point for one end of the X-axis. Next, a jig transit was positioned 87.5 inches from the #5 bearing. It was centered on the #1 and #5 target
centerlines, plus the scribe lines on the wall mounted aluminum bench mark.
The transit was then sighted down through the hollow center to the concrete
floor, and this location was marked. The transit was then sighted to a floor location 75 inches behind the transit, and the floor was again marked.
Holes were drilled into the concrete floor, and brass plugs were epoxied in
place at each location. Both plugs were recessed below the floor surface to minimize damage. After the epoxy hardened, the jig transit was reset above the plug
at 87.5 inches from the #5 bearing. The transit was again centered on the targets
at bearings #1 and #5, plus the scribe lines on the aluminum wall bench mark.
The transit was again sighted down through the hollow center to the brass plug,
and the plug was center punched to the center of the telescope crosshairs. The
same procedure was repeated for the second brass plug 75 inches behind the
transit. These two floor mounted brass bench marks combined with the scribed
aluminum plate to uniquely define the X axis.
Based on these bench marks, the exact crankcase location was identified
and recorded. The locations of the main bench marks in the immediate vicinity of
the crankshaft are shown in Fig. 12-7. Following these initial position measurements, the machinery was removed, and sent to the OEM for rejuvenation. At
this point, the old foundation was removed past the cracks and contamination.
New rebar was installed in the retained foundation to connect the existing with
the new concrete. In addition, a precision aluminum anchor bolt template was
fabricated based upon the OEM anchor bolt locations. This template was optically positioned, and new full length anchor bolts were installed. After the
anchor bolts were tightened, and the template secured, a final set of optical position readings was taken to verify that nothing was moved during the securing
process.
657
The concrete was successfully poured, and the new foundation was prepared to accept the crank case for setting and grouting. The rebuilt crankcase
was set and leveled with the jack bolts to the top gasket surface. All 26 anchor
bolts lined up perfectly with the crankcase holes. Slipper guide targets were once
again set and centered in the #1 and #5 bearing bores. To reference the previously established bench marks, a jig transit was centered with the two brass
floor plugs, and the scribe lines on the aluminum wall plate. A second transit was
bucked-in to the motor sole plate scribe lines, and a tilting level was set to measure the crankcase elevation verses the vertical bench mark. Following this
setup of optical equipment, the crankcase was moved back to the original position. The bore alignment was checked (case history 42), and then all 26 anchor
bolts were torqued to an initial value of 50 foot-pounds to hold the machine during grouting.
After the crankcase was grouted, and the epoxy cured, the optical equipment was used to obtain position readings before and after the anchor bolts were
fully torqued. The final results of this work are depicted in Fig. 12-7. As noted,
the #1 bearing position was 5.0 Mils to the North, and 2.0 Mils lower than the
original position. The #5 bearing was 16.0 Mils North, and 20.0 Mils lower than
initially found. The crankcase face position was 3.0 Mils West of the original location, and the compressor top cover elevations varied from 1.0 to 6.0 Mils from
their initial positions. The final offsets at the motor #5 bearing were larger than
desired, but the remaining locations were judged to be quite acceptable. Additional bore alignment work was performed on this compressor, and this is discussed in case history 42. Overall, the efforts to accurately measure the
crankcase position prior to removal, and during installation, allowed a clean
reinstallation of this machinery. The final success of this work is evident from
consecutive years of smooth operation, plus the absence of bearing failures.
Rebuilt Position:
16 Mils North
20 Mils Low
Benchmark Of
Scribed Line On
Motor Sole Plates
Brass Floor Plug

Benchmarks
Rebuilt Position:
1 Mil Low
Rebuilt Position:
4 Mils High
Brg. #3
Brg. #5
Rebuilt Position:
5 Mils North
2 Mils Low
Brg. #1
3.200
75
87.5
17
41
20
Motor Sole Plate
Diagram Not To Scale
Rebuilt Position:
3 Mils West
Rebuilt Position:
3 Mils High
42
Aluminum
Benchmark
On Wall
Rebuilt Position:
6 Mils High
Fig. 127 Installed Bench Marks Plus Final Position Of Compressor Crankcase
658
Chapter-12
LASER POSITION ALIGNMENT

For many years, optical alignment was the only versatile method for positioning process machinery. Since 1975, laser alignment equipment has been
developed, and utilized in the machinery business. The accuracy of the laser has
been demonstrated in many calibration and inspection labs. As technology and
instrumentation has improved, the industrial laser has evolved as a rugged and
versatile device. The greatest advantage of a laser is the high degree of accuracy
over a long distance. Some surveying lasers have effective ranges of over 10
miles with an accuracy of 0.2 inches. With respect to machinery alignment, the
laser alignment methods are based upon the same line-of-sight principles as the
optical telescope. However, the laser beam represents a straight line in space
that can be positioned and accurately measured.
Lasers offers several advantages over optical tooling. For instance, laser
alignment systems for machinery applications typically have a range up to 150
feet, with an accuracy of 1.0 Mil. The laser beam is not subject to interpretation
like optics, and the readout displays the target offset directly. Since the readout
is a continuous display that tracks the laser beam, any movements are immediately displayed by the readout system. This allows components to be moved and
recorded with the laser. Furthermore, the laser beam is not affected by physical
sag like many mechanical measurement methods.
The laser does have several drawbacks over optics. A laser is very sensitive
to moisture, and high humidity, water vapor, steam, and heat hazes can all influence the laser beam. In many cases the beam may be bent due to water vapor.
The removal of the human factor from the data interpretation requires that the
operator be experienced, and able to recognize good versus questionable data.
From a commercial standpoint, lasers are expensive to purchase and repair. Furthermore, most lasers and accessories are not interchangeable between vendors,
and the user is committed to purchase everything from one supplier.
In virtually all cases, a calibrated laser used in its proper application with
qualified personnel will provide results comparable to that of optics. The operation and procedures of the laser are very similar to optics, and the intended
results are the same when measuring for flat, level, square, or plumb. Laser tooling consists of six major components. Specifically, this includes the laser head,
detector, beam directors, beam splitters, power supply, and readout. The stands
and holders in some cases are identical to the optical equipment. In other situations, custom holders and fixtures are needed to locate and set the laser tooling.
A laser head is the device that generates and directs the laser beam. For
machinery applications, the laser head is often a rectangular box with the laser
beam emitted from one end. Lasers are also available in tri-axial units, where
beams are directed out of two locations 90 apart. The physical size, shape, and
construction of the laser differs for each manufacture. Fortunately, some vendors
of alignment lasers build units in the shape of long tubes with the same diameter
as an optical telescope. This allows much of the traditional optical tooling to be
utilized with the alignment laser. Currently, most lasers are visible beam systems, and this eliminates the need for a beam finder. The effective working range
Laser Position Alignment
659
and accuracy of the laser depends on the specific unit, and commercial systems
typically identify an operating range between 50 and 300 feet.
The detector is the target used in laser alignment. For machinery alignment, the detector typically contains four cell units to cover the horizontal and
vertical axes. The detector has a finite range over which the beam can be sensed.
If the measurement range is not compatible with the detector, it may be difficult
to retain the beam in the detector. This is especially true during large changes
during rough alignment moves. The physical size and operation of the detector
dictates the mounting. Typically, detectors are mounted on magnetic bases or
angle plates. As noted in the optical instruments, special holders for bores and
irregular surfaces may be fabricated if the vendor does not offer such a mount.
Since a laser beam operates in a straight line, it is sometimes necessary to
use a beam director. This is a mirror and prism device that deflects the beam
by 90. This allows a vertical plane to be established, and it also provides a way
to deflect the beam around stationary obstructions. Beam splitters allow the
laser beam to be split into two beams with a 90 separation. With the proper
placement of beam splitters and deflectors, a total of 8 axes can be measured
with one laser head.
A power supply is used to drive the laser, detector, and readout. This
device converts AC or DC power to the voltage and frequency required to operated the laser. Amplifiers, if needed, are typically located in the power supply,
and many systems combine the power supply and readout into one unit.
The laser readout displays the current position of the laser beam on the
detector. Machinery alignment readouts generally show a two axis digital display with readings in the vertical and horizontal directions. The values are typically labeled as plus or minus to indicate the direction of offset. Some systems
have the ability to display offsets as position change vectors, which is quite desirable in some applications. Current systems incorporate computers to record, display, and manipulate the readings. This provides the operator with quick and
accurate results. In addition, the software allows the results to be logged and
saved. In this situation, the computer does reduce the chance for a human
recording error. It should be recognized that computers have a tendency to isolate the diagnostician from the machine, and provide a set of concise alignment
moves. This is fine when everything works properly, but if measurement or
mechanical problems appear, the computer solution may not provide sufficient
visibility to identify and correct the error.
It should also be noted that the methods behind laser alignment are the
same as optical alignment. Measuring flat, level, square, and plumb are all possible with proper placement and beam manipulation. The setting, leveling, and
zeroing of each laser will vary depending on the vendor. For instance, some units
have internal automatic levels, others have manual levels, and some lasers are
not equipped with levels. Regardless of the specific laser, the important point to
remember is that the principles and procedures for laser position alignment are
similar to that of optical alignment. In all cases, if the basic concepts are understood, the diagnostician has two viable instruments that may be applied to a
field machinery problem.
660
Chapter-12
OPTICAL AND LASER BORE ALIGNMENT

Bore alignment is the process of measuring and repositioning bore center
lines in relation to a fixed point. This type of alignment is performed on a daily
basis in many machine shops, where machine quills or boring bars are used to
indicate internal machine bores and faces. With respect to process machinery,
bore alignment is also performed with optical tooling, laser instruments, and
wire alignment equipment. These three methods can be applied in the shop and
in the field on virtually any type of equipment.
The optical tooling used in bore alignment is similar to devices used for
position alignment. Telescopes, targets, and supports are required, and many of
the tools and procedures used for position alignment are directly applicable to
bore alignment. Optical tooling for bore alignment is generally expensive, and
specific fixtures are often required for each individual machine. These measurements are subject to environmental conditions such as vibration and heat hazes,
plus the resultant data is subject to interpretation. As with any technical topic,
the understanding and interpretation of results improves with experience, and a
sound working knowledge of the equipment and procedures.
The heart of an optical bore alignment system is the alignment telescope
that is used to set a single line of sight. The K&E telescope shown in Fig. 12-8 is
similar to that of a transit telescope. However, this model offers specific features
required for bore alignment. The telescope barrel has been hard chromed, and
ground to a diameter of 2.2498 +0.0000
0.0003 inches. This provides a precision surface to
mount the scope on a cone mount, or a sphere and cup mount. The telescope can
be focused from zero to infinity with a magnification range from 4X to 46X. The
alignment scope is also equipped with two micrometers for horizontal and vertical movement. The micrometer numbers are colored black and red on opposite
sides of zero to indicate the direction of displacement. Alignment telescopes can
be fitted with auto-collimation eyepieces as well as right angle eyepieces. Attach-
Fig. 128 K&E Alignment

Telescope Supported On
Cone Mounts For Borescope Measurements
Optical and Laser Bore Alignment
661
ments such as optical squares, auto-reflection targets, and angle reading attachments are also available.
The telescope cone mount shown in Fig 12-8 is a four cone mounting system that allows precise adjustment of the telescope in the vertical and horizontal
directions. The mount consists of front and back twin cone assemblies with top
retaining strips. The support cones are precision devices with threaded shafts
that screw into the base. The individual cones are identified as 1 through 4. As
viewed from the eyepiece end of the telescope, cones 1 and 3 are on the left side,
and cones 2 and 4 are mounted on the right side. Furthermore, cones 1 and 2 are
at the objective lens end of the telescope, whereas cones 3 and 4 at located back
at the eyepiece end. If all four cones are initially set to the same height, and in
the center of the total threaded length, the telescope mount will have the maximum adjustment range. For reference purposes, the movement characteristics of
the telescope with specific cone rotations are identified in Table 12-3.
Table 123 Adjustment Of Support Cones For Specific Optical Telescope Movements
Operation
CW Cone
Rotation
CCW Cone
Rotation
Raise
1,2,3,4
Lower
1,2,3,4
Move Right
2,4
1,3
Move Left
1,3
2,4
Aim Up
3,4
1,2
Aim Down
1,2
3,4
Aim Right
2,3
1,4
Aim Left
1,4
2,3
As discussed in the position alignment section, optical targets are used to

establish a line of sight with the telescope. For bore measurements, the targets
are also used to determine offset values with respect to the line of sight. Targets
used for bore alignment typically contain paired lines in both the horizontal and
vertical directions. The construction varies from glass or plastic printed targets,
to wire strung targets. These targets are precision ground on the outer diameter,
and the crosshairs are centered to the outer diameter within 1.0 Mil. For bore
alignment, two distinct types of target holders exist. Specifically, there are threelegged adjustable holder that will accommodate both glass and wire targets. The
leg extensions allow the holder to be used on bore diameters ranging from 5.5 to
21.5 inches. The targets are centered in the bore with a Y centering device. This
works well for rough centering, but can cause some difficulties for final setting.
Another target holder configuration is shown in Fig. 12-9. Bayshore Surveying in
Deer Park, Texas, produces this target holder, which adapts to most applications.
662
Chapter-12
Fig. 129 Bore Target

Holder For Slipper Guide
By Bayshore Surveying
The holder accepts all standard glass, and wire targets. The device is designed
with a solid support base, and two vertical rails. The cross rail assembly is
adjustable on the vertical rails, and the target assembly is mounted on the cross
rail with position adjustment screws. Thus, two position adjustment screws control the vertical and horizontal movement. The system is extremely versatile,
and can easily be modified to fit in special locations.
For accurate bore measurements, the targets must be located in the center
of each bore. To accomplish this task, Bayshore Surveying manufactures a bore
sweeper. This device is a brass fixture that clamps to the outside of the target,
and it supports an adjustable arm with a dial indicator for sweeping the bore.
The bore may be indicated, and the target centered within 0.25 Mils. Ideally, the
bore sweeper is set to indicate the bore in the same plane as the crosshairs. Also,
if the bore is elliptical, the indicator readings between horizontal and vertical
will not be equal. However, the target will be centered in the bore as long as the
opposing readings are equal (i.e., left equals right, and top equals bottom).
As a cautionary note, an improperly set target can produce significant
errors that may not be obvious. Before the equipment is installed, all targets and
holders must be inspected for damage. Glass targets should be clean, and wire
targets should be straight and free of kinked or broken wires. In addition, the
targets should be calibrated on a routine cycle by a qualified shop.
After the targets have been positioned in each bore, it is necessary to set the
alignment telescope. The telescope line of sight must be set to two of the target
locations (i.e., two points determine a line). Each individual machine will dictate
which two bores should be used as the zero points. Many times centrifugal
machines will set zero on the two bearing bores while the seal fits are checked. In
the case of an engine or reciprocating compressor, the two most outer bores are
typically set as zero. The procedure for setting the alignment telescope is dependent upon the type of scope and mount (e.g., Fig. 12-8). The physical setup may
require the use of a portable tripod, or the telescope assembly may be bolted
Wire Bore Alignment
663
directly to the end cover of the machine. In either case, once the telescope has
been centered on the targets, the offsets at each point of interest (i.e., other
bores) may be directly read, recorded, and physically adjusted as required.
The laser is well suited for bore alignment applications, and this has been
common practice for many years. The instrumentation is the same basic equipment previously discussed in the laser position alignment section of this chapter.
Once again, the bore alignment procedure for lasers is based upon the optical
method. The specific procedures are unique to the laser system supplied by each
vendor. It should be mentioned that most laser system targets are not indicated
to the bore like the optical targets. In laser bore measurements, the detectors are
set to the laser head alignment, and the detectors are then rotated 180 and
zeroed. The detectors are then moved from location to location for the individual
offset readings. Like any optical device, heat hazes and steam leaks will influence the laser beam, and they should be removed from the line of sight.
WIRE BORE ALIGNMENT

Wire alignment has been utilized for many years to check engine, compressor, and cylinder bores. This technique is adaptable to any situation where wire
can be strung, centered, and read. The wire alignment equipment is simple, compact, and readily available at a reasonable cost. The approach is quite flexible for
checking various machine components. Once a wire is strung and set in a cylinder bore, any gland, shoulder, or surface can be measured with the same setup.
Like any measurement system, wire alignment has limitations. Even
though the wire is tensioned, it still sags due to gravity. For spans over 35 feet
(420 inches), the sag becomes considerable. The sag must be determined and
compensated for in every wire alignment. The required sag calculations are simple, but due to the repetitive nature, unconscious errors can occur. In all cases,
the wire must be straight, and free from kinks or corrosion. Wire vibration due to
inadvertent twanging of the wire, or induced vibration from adjacent machines
will produce measurement errors. The wire is also sensitive to thermal changes,
and mounting brackets will grow faster than the machinery case when exposed
to sunlight. In addition, wire alignment requires direct physical access to all
measurement areas. Small bores, long and closed spans, and tight locations can
make the process difficult, and in some cases impossible.
Wire alignment requires the use of a wire, anchors, oscillator, micrometer,
and a tensioner. In most commercially available wire alignment kits, the primary
components are the same, with some variance in wire diameters and the tensioner. For consistency, the following discussion will be based upon the hardware
contained in a standard Dresser-Rand wire alignment kit.
In the alignment kit, a tensioner is used to preload the wire. The two common means of tensioning the wire are by using a spring or a dead weight. The
weight method offers the advantage of constant load; however, the weight is
heavy and often cumbersome to hang in the field. The spring design used by
Dresser-Rand provides a compact package that accommodates most applications.
664
Chapter-12
The drawback to the spring tensioner is that it must be periodically calibrated to

insure proper wire tension.
The piano wire used by Dresser-Rand has a diameter of 18.0 0.3 Mils.
During setup, the wire must be inspected to insure that it is both clean and
straight. If any physical problems are discovered with the wire, it should be discarded and replaced. The wire is supported between two anchors, one fixed and
one adjustable. The anchors are designed to retain the wire and allow position
adjustment of the wire within the bore. The anchor and wire must be electrically
insulated from the machine casing. The fixed anchor is designed to clamp and
retain, and it has four adjustment screws to allow positioning of the clamp in the
vertical and horizontal directions. The adjustable (or tension) anchor clamps the
wire, provides vertical and horizontal adjustment, and it contains the calibrated
spring used to tension the wire. For the following discussions, the Dresser-Rand
tensioner is presumed to be calibrated to 60 pounds of tension.
Physical alignment measurements are obtained between the wire and the
bore with standard rod style inside micrometers such as the Starrett 124.
This micrometer configuration allows variable length rods to be inserted in the
head to reach various diameters. These straight extension rods can be fabricated
from 1/4 or 5/32 inch drill rod. If different diameters are to be measured, it is convenient to use several micrometers, and have one mic setup for each diameter.
The inside micrometer is used to measure the distance between the stationary bore, and the tensioned wire. Obviously, the micrometer will have a solid surface on the bore. The trick to wire alignment is to measure the distance from the
bore to the wire without deflecting the wire. One way to accomplish this task is
by connecting a low power electrical oscillator between the machine case and
the insulated wire (usually with alligator clips). During the measurement process, the micrometer completes the circuit between the wire and ground, and an
audible tone is generated through a headset connected to the oscillator. The
oscillator may be supplied with volume and frequency controls to allow adjustment of the sound to a level compatible with the surroundings. The same information may be obtained by connecting a digital multimeter (DMM) between the
wire and the machine case. In this technique, when the micrometer stem
approaches the wire, the circuit is completed, and the continuity checker on the
DMM emits an audible tone. Using a DMM eliminates the oscillator and the
headset, but the resultant tone may not be loud enough in a noisy environment.
Regardless of the method used to identify circuit completion, the wire must
be stretched between fixed and adjustable mounting points. The wire must be
centered at each reference bore (points of zero center), and the sag at each measurement point determined. To establish the gravitational wire sag, it is necessary to obtain accurate span dimensions between the reference points of center,
and the distance from the center of the wire to each measurement point. Obviously, the wire will deflect as a catenary curve, and the maximum deflection will
occur at the midspan. For the Dresser-Rand system employing 18 Mil piano wire
with a 60 pound tension, the wire will sag in as shown in Fig. 12-10 (rendered
from referenced Dresser-Rand service manual5). This diagram covers an overall
Wire Bore Alignment
665
25
Wire Sag (Mils)
20
15
10
0
-200
-150
-100
-50
50
100
150
200
Total Wire Span (Inches)

Fig. 1210 Sag of 18 Mil Piano Wire With 60 Pound Tension - Original by Dresser Rand
range of 400 inches (200 inches from the center of the sagging wire). If a wire is
stretched for the total length of Fig. 12-10, the center or midspan deflection will
be 24 Mils. This is an appreciable sag, and this physical characteristic restricts
wire alignment to moderately short spans.
As an example of wire sag, consider the bearing spacing shown in Fig. 12-7
in case history 41. On this machine, the overall span between the end bearings is
120 inches (=42+20+41+17). Thus, the range on Fig. 12-10 would be 60 inches
(=120/2). For determination of wire sag at the #2 bearing, the axial location of
this bearing from #1 bearing is 42 inches. The above curve can now be read at the
location of the #2 bearing of 18 inches (=60-42). The wire sag at 18 inches is
about 0.1 Mil, and the value at 60 inches (#1 bearing) is about 2.1 Mils. Thus the
differential wire sag between the center point at #1 bearing and the measurement point at #2 bearing is approximately 2.0 Mils (=2.1-0.1). This 2.0 Mil correction would be applied to all of the vertical wire alignment data at this bearing.
The sag can also be calculated from the following Dresser-Rand equation.
Y bet = 0.0006 D { S D }
where:
Ybet
D
S
0.0006
=
=
=
=
(12-3)
Wire Sag Between Centers at Distance D from the Point of Center (Mils)
Axial Distance from the Point of Center to the Measurement Point (Inches)
Total Axial Span of Wire Between Points of Center (Inches)
Sag Constant for 18 Mil Diameter Wire with 60 Pound Tension (Mils / Inch2)
5 Gas Engine and Compressor Field Service Manual - Section 16 - Drawing R19307A, (Painted
Post/Corning, New York: Dresser-Rand, 1986), pp. 1-12.
666
Chapter-12
If the data from the previous example is inserted into equation (12-3), the
span S would be 120 inches, and the distance from the zero point at #1 bearing to
the #2 bearing D would be 42 inches. The sag is calculated as follows:
Y bet = 0.0006 D { S D } = 0.0006 42 { 120 42 } = 1.96 2.0 Mils
This is the same result previously obtained with the graphical solution
from Fig. 12-10. In all cases, the calculated sag values should be compared to the
graphical approach, and the answers should match within 0.25 Mils. Sag outside
of the points of zero center may also be obtained by plotting the machine dimensions on Fig. 12-10. The graph is read the same as before, with the exception that
the sag is now opposite to the direction of gravity. This sag (actually a rise above
the zero point) can be calculated for the locations outside the points of center
with the following Dresser-Rand equation:
2 S 2
Y out = 0.0006 d ----
2
where:
(12-4)
Yout = Wire Sag Outside of Centers at Distance d from the Center of Span (Mils)
d = Axial Distance from Center of Span to the Measurement Point (Inches)
Note that the reference points of center always have zero sag. The wire sag
is determined with respect to these zero sag points. For locations inside the
points of zero center, the sag value must be added to the bottom vertical reading.
Conversely, for locations outside of the points of center, the sag must be subtracted from the bottom vertical readings. Horizontal readings are not influenced
by sag, and do not require correction. Furthermore, if the micrometer readings
are obtained with calibrated stems, the distance readings are dimensionally
accurate values. However, if the readings are taken with uncalibrated drill rods
in the micrometer, the resultant values are not actual lengths, and they are only
suitable for differential comparisons (i.e., right versus left).
Once the wire has been centered, and the sags determined, the bore readings may be taken. As always, physical orientation must be clearly identified.
Most turbomachinery trains are viewed from the driver to the driven, which
clearly establishes left and right. Engines and reciprocating machines are typically viewed from the oil pump end. This may be changed, but the direction of
view must be clearly identified on the alignment documentation.
After all data has been recorded and verified, the wire sag values must be
applied to each measurement location. The vertical offset of a full bore is one half
of the difference between the corrected top and bottom readings. Naturally, the
direction of offset is governed by the larger number. Thus, if the top value is
greater than the bottom value, the bore is higher than the established centers. In
the case of a half bore, the bottom readings are corrected as before, and the bore
offset is the difference between the bottom corrected value and the bottom reading of the center points (presumably zero). In the horizontal direction, the bore
offset is one half of the difference between the left and right readings. The final
bore alignment results may be presented in a graphical or a tabular format.
Wire Bore Alignment
667
Case History 42: Hyper Compressor Bore Alignment

The removal and replacement of the high pressure reciprocating compressor previously discussed in case history 41 required extensive position alignment. In addition, it is mandatory to obtain proper bore alignment of the five
main bearings. Since bore alignment addresses the offset between main bearings, this type of alignment is critical to the life and successful operation of the
machine. For the hyper compressor from case history 41, the bores should be
aligned to a tolerance of 1.0 Mil step between each of the main bearings.
The crankcase bore alignment was checked prior to the removal of the
crankcase to establish a pre-overhaul base line. The initial bore alignment
revealed steps between bearings that varied from 1.0 and 1.5 Mils. In addition,
the bores were out of round (elliptical) and oversized. As previously mentioned,
during disassembly the #2 main bearing was found to be broken.
A machine of this size and construction may easily be twisted or distorted
during installation. Once the machine has been grouted in place, major alignment movements are impossible. Due to the importance of the bore alignment,
both wire and optical alignment measurements were performed. Slipper guide
targets were set into all five bearing bores. The targets were centered to within
0.25 Mil, and the alignment telescope set on cone mounts at the #5 bearing. The
telescope was bucked-in to the #1 and #5 targets, and the offsets at bearings #2,
#3, and #4 were read. These readings were taken prior to grout, and they were
recorded and plotted. The optical equipment was removed, and the wire alignment equipment was installed. The wire readings were found to be within 0.5
Mils of the optical results.
After the crankcase was grouted, and allowed to cure, the crankcase was
ready for the bore alignment checks. Optical tooling was again installed and the
measurements after grouting were compared to the pre-grout values. There was
no significant distortion of the crankcase due to the grouting, and optical equipment was then used to check the bore while the frame anchor bolts were hydrauTable 124 Summary Of Final Wire Bore Alignment Measurements And Vertical Sag
Corrections All Measurements Shown With Consistent Units Of Mils.
Bearing Number
Parameter
72.0
70.5
70.0
70.5
72.0
2.0
2.0
1.0
72.0
72.5
72.0
71.5
72.0
0.5 Low
0.5 High
Right Horizontal
72.0
71.5
72.5
72.5
72.0
Left Horizontal
72.0
72.5
72.5
72.0
72.0
0.5 Left
0.25 Right
Bottom Vertical
Wire Sag
Corrected Vertical
Vertical Offset
Horizontal Offset
668
Chapter-12
lically torqued to 1,400 foot pounds. The anchor bolt torquing was performed in
several steps, and the bore alignment was continually monitored. Once all
twenty-six frame anchor bolts were torqued to their final values, a set of alignment readings were obtained and recorded. The optical equipment was then
removed, and the wire alignment equipment was strung and read. The final set
of wire alignment readings are shown in Table 12-4. The top four rows of this
summary table display the vertical wire bore readings. Since this was a half bore
measurement, the bottom vertical readings were obtained, and corrected for sag.
The corrected vertical readings were then compared with the end points, and the
resultant differentials listed as the vertical offsets. The horizontal offsets in
Table 12-4 were obtained by one half of the difference between the horizontal
readings. Again, wire sag will not appreciably influence these values.
Finally, the results between the optical alignment and the wire alignment
measurements are presented in Table 12-5. It is noted that these offset measurements agree within 0.5 Mils at all locations. The tolerance of 1.0 Mil maximum
per step between bearings was met, and there was good confidence that the main
bearing bores were properly aligned. Clearly, both techniques provide accurate
dependable results for bore alignment. The optical alignment is quicker and easier to use during the setting, adjusting, and torquing. The wire may have been
used during some of those operations, but would have been considerably slower.
Certainly, laser measurements could also be employed for this work, and similar
results would be anticipated. Finally, it should be restated that the overhaul on
this high pressure reciprocating compressor was quite successful, with extended
run times, and the elimination of the periodic bearing failures.
Table 125 Final Bore Position At Five Main Bearings On High Pressure Reciprocating
Compressor Crankcase Comparison Between Optical And Wire Alignment Techniques
Measured Bore Offset (Mils)
Optical Measurements
Wire Measurements
Bearing
Number
Vertical
(Elevation)
Horizontal
(Plan)
Vertical
(Elevation)
Horizontal
(Plan)
0.5 Low
0.5 Low
0.5 Left
0.5 Low
0.5 Right
0.5 High
0.5 Right
0.5 High
0.25 Right
Shaft Alignment Concepts
669
SHAFT ALIGNMENT CONCEPTS

Shaft alignment is the most common form of machinery alignment practiced and discussed in industry. The general topics of discussion range from
alignment related failures, shaft alignment methods, plus modern tools and
instruments available. Study after study has ranked shaft misalignment as the
primary cause of rotating equipment failures. Hence, shaft alignment is commonly performed on virtually all rotating equipment from small pumps to massive turbine generator sets. Shaft alignment is the final step to aligning a piece
of rotating equipment. As discussed in this chapter, machinery alignment begins
with the position of baseplates, sole plates, and the initial location of machinery
casings. As part of this installation, the support members are checked to verify
proper level and flatness. Next, bore alignment is employed to position all critical
elements within a machine casing. This may be achieved by sweeping the
machine internals with a bearing mounted mandrel, or by one of the measurement techniques discussed in the previous section.
Many of the techniques and instruments used for position and bore alignment are directly applicable to shaft alignment. However, before addressing the
specific tools and procedures for shaft alignment, it is meaningful to understand
the principles of shaft alignment. In essence, shaft alignment is the placement of
two shafts in a coaxial position. Ideally, in a perfect alignment, the centerline of
one shaft would be coincident with the centerline of the other shaft. In reality
this type of perfect shaft alignment is seldom, if ever achieved. Due to the influence of variable thermal conditions, manufacturing tolerances, assembly clearances, runouts, and different operating loads a perfect alignment is often
considered to be an acceptable alignment that falls within the misalignment tolerance of the coupling. Hence, shaft alignment really is the final positioning of
the machinery in a cold condition to compensate for the machinery full load condition, and allow for a hot running alignment that is within the coupling misalignment tolerance.
Interestingly enough, machines such as gas turbines may have a large cold
offset to allow for thermal growth as the turbine reaches full operating temperature. In many units it is common to encounter cold offsets of 250 Mils or more.
With this type of cold offset across a coupling, it is clear that the machinery train
starts up with a significant misalignment. The unit then moves into an acceptable alignment state as the machines reach thermal equilibrium. In more than
one case, a coupling has been subjected to serious distress due to extended operation at presumably safe slow roll speeds.
To appreciate the requirements and problems associated with shaft alignment, it is reasonable to begin with a basic understanding of the three fundamental types of shaft misalignment. This includes a simple parallel offset
between shafts, and angular offset of one shaft to another, plus a combination of
the two offsets into a parallel plus an angular offset. For demonstration purposes, consider the example of parallel misalignment shown in Fig. 12-11. In this
diagram, two shafts are shown to be perfectly parallel. That is, the centerline of
one shaft is parallel to the centerline of the other shaft. The misalignment occurs
670
Chapter-12
0
-10
0
-10
+10
-20
+10
+20
Fig. 1211 Parallel Shaft

Misalignment
10 Mils
when the centerline of the B shaft is dropped by 10 Mils below the centerline of
the A shaft. If reverse dial indicators are mounted across these two shafts, the
resultant sweep readings are shown at the top of Fig. 12-11. Reverse dial indicators will be discussed later in this chapter, but the immediate comment on Fig.
12-11 might be: if the shafts are displaced vertically by 10 Mils, how come the
indicators show a 20 Mil change? Initially, this might be a difficult concept to
grasp. However, the dial indicator readings are directly explainable.
Specifically, assume that a pair of dial indicators are mounted on the A coupling, and they are indicating on the B coupling. Assume that both shafts are
perfectly aligned, and that one indicator is mounted on top at the 12 oclock position, and the other dial indicator is mounted on the bottom at 6 oclock. Further
assume that both indicators are set to zero with the shafts perfectly aligned. If
shaft B is lowered by 10 Mils as shown in Fig. 12-11, the top dial indicator would
show a reading of -10 Mils indicating that the stem was moving away from the
indicator. Simultaneously, the bottom indicator would show a reading of +10 Mils
indicating that the stem was collapsed, or moving into the indicator. Therefore,
the total change sensed by both indicators would be +20 Mils {=+10-(-10)}. Looking at this move another way, if the top indicator that is reading -10 was reset to
zero, and the A shaft rotated by 180, the resultant reading at the 6 oclock position would be +20 Mils. This simple example identifies a fundamental rule in
shaft alignment as expressed by the following equation:
Bottom Top
Vert o = --------------------------------------2
where:
(12-5)
Verto = Vertical Offset (Mils)

Bottom = Indicator Reading at Bottom of Sweep (Mils)
Top = Indicator Reading at Top of Sweep, Usually set to Zero (Mils)
The same argument may be applied to the horizontal offset as follows:

Right Left
Horiz o = ----------------------------------2
where: Horizo = Horizontal Offset (Mils)
Right = Indicator Reading at Right Side of Sweep (Mils)
Left = Indicator Reading at Left Side of Sweep (Mils)
(12-6)
Shaft Alignment Concepts
671
The next common type of misalignment is the angular condition where the
centerline of the B shaft intersects the center line of the A shaft at the coupling.
This condition is shown in Fig. 12-12, and the associated reverse indicator readings are presented at the top of the diagram. Since the shaft centerlines intersect
at the A coupling hub, the dial indicator on the B coupling does not detect any
change during a sweep. However, the indicator mounted on the A coupling
reveals that the B coupling centerline is low by 10 Mils (=TIR/2). The only way
that both shafts can coincide at the A coupling, and simultaneously be 10 Mils
low at the B coupling is the angular misalignment condition of Fig. 12-12.
0
0
0
0
+10
+10
+20
10 Mils
Fig. 1212 Angular Shaft

Misalignment
The third type of shaft misalignment is the combination of angular plus

parallel misalignment as depicted in Fig. 12-13. In this condition the shaft centerlines cross at any point other then the coupling faces. This is a much more
common situation where adjustments in offset and angularity are required. For
comparative purposes, the reverse indicator readings across the span are again
shown at the top of Fig. 12-13. A casual observation of these readings might lend
one to conclude that the offset is a simple parallel offset as previously shown in
Fig. 12-11. However, the key element in this combination of parallel and angular
offsets is the fact that both indicators show a value of +20 Mils. Whereas on
Fig.12-11, one indicator read +20 Mils and the other indicator showed -20 Mils.
This subtle distinction of polarity on the dial indicator reading makes a significant difference in the type of misalignment to be addressed.
Once again the need to properly document plus and minus signs and maintain a fixed viewing position of the alignment job is reinforced. In all cases, if the
machinery diagnostician starts changing viewing directions, right becomes left,
and vice versa. Basically, the entire alignment project crumples into disarray.
0
+10
0
+10
+10
+20
Fig. 1213 Combination

Parallel And Angular Shaft
Misalignment
+10
+20
10 Mils
672
Chapter-12
Obviously the previous three cases represent the most simple types of shaft
misalignment. Process machines seldom display misalignment in only one plane.
In most instances, the diagnostician has to deal with a complex offset in both the
vertical and the horizontal directions. In an effort to minimize these offsets, it
often makes sense to place the machines in a rough aligned position. This can be
accomplished using a straight edge or precision scale, and generally will bring
the alignment close enough to begin the precision alignment.
Most shaft alignment techniques use dial indicators and indicator brackets.
These brackets are attached to a machine shaft or coupling hub. A dial indicator
is then mounted on the end of the bracket, and it is positioned to indicate the
mating hub or shaft. Brackets are constructed of everything from solid re-bar to
hollow aluminum tubes. Unfortunately, as rigidity increases so does the bracket
weight. The diagnostician should strive for brackets that provide a good compromise between high rigidity and low weight. In all cases, the indicator bracket
will sag due to the weight of the bracket plus the indicator. For short distances
the sag is minimal and repeatable. For long coupling spans the bracket sag can
become both excessive and unpredictable. Regardless of the span and bracket
size, the sag must be checked and documented prior to use. The sag value must
be recorded and used to correct the alignment readings. To check the indicator
bracket sag the following procedure may be used.
1. Determine where brackets will be mounted on the equipment to be aligned.
2. Measure and record the length between alignment planes (bracket span).
3. Attach indicator(s) and bracket to a lightweight rigid pipe (mandrel) at the
same position as they will be mounted on the machine.
4. Zero the radial indicator, and lightly tap the indicator face to check for stability. If the needle does not return to zero, check and retighten all clamps.
If the indicator does not re-zero, replace the indicator and/or bracket.
5. Pick up the assembly and hold it in front of your body.
6. Verify that the radial indicator is still reading zero. If the indicator has
moved the mandrel is flexing, and should be replaced with a rigid unit.
7. Rotate the assembly to the 6 oclock position, and hold it over your head so
the bracket and indicator(s) are hanging towards the ground.
8. Read the radial indicator and record this value.
9. Next, hold the assembly at the 3 or 9 oclock position and read the radial
indicator. The value should be one half of the sag at the 6 oclock position.
10. Return the assembly back to the 12 oclock position and re-check zero.
In specialty situations with large or heavy brackets, it may not be possible
to physically lift the assembly. In this case, the assembly may be set in V-blocks,
rollers, or the mandrel may be positioned in a lathe for the sag test. Regardless of
the physical setup, the indicated bottom value is twice the bracket sag. This is
due to the fact that the bracket sags both at the top vertical 12 oclock position,
as well as the bottom vertical 6 oclock position. Hence, the overall bracket sag
may be obtained by dividing the bottom reading by 2. This relationship is an
Rim and Face Shaft Alignment
673
obvious extension of equation (12-5), and it may be stated as:

Bottom Reading
Bracket Sag = ------------------------------------------------2
(12-7)
This expression requires that the indicator is zeroed at the top 12 oclock
position, and the bottom reading is obtained with the entire assembly (mandrel,
bracket, and indicator) inverted by 180. This bracket sag must always be a negative number. If the dial indicator displays a positive value, the mandrel is probably more flexible than the indicator bracket and it should be replaced.
It should also be mentioned that a bracket counterweight may be applied to
balance out some of the overhung weight associated with the dial indicator(s).
The position and size of the counterweight may be optimized during the bracket
sag test. Generally, if bracket sag is less than 20 to 25 Mils, an indicator counterweight is not necessary. As always, the repetitiveness of the data may carry more
significance than the actual magnitude and the diagnostician must be aware
of the specific measurement situation, plus the associated trade-offs. From a utilization standpoint, the bracket sag correction will be demonstrated during the
following discussions about rim and face, plus the reverse indicator sections.
RIM AND FACE SHAFT ALIGNMENT

Rim and face shaft alignment is one of the oldest and most common forms
of aligning two shafts. This method measures the shaft offset by indicating one
shaft with respect to the rim and face of a mating shaft. A typical machinery configuration subjected to this type of alignment is shown in Fig. 12-14. The rim dial
indicator readings determine vertical and horizontal offsets, and the face readings identify angular misalignment. The face readings may be obtained with a
dial indicator. Alternatively, inside micrometers may be used to measure the distance between coupling hubs. On couplings with close axial clearances, taper or
feeler gauges may be used to measure the gap between faces. Although this
method has lost some appeal as newer techniques were developed, rim and face
alignment is still used in the following applications:
Trains where one shaft cannot be rotated during the alignment process.
Machines with coupling hubs that are axially close to each other.
Machines that have large diameter couplings (i.e., the coupling diameter is
much greater than the coupling span).

Small general purpose machines that are typically less than 5 HP.
Rim and face readings provide a good visualization of the relative positions
of two shafts. However, this technique has several disadvantages, and many
applications require reverse indicator measurements (to be discussed in the next
section). With respect to the current topic, the diagnostician generally has to contend with the following limitations associated with rim and face alignment.
674
Chapter-12
Coupling hub runout will induce an error into the alignment readings. If
runout is excessive or inconsistent, it can be very difficult to compensate.

Machines with sleeve bearings can have a face reading error due to rotor
axial float. It is necessary to locate the rotor in a fixed axial position for
each sweep. On machines with a large axial floats, such as motors with
sleeve bearings, some type of axial stops must be employed.
Generally, the coupling spool piece must be removed.
Provides marginal accuracy on units with small diameter couplings and/or
long spans (i.e., coupling diameter is much less than the coupling span).
Any indicator face sag can be difficult to accurately compensate, and this
may adversely influence the angularity corrections.
On long spans, the indicator brackets can become very complex, and flexible.
Along with the standard pre-alignment considerations discussed earlier in

this chapter, rim and face alignment requires several additional steps. Specifically, the coupling spool piece must be disconnected, so that one rotor will remain
stationary while the other rotor is rotated. The coupling spool also interferes
with the face readings, and in most cases it should be removed.
Rim and face coupling hub runouts must be documented on the stationary
unit. If the shaft in the stationary machine can be turned, it is a simple matter to
mount a dial indicator on any fixed mechanical element, and measure the coupling hub runout (every 90). However, if the rotor of the stationary machine cannot be rotated, the radial hub runout can be determined by a dial indicator
mounted on the shaft that can be turned. In this situation, the rotatable shaft is
turned in 90 increments, and readings are obtained from the stationary coupling hub. The dial indicator is then repositioned to the shaft of the stationary
machine, and a second set of radial readings are obtained. Keep in mind that the
bracket sag may change with this move, and must be rechecked. The difference
between the shaft and hub readings will be the radial coupling hub runout
(assuming minimal shaft runout). Excessive radial runout (greater than 2.0
Mils) may be indicative of other problems with the machinery, and this should be
investigated and corrected prior to any alignment activities. Similarly, excessive
face runout of greater than 0.5 Mils should also be queried and rectified.
The tools required to perform rim and face shaft alignment include the dial
indicator and some form of indicator bracket. The dial indicator is a precision
mechanical distance measurement device consisting of a face, plunger, and body.
The circular face has a graduated scale with an indicator needle. For shaft alignment applications, the device will typically read in graduations of 1.0 Mil per
division. Indicators are also available in 0.5 and 0.1 Mil steps. The graduated
face contains a zero point with circular scales progressing in a positive clockwise
direction, and a negative counterclockwise direction. As the plunger is
depressed, the needle moves in a positive clockwise direction. Similarly, as the
plunger is extended, the needle moves in a negative or counterclockwise direction. In actual use, the plunger is placed in direct contact with the surface to be
measured, and the indicator face is rotated to the zero point. Subsequent mea-
675
surements may then be identified with plus and minus signs to define the direction of plunger travel. This polarity information identifies the convergence or
divergence of the indicator with respect to the measured surface.
An inclinometer is a shaft mounted level that displays degrees of rotation.
This device is used for determination of true horizontal and vertical locations. In
many cases, a pair of perpendicular bubble levels attached to a magnetic base
may be mounted on the coupling face. By observation of the two bubbles, the
shaft may be rotated in fairly concise 90 increments.
A strap wrench or chain wrench is typically used to rotate the moveable
shaft during alignment. With very heavy rotors, a specialized cheater bar with
pins to fit into the coupling holes may be constructed. As a precautionary note,
the shaft must never be turned by twisting the dial indicator bracket. Application of torque at the bracket will generally invalidate any useful alignment data.
In addition to the standard millwright hand tools and measurement devices, the
diagnostician should address the tool requirements for actually moving the
machinery. These tools will vary between jobs, and options such as slings and
chainfalls, hydraulic jacks, pry bars, jack bolts, or jack screws are acceptable
methods of moving equipment. Hammers and wedges should not be used to move
equipment, and hydraulic jacks should be very carefully applied. More than one
machine has been damaged due to excessive force from a hydraulic jack.
In order to convert the rim and face measurements to physical moves of the
machinery, similar triangles are established between the dial indicator readings
and the machinery arrangement. Specifically, Fig. 12-14 describes a typical setup
of rim and face dial indicators, plus the required support dimensions. The coupling diameter observed by the rim indicator is identified as the A dimension.
The distance from the face indicator plane to the center of the near foot is identified as the B dimension. Similarly, the C dimension defines the distance between
the far foot and the face indicator plane. Based upon these distances, and the rim
Rim Indicator
Viewing Direction From

Coupling Diameter
at Face Indicator (A)
Fixed to Moveable
Face Indicator
Far
Foot
Moveable
Machine
Near
Foot
Fixed
Machine
Distance to
Near Foot (B)
Distance to Far Foot (C)
Fig. 1214 General Arrangement For Rim And Face Shaft Alignment
676
Chapter-12
and face dial indicator readings, the two shafts may be aligned. Since this is such
an important activity, the specific details and requirements to check, document,
and correct the machinery alignment are presented in the following procedure.
1. The rough shaft alignment should be verified with a straightedge, and precision scale. This is a three-dimensional check that covers vertical and horizontal position, plus axial spacing across the coupling. Large errors in any
direction should be corrected at this time. In all cases, the expected alignment moves should be well within the travel range of the dial indicators.
2. Measure and record the coupling diameter A, the distance from face indicator plane to the moveable near foot B, and distance from the face indicator
plane to moveable far foot C. Check and record the coupling span.
3. Based on the machine dimensions, set the indicator brackets on a mandrel
and check bracket sag in accordance with the previously stated procedure.
4. Check and correct for soft foot as previously discussed.
5. Mark the 3, 6, 9 and 12 oclock positions on both coupling hubs with a paint
or ink marker. It is very important to maintain the two shaft orientations,
especially when runout compensation is necessary.
6. Mount the indicator bracket and the dial indicator(s) on the shaft of the
moveable machine. This shaft must be able to be rotated for the alignment
readings. If a face indicator cannot be used, establish an alternate measurement of axial coupling span (e.g., inside micrometer or feeler gauge).
7. Position both rim and face dial indicators at the top of the coupling hub.
Make certain the rim indicator is perpendicular to the shaft centerline, and
the face indicator is perpendicular to the coupling face. Set both indicators
to the middle of their respective travel range, and zero both indicators.
8. If possible, rotate the shaft on the stationary machine, and record coupling
hub and face runouts every 90. If this stationary rotor cannot be turned,
then alternate methods for determination of runout should be employed.
9. Turn the moveable shaft through one complete rotation and check for adequate indicator travel. It is always a good idea to follow the indicator
around with a mirror to make sure that the plunger is in contact with the
shaft surface throughout the sweep, and to confirm the reading polarity.
Also check for obstructions that may interfere with the indicator bracket.
10. Reposition the indicators at the top 12 oclock location, and re-zero. Again
rotate the shafts through one complete rotation and verify the indicators
return to zero at the 12 oclock position. If the indicators do not return to
zero, check the setup for any possible looseness or movement, and repeat.
11. Rotate the moveable shaft around to 3, 6, and 9 oclock positions, and record
the dial indicator readings and direction (polarity) at each location. Orientation is typically viewed from the fixed to the moveable, and this must be
stated with the recorded indicator readings for future reference. Fig. 12-15
describes the traditional method used for documenting this type of dial
677
indicator information. The zeroes at the 12 oclock position represent the

manually set zero points. The rim readings are typically shown on the outer
diameter of the circle, and the face readings are presented inside the circle.
It is always a good idea to obtain several sweeps, and present a realistic
average dial indicator reading at the 3, 6, and 9 oclock positions.
Face Readings
Rim Readings
0
0
+7 -4
Fig. 1215 Direct Indicator
Readings For Rim And Face
Measurements
-2 +5
-6
+12
12. Check the validity of the averaged dial indicator readings. In all cases, the
following expression must be satisfied:
Top + Bottom = Right + Left
(12-8)
Since the top readings are set to zero, this equation may be simplified as:
Bottom = Right + Left
(12-9)
For the rim indicator readings shown in Fig. 12-15, it is clear that +12 is
equal to +7 added to +5. This is a fundamental validity test, and if it does
not agree to within 1.0 or 2.0 Mils, the readings are corrupted. In this situation, the cause of the inequality must be determined and corrected before
proceeding with the alignment. In all cases, the plus or minus polarity of
the indicator readings must be considered and included into (12-9).
13. The averaged readings from step 11 must now be corrected for bracket sag.
As previously stated, bracket sag is a negative number, and it must be subtracted from the dial indicator readings. The bottom vertical reading must
have the total measured sag, or twice the actual sag subtracted from the
indicator reading. The horizontal readings must have the actual sag, or one
half of the measured sag subtracted from the indicator reading. For example, if the indicator bracket used for the measurements shown in Fig. 12-15
displayed a total bottom sag reading of -6.0 Mils, and a pair of side sag
readings of -3.0 Mils, the sag corrections would be performed as follows:
Left = ( +7 ) ( -3 ) = +10
Right = ( +5 ) ( -3 ) = +8
Bottom = ( +12 ) ( -6 ) = +18
678
Chapter-12
These corrected rim readings are combined with the face readings, and the
results are shown in Fig. 12-16. In most instances face sag is insignificant
and it is not considered. Hence, the rim and face readings in Fig. 12-16 are
the values that will be used to determine the vertical alignment changes.
Face Readings
Rim Readings
Corrected For Sag
0
0
+10 -4
Fig. 1216 Face Measurements Combined With Rim
Measurements That Are Corrected For Bracket Sag
-2 +8
-6
+18
14. It may be possible to save time by performing vertical and horizontal alignment moves simultaneously. However, machines have a tendency to slide
horizontally during vertical adjustments. The most direct approach is to
make the vertical moves first and then proceed with the horizontal shifts.
For face measurements, the following equations (12-10) and (12-11) may be
used to calculate the required correction at each foot of the moveable equipment for angular misalignment.
where:
Angnf
Angff
Face
A
B
C
=
=
=
=
=
=
B
An g nf = ( Face ) ----
A
(12-10)
C
Ang ff = ( Face ) ----
A
(12-11)
Shim Change at Near Foot for Vertical Angular Misalignment (Mils)

Shim Change at Far Foot for Vertical Angular Misalignment (Mils)
Face Dial Indicator Reading At 6 oclock on The Coupling Face (Mils)
Diameter of the Coupling Hub where the Face Reading Was Obtained (Inches)
Distance Between the Indicator Plane and the Center of the Near Foot (Inches)
Distance Between the Indicator Plane and the Center of the Far Foot (Inches)
These dimensions in equations (12-10) through (12-13) are consistent with

the machinery arrangement diagram presented in Fig. 12-15. Intuitively, if
the face readings are zero, there is no angular misalignment, and the problem reverts to a simple parallel offset, such as shown in Fig. 12-11. If the
bottom face reading is positive, the results from (12-10) and (12-11) will be
positive. The plus sign will indicate the requirement to add shims, and raise
the moveable machine. Conversely, a negative value is indicative of a need
to remove shims and lower the moveable machine.
679
15. The vertical offset in rim and face alignment is determined by the bottom
corrected indicator reading. The bottom corrected reading is divided in half
to give the centerline offset as previously described by equation (12-5). If
the angular offset from equations (12-10) and (12-11) are combined with the
vertical offset, the following two equations may be used to calculate the
total vertical correction at each support foot of the moveable machine.
where:
B Rim
Vert nf = ( Face ) ---- ------------
A 2
(12-12)
C Rim
Vert ff = ( Face ) ---- ------------
A 2
(12-13)
Vertnf = Total Shim Change at Near Foot for Vertical Angular and Offset Misalignment (Mils)
Vertff = Total Shim Change at Far Foot for Vertical Angular and Offset Misalignment (Mils)
Rim = Rim Dial Indicator Reading At 6 oclock on the Coupling Hub (Mils)
Again, the required movement direction is determined by the sign (). A

positive (+) sign for equations (12-12) and (12-13) indicates that the moveable machine must be raised. A final negative (-) sign requires that the
moveable machine be lowered at both support feet.
As an example of these calculations, consider the corrected rim and face
readings shown in Fig. 12-16. From these dial indicator readings, the Face
value was equal to -6 Mils, and the corrected Rim reading was +18 Mils.
Assume that the swept face diameter A was 8 inches, and the distances
from the indicator plane to the support feet B and C were equal to 21 and
40 inches respectively. Based on this information, the required vertical corrections may be computed from equations (12-12) and (12-13) as follows.
21 18
Vert nf = ( 6 ) ------ ------ = 15.75 9 = 24.75 25 Mils
8 2
40 18
Vert ff = ( 6 ) ------ ------ = 30 9 = 39 Mils
8 2
From these calculations, it is clear that the near foot should be lowered by
25 Mils, and the far foot should be lowered in elevation by 39 Mils.
16. After completion of the vertical elevation changes on both of the moveable
machine supports the rim and face indicators should be re-zeroed, and
another sweep made of the coupling. At this time, the vertical angularity
and offset should be corrected. However, if residual misalignment still persists, the previous steps 9 through 15 should be repeated.
680
Chapter-12
17. Once the vertical corrections have been completed, the horizontal movements can be made. If the machine was not moved horizontally, the previous sweeps may be used to calculate the required horizontal movements. In
essence, the vertical readings identify the vertical misalignment, and the
horizontal dial indicator readings are used to identify and correct the horizontal misalignment between shafts. Stated in another way, the vertical
dial indicator readings (top and bottom) are used to compute the vertical
shim changes at the near and far support feet. During these vertical calculations the horizontal dial indicator readings are not used. Similarly, the
horizontal dial indicator readings (left and right) are used to determine the
horizontal or sideways moves. For the horizontal changes, the vertical dial
indicator readings are not required.
18. In most cases it is desirable to re-sweep the coupling to obtain new dial
indicator readings. This will identify the current horizontal position, and
check the vertical position. The horizontal shifts are calculated in the same
manner as the vertical movements. Specifically, the horizontal alignment
changes may be determined with the following equations:
B Rim
Horiz nf = ( Face ) ---- ------------
A 2
(12-14)
C Rim
Horiz ff = ( Face ) ---- ------------
A 2
(12-15)
where: Horiznf = Shift at Near Foot for Horizontal Angular and Offset Misalignment (Mils)
Horizff = Shift at Far Foot for Horizontal Angular and Offset Misalignment (Mils)
The Face and Rim numbers used in equations (12-14) and (12-15) are based
upon the horizontal dial indicator readings. Some individuals prefer to rezero the indicators horizontally prior to the horizontal moves. This is perfectly acceptable as long as consistency is applied, and the validity rule of
equation (12-8) is satisfied.
Also, if equations (12-14) and (12-15) display positive (+) signs, the moveable machine must be moved to the left. If these two expressions produce
negative (-) signs, then the machine must be moved to the right.
19. After calculating the required horizontal moves in the previous step 18, the
diagnostician should position horizontal dial indicators at all four corners of
the moveable machine. This will allow the horizontal machine movement to
be accurately monitored as the machine is shifted into position.
20. Repeat steps 18 and 19 until the desired horizontal alignment has been
achieved. Double-check that the coupling gap is still within specification.
Reverse Indicator Shaft Alignment
681
21. Once the final alignment condition has been reached, obtain one final set of
indicator readings. Correct these readings for sag, and identify this information as the final alignment readings.
This procedure and the associated discussion was directed at achieving a
straight 0-0-0-0 rim and face alignment between two shafts. In reality, thermal
growth and other running conditions must be factored into the final alignment.
In most machinery applications, target values are established to accommodate
the changes between the cold alignment and the normal running condition. The
methods to obtain these offsets are discussed at the end of this chapter. As with
any other form of machinery alignment, the previously outlined procedure can be
adapted to a variety of specific applications. In all cases, once the fundamental
concept is understood, the modification of a standard procedure to address a particular mechanical situation is easily accomplished.
REVERSE INDICATOR SHAFT ALIGNMENT

Reverse indicator shaft alignment is similar to the rim and face technique.
This method measures the shaft centerline offsets between two adjacent shafts
or couplings. In most cases, the two shafts are rotated together, and simultaneous rim readings are acquired. A typical machinery arrangement subjected to
reverse indicator readings is shown in Fig. 12-17. The machine on the left side of
this diagram is designated as the fixed or stationary machine, and the machine
on the right side is considered as the moveable unit. A dial indicator bracket is
attached to the shaft or coupling of the fixed machine, and the indicator is positioned in contact with the shaft or coupling hub of the moveable machine. A second indicator bracket is mounted on the shaft or coupling of the moveable
machine, and the attached dial indicator is placed in contact with the shaft or
Distance Between
Dial Indicators (D)

Fixed to Moveable
Moveable
Machine
Near
Foot
Fixed
Machine
Dial Indicator Reading

Moveable Machine Hub
Far
Foot
Dial Indicator Reading

Fixed Machine Hub
Distance to
Near Foot (B)
Distance to Far Foot (C)
Fig. 1217 General Arrangement For Reverse Indicator Shaft Alignment
682
Chapter-12
coupling hub of the stationary machine. This setup for shaft alignment has
become quite popular due to the following advantages.
Reverse indicator provides better dimensional accuracy for small diameter,
long span machines (i.e., the coupling diameter is less than coupling span).
Reverse indicator does not require removal of the coupling spool piece.
In most cases, both rotors are turned together, and coupling hub runout does
not influence the dial indicator readings.

Axial float does not significantly influence the alignment measurements.
Only one indicator is mounted on each indicator bracket and this decrease
in overhung weight reduces the dial indicator bracket sag.

A graphical display of the relative shaft positions and the final alignment
condition are easily generated, and are visually meaningful.

Virtually all mechanical procedures display various benefits combined with
some drawbacks. Reverse indicator shaft alignment is no exception and the traditional disadvantages of this technique are summarized as follows:
Reverse indicator shaft alignment requires the rotation of both rotors. When
one machine cannot be turned, reverse indicators cannot be used.

Provides marginal accuracy on close coupled machines with large diameter
couplings (i.e., the coupling diameter is greater than coupling span).

Cannot be used on small machines with insufficient room to install two indi-
cator brackets and associated dial indicators.

On machines with large diameter couplings and extremely long spans, the
indicator brackets become difficult to handle and obtain consistent data.

The tools and equipment used for reverse indicator shaft alignment are the
same devices required for rim and face measurements. The setup is fundamentally the same, except for the use of brackets with only one dial indicator, and the
fact that the coupling spool piece is generally not removed. From Fig. 12-17, the
distance between the indicator plane on the fixed machine to the center of the
near foot of the moveable machine is again identified as the B dimension. Similarly, the distance between the dial indicator plane on the fixed machine to the
center of the far foot on the moveable machine is once again called the C dimension. The axial distance between dial indicator planes will be referred to as the D
dimension. In most installations, this D distance will be greater than the shaft
end spacing, but D will always be less than the B or C dimensions.
It should be mentioned that reverse indicator alignment may be performed
in several different ways. For instance, one bracket with one dial indicator may
be mounted on the fixed machine and used to sweep the coupling hub of the
moveable machine. This bracket and indicator may then be removed from the
fixed machine, and then reversed and attached to the moveable machine. In this
location, the indicator is used to sweep the coupling hub of the fixed machine.
Moving of brackets and dial indicators does minimize the required hardware to
perform the alignment. However, it is time consuming, and prone to error if the
683
mounting positions are not accurately repeated. In most industrial applications,

a pair of dial indicators and associated support brackets are mounted across the
coupling. This takes slightly longer during the initial setup, but it is much easier
and faster during the actual alignment work. The method discussed in the following procedure locates both indicators at the same angular position on the
shafts. Some methods use indicators that are opposed by 180 across the coupling. Although the results are identical, the sign convention can become confusing. In all cases, the sign convention, orientation, and equipment viewing
direction must be documented and maintained throughout the alignment work.
Reverse indicator alignment is a common practice in many industries. Due
to the widespread use of this technique, the following detailed procedure and discussion is presented for both information and reference purposes.
1. Inspect the coupling to verify proper assembly and correct bolt torque.
2. The rough shaft alignment should be verified with a straightedge and precision scale. This is a three-dimensional check that covers vertical and horizontal position, plus axial spacing across the coupling. Large errors in any
direction should be corrected at this time. In all cases, the expected alignment moves should be well within the travel range of the dial indicators.
3. Measure and record the distance D between indicator planes, the distance
from the indicator plane on the fixed machine to the moveable near foot B,
plus the distance from the indicator on the fixed machine to moveable far
foot C. Also, measure and record the coupling span.
4. Based on the machine dimensions, set the indicator brackets on a mandrel
and check bracket sag in accordance with the previously stated procedure.
5. Check and correct for soft foot as previously discussed.
6. Mount both indicator brackets and the associated dial indicators on both
shafts (or couplings), and verify that all brackets and clamps are tight.
7. Position both dial indicators to read their respective hub or shaft surfaces.
Make certain that the indicators are perpendicular to the shaft centerline.
Set both indicators to the middle of their respective travel range, and zero
both indicators.
8. Turn the shaft through one complete rotation and check for adequate indicator travel. It is always a good idea to follow the indicators around with a
mirror to make sure that the plungers are in contact with the shaft surface
throughout the sweep, and to confirm the reading polarity. Also check for
obstructions that may interfere with the indicator bracket.
9. Reposition the indicators at the top 12 oclock location, and re-zero. Again
rotate the shafts through one complete rotation and verify that the indicators return to zero at the 12 oclock position. If the indicators do not return
to zero, check the setup for any possible looseness or movement, and repeat.
10. Rotate both shafts to the 3, 6, and 9 oclock positions, and record the dial
indicator readings and the direction (polarity) at each location. Use an incli-
684
Chapter-12
nometer or dual bubble levels to locate the precise positions at 3, 6, and 9

oclock. In most instances, the shafts should be stopped precisely at 90
increments. If there is any overshoot of the angular position, it is usually
best to continue around in the direction of rotation until the correct angular
position is obtained. This is normally preferable to stopping and reversing
direction (especially with LOP tilt pad bearings). However, there is an
exception to this rule when dealing with machines coupled to gear boxes. In
this situation, the shafts may get hung up in the gear mesh. It is often
advisable to rotate slightly past the desired angular positions, and then
reverse rotation to allow the gear to back off the mesh.
11. Document the reverse indicator data in the manner shown in Fig. 12-18.
Again, these dial indicator readings are obtained with a viewing direction
that originates at the outboard end of the fixed machine and looks towards
the moveable machine. The zeros at the 12 oclock position on both sweeps
represent the manually set zero points. The readings shown on the outside
of each circle are the radial or rim readings from each dial indicator. Again,
it is always a good idea to obtain several sweeps and present a realistic
average dial indicator reading at the 3, 6, and 9 oclock positions.
0
+7
Fig. 1218 Direct Reverse
Indicator Readings
Fixed
Machine
Hub
+12
+5
-15
Moveable
Machine
Hub
-21
-36
12. Check the validity of the dial indicator readings with equation (12-9). This
expression requires that the sum of the left and right indicator readings are
equal to the bottom value. A variance of 1.0 to 2.0 Mils is considered acceptable. However, if these tolerances are exceeded, the dial indicator readings
are corrupted. Once again, the source of the error must be determined and
corrected before proceeding with the shaft alignment.
13. The averaged readings from step 11 must now be corrected for bracket sag.
As previously stated, bracket sag is a negative number and it must be subtracted from the dial indicator readings. The bottom vertical reading must
have the total measured sag or twice the actual sag subtracted from the
indicator reading. The horizontal readings must have the actual sag or one
half of the measured sag subtracted from the indicator reading. For
instance, if the indicator brackets used for the readings shown in Fig. 12-18
displayed a total bottom sag reading of -6.0 Mils, and a pair of side sag
readings of -3.0 Mils, the sag corrections would be performed as follows:
685
Fixed Hub Readings:
Moveable Hub Readings:
Left = ( +7 ) ( -3 ) = +10
Right = ( +5 ) ( -3 ) = +8
Bottom = ( +12 ) ( -6 ) = +18
Left = ( -15 ) ( -3 ) = -12

Right = ( -21 ) ( -3 ) = -18
Bottom = ( -36 ) ( -6 ) = -30
These sag corrected readings are shown in Fig. 12-19. These vertical dial
indicator readings will be used to determine the vertical shim changes, and
the horizontal values will be used to compute the horizontal moves.
0
+10
Fig. 1219 Reverse Dial
Indicator Readings Corrected For Bracket Sag
Fixed
Machine
Hub
+8
-12
+18
Moveable
Machine
Hub
-18
-30
14. Based on the machine geometry (dimensions B, C, and D) and the sag corrected dial indicator readings the vertical and horizontal alignment
moves may now be determined with direct calculations or a graphical solution. These computations may be accomplished with a variety of handheld
calculator alignment programs, or the more sophisticated personal computer programs that provide both computed and graphical results. These
programs are excellent tools; however, it is necessary to understand the
basic concepts before depending on any type of automated system. Thus,
the continuation of this reverse indicator procedure will address the specific
details for calculating the moves, plus plotting the graphical solution. This
work begins by determining the vertical and horizontal centerline offsets
using the following set of equations:
Vo
Vo
fix
mov
Bottom Top
Bottom
= ---------------------------------------
= ---------------------
2
2
fix
fix
(12-16)
Bottom Top
Bottom
= ---------------------------------------
= ---------------------
2
2
mov
mov
(12-17)
Right Left
= -----------------------------------
2
fix
(12-18)
Right Left
= -----------------------------------
2
mov
(12-19)
Ho
Ho
fix
mov
686
Chapter-12
where:
Vo-fix
Vo-mov
Ho-fix
Ho-mov
Bottom
Top
Right
Left
=
=
=
=
=
=
=
=
Vertical Shaft Offset at Fixed Machine (Mils)

Vertical Shaft Offset at Moveable Machine (Mils)
Horizontal Shaft Offset at Fixed Machine (Mils)
Horizontal Shaft Offset at Moveable Machine (Mils)
Indicator Reading at Bottom of Sweep (Mils)
Indicator Reading at Top of Sweep, Normally Equal to Zero (Mils)
Indicator Reading at Right Side of Sweep (Mils)
Indicator Reading at Left Side of Sweep (Mils)
Obviously, equations (12-16) through (12-19) are nothing more than extensions of the previous shaft offset equations (12-5) and (12-6) applied to each set of
dial indicator readings. It is also clear that bracket sag corrected numbers must
be used from this point on. Again, it is very important to view the machine from
a constant position and maintain the proper sign convention. For example, the
sag corrected dial indicator readings from Fig, 12-19 may be used to compute the
vertical and horizontal offsets using equations (12-16) through (12-19) as follows:
Vo
Vo
Ho
Ho
mov
fix
mov
fix
Bottom
18
= ---------------------
= ------
= +9 Mils
2
fix
2 fix
Bottom
30
= ---------------------
= ---------
= -15 Mils
2
mov
2 mov
Right Left
8 10
= -----------------------------------
= ---------------
= -1 Mil
2
fix
2 fix
Right Left
( 18 ) ( 12 )
= -----------------------------------
= ----------------------------------
= -3 Mils
2
2
mov
mov
15. Based on the vertical and horizontal offsets, the following common set of
equations may be used to compute the required movements for the near
foot and the far foot of the moveable machine.
B
Vert nf = ( V o + V o ) ---- { V o }
D
fix
mov
fix
(12-20)
C
Vert ff = ( V o + V o ) ---- { V o }
D
fix
mov
fix
(12-21)
B
Horiz nf = ( H o + H o ) ---- { H o }
D
fix
mov
fix
(12-22)
C
Horiz ff = ( H o + H o ) ---- { H o }
fix
mov
fix
D
(12-23)
687
Once again, the B and C dimensions are the distances between the indicator plane on the fixed machine to the center of the near foot and far foot on
the moveable machine as shown in Fig. 12-17. The dimension D is the axial
distance between indicator planes. As previously noted, the B, C, and D
dimensions are all expressed in inches. The final results for vertical shim
changes at the near foot Vertnf are computed with equation (12-20), and the
shim changes at the far foot Vertff are determined with (12-21). Again, a
plus sign (+) reveals a need to add shims and raise the machine. Conversely,
a final negative sign (-) requires a removal of shims to lower the machine.
With respect to the horizontal equations (12-22) and (12-23), the horizontal
shift is defined by Horiznf at the near foot, and Horizff at the far foot respectively. A positive sign (+) on the horizontal value indicates a need to move
the machine to the left. A negative sign (-) requires a shift of the moveable
machine to the right. Since all of the dial indicator readings carry the units
of Mils, the final alignment shift values will also be in Mils.
As an example of these movement calculations, equations (12-20) through
(12-23) will be used to compute the overall vertical and horizontal moves
based upon the previously calculated vertical and horizontal offsets. These
values will be combined with the same machine geometry used for the rim
and face example. Namely, the distance to the near foot B will be maintained at 21 inches, and the distance to the far foot C will remain at 40
inches. If the spacing between reverse indicator planes D is 8 inches, the
required movements are calculated as follows.
21
( 6 ) 21
Vert nf = ( 9 15 ) ------ { 9 } = ----------------------- ( 9 ) = 24.75 25 Mils
8
8
40
( 6 ) 40
Vert ff = ( 9 15 ) ------ { 9 } = ----------------------- ( 9 ) = 39 Mils
8
8
21
( 4 ) 21
Horiz nf = ( 1 3 ) ------ { 1 } = ----------------------- ( 1 ) = 9.5 Mils
8
8
40
( 4 ) 40
Horiz ff = ( 1 3 ) ------ { 1 } = ----------------------- ( 1 ) = 19 Mils
8
8
From these calculations and the associated sign convention it is clear that
shims must be removed from both feet of the moveable machine. Furthermore, the moveable machine must be shifted to the right. Although these
calculations are fairly straightforward, it is easy to make a numerical or a
sign mistake. Hence, it is always desirable to perform a graphical solution
of the alignment move to check the calculations.
688
Chapter-12
16. A graphical solution of the alignment data yields the required movement at
the feet of the moveable machine. This procedure is based upon plotting the
calculated offsets at each dial indicator plane, and then extending a
straight line to the support feet of the moveable machine. Traditionally, the
horizontal axis is used to define the axial locations of both dial indicator
planes, plus the center of the two support feet on the moveable machine.
Units of inches are generally employed on this axis. The vertical axis displays the offsets in engineering units of Mils. In all cases, the offsets
between shafts are calculated for both vertical and horizontal directions
using equations (12-16) to (12-19).
17. It is good practice to plot the vertical and the horizontal data separately to
avoid confusion, and errors in polarity. The actual graph is generated by
plotting the offsets at both dial indicator planes, and extending the straight
line out to the moveable support feet. When plotting the point at the stationary plane a positive number (+) is placed above the centerline, and a
negative value (-) is plotted below the centerline. When locating the point
on the moveable dial indicator plane, a positive number (+) is place below
the centerline, and a negative (-) offset is located above the centerline. A
line intersecting the two dial indicator offset points is then drawn and
extended past the locations of the support feet. The required offsets are the
distance from the centerline to the point where the extended shaft centerline intersects the lines associated with the moveable support feet. If the
intersection occurs above the centerline, that indicates that the moveable
machine is high and it must be lowered. Similarly, if the intersection point
of the extended centerline falls below the desired centerline, the moveable
machine is low and it must be elevated.
18. For demonstration purposes, the vertical offset data in the current example
is plotted on Fig. 12-20. In this case, equation (12-16) revealed a vertical offset at the indicator plane on the fixed machine of +9 Mils. This point is plotted at the 0 inch location on Fig. 12-20. Across the coupling at the dial
indicator plane on the moveable machine, equation (12-17) showed an offset
of -15 Mils. Since the span between indicator planes was stated as 8 inches,
the -15 Mil vertical offset was plotted at 8 inches. Since this value carried a
negative sign, it was plotted above the desired centerline. These two measurement points establish the extended shaft centerline through the location of the near foot and the far foot. From Fig. 12-20, it is clear that the
near foot location intersects the extended centerline at 25 Mils. This means
that the moveable machine is 25 Mils high and shims equal to this distance
should be removed. At the far foot, the extended shaft centerline is 39 Mils
above the desired centerline. Hence, a shim thickness equal to 39 Mils
should be removed at this location to lower the outboard end. These values
and directions are identical to the previously calculated moves.
Vo
=
mov
-15 Mils
10
Vo =
fix
+9 Mils
+
-
-10
Desired Centerline
D
B Dimension
C Dimension
10
20
30
Distance (Inches)
Down
Far Foot
25 Mils Above
Centerline
Move Down
20
ed
nd ine
e
t
l
Ex nter
Ce
39 Mils Above
Centerline
Move Down
Near Foot
30
Indicator Plane on
Moveable Machine
Indicator Plane on
Fixed Machine
Vertical Offset (Mils)
40
689
Up
40
50
=-3 Mils
mov
+
-
Ho =
fix
Far Foot
Near Foot
Ho
19 Mils to Left
of Centerline
Move to Right
10
d
de e
en rlin
t
Ex nte
Ce
9.5 Mils to Left

of Centerline
Move to Right
15
Indicator on Moveable Machine
Indicator Plane on Fixed Machine
Horizontal Offset (Mils)
20
Left
Fig. 1220 Graphical Solution Of Vertical Alignment Based On Reverse Dial Indicators
Desired Centerline
-10
Right
-1 Mil
10
20
30
Distance (Inches)
40
50
Fig. 1221 Graphical Solution Of Horizontal Alignment Based On Reverse Dial Indicators
690
Chapter-12
19. The graphical solution may now be repeated for the horizontal data as
shown in Fig. 12-21. Equation (12-18) revealed a horizontal offset at the
fixed machine of -1 Mil. This point is plotted at the 0 inch position on Fig.
12-21. Across the coupling at the moveable machine indicator plane, equation (12-19) displayed a -3 Mil offset. Again, the span between indicator
planes was 8 inches, and the -3 Mil horizontal offset was plotted at 8 inches.
Since this value carried a negative sign, it was plotted above the desired
centerline. These two horizontal offsets establish the extended shaft centerline up through the location of the near and the far foot. From Fig. 12-21,
the near foot location intersects the extended centerline at 9.5 Mils. This
means that the moveable machine is 9.5 Mils to the left of the desired centerline, and the machine should be shifted 9.5 Mils to the right. At the far
foot, the extended shaft centerline is 19 Mils to the left of the desired centerline, and the outboard end of the moveable machine should be shifted 19
Mils to the right. Again, these values duplicate the previous calculations.
20. As previously discussed, the most direct approach is to make the vertical
shim changes first, and then proceed with the horizontal moves.
21. For the horizontal moves, position horizontal dial indicators at all four corners of the moveable machine. This will allow the horizontal machine movement to be accurately monitored as the machine is shifted into position.
22. Repeat steps 10 through 21 until the desired alignment has been achieved.
Double-check that the coupling gap is still within specification.
23. Once the final alignment condition has been reached, obtain one final set of
indicator readings. Correct these readings for sag, and identify this information as the final alignment readings.
The diagnostician is encouraged to use both the calculation and the graphical procedure to cross-check results. In all cases, clearly label the data with consistent designations. On graphical solutions, do not clutter up the page with
sloppy or meaningless information. Always separate the vertical and horizontal
alignment plots, and use separate pages for the calculation of the cold offsets. It
is also a good idea to write the date and time on each piece of acquired and calculated data. This date and time stamping helps keeps things organized and
sequential on difficult alignment projects. As with most technical endeavors,
clear and accurate work will be rewarded by excellent results.
As stated at the end of the section on rim and face alignment, the presented
procedures are directed at achieving a perfect alignment between shafts under a
static condition. Obviously, as the machinery journals move around in their
respective bearings, and process fluids are introduced, and thermal growth (or
contraction) initiates, the shaft alignment will change. The variation of shaft
position under normal operating conditions versus the cold alignment condition
is presented in the text section on hot alignment techniques. This is a very
important consideration, but it cannot be properly addressed until the fundamental concepts of rim and face plus reverse indicator alignment are understood.
Optics, Lasers, and Wires for Shaft Alignment
691
OPTICS, LASERS, AND WIRES FOR SHAFT ALIGNMENT

Optical tools have many machinery applications including shaft alignment.
As discussed earlier in this chapter, optics may be used to measure the relative
positions of two shafts, or the differential machinery growth between at stop and
normal operating conditions. This makes optical measurements a strong candidate for measurement of the thermal movement of process machinery. In addition, optics is quite useful for tasks such as setting shaft centerlines. By placing
a mirror on a shaft end, and auto-collimating a scope, shaft centerlines may be
set and measured. Optical tooling is quite adept for measuring shaft gaps, and
face parallelism. Although the potential uses of optical measurements on
machinery alignment are extensive, many optical methods are slow, and they are
often quite complex. There is also considerable skill required to correctly execute
these optical measurements. Unless the machinery diagnostician is specialty
trained in all of the particular nuances associated with optical equipment, the
application of this technology may be both difficult and time consuming.
Laser alignment has become an extremely popular form of shaft alignment
due to the ease and speed of operation. The tools for laser alignment are identical
to the devices discussed in the section on laser position alignment. For shaft
alignment, specialized mounting brackets are available to allow the laser and
detector to be secured to the shaft surfaces. Most systems utilize some sort of
chain clamp for this purpose. Magnetic holders are also available to allow units
to be mounted on faces where circumferential clamps may not be practical.
Although variations exist between systems from different vendors, most laser
measurements mimic the operation of shaft mounted dial indicators.
The laser provides a high degree of accuracy, combined with computerbased automation of the measurements. These computer controls have become
quite sophisticated, and are capable of determining hot and cold alignment offsets, acceptable alignment tolerances, plus many other functions. The associated
computer can generally display the results in numeric and graphical formats. As
with any computer-based system, the final results are totally dependent on operator input, plus proper operation of the system. Hence, the diagnostician must
have a complete working knowledge of the particular laser system, plus a full
understanding of basic alignment principles.
Shaft wire alignment techniques are generally not performed on horizontal
machines due to obvious limitations. However, wire alignment is commonly
applied on large vertical machines, such as hydro turbines and large motor
driven pumps. These types of units are often several stories tall and they employ
a series of guide bearings at various elevations. Many of these machines are built
with access passages to allow dropping alignment wires from the top to the bottom guide bearing. In this application wire sag is not a problem, but the wires
must be secured in a perfectly vertical position to have meaningful distance comparisons between elevations. The measurement tools previously discussed for
wire bore alignment are fully applicable to these large vertical units. It should
also be mentioned that laser alignment is becoming popular for this application
due to reduced setup time and faster processing of the results.
692
Chapter-12
HOT ALIGNMENT TECHNIQUES

As previously mentioned in this chapter, the desired alignment for many
applications must include offsets to compensate for thermal expansion (or contraction) of the machinery at normal operating conditions. Typically, the thermal
related movements are calculated or measured, and these values are converted
into vertical and horizontal offsets at the coupling hubs. These desired offsets are
then combined with the current position of the shafts, and appropriate corrections computed or charted.
For instance, if the final desired position of the moveable machine is 10 Mils
high and 4 Mils to the right, the equivalent reverse indicator readings are shown
in Fig. 12-22. These desired indicator readings do not include bracket sag, and
they may be directly converted to vertical and horizontal offsets.
0
Fig. 1222 Desired

Reverse Indicator Readings For 10 Mil Vertical,
And 4 Mil Horizontal Offsets Between Shafts
+14
Fixed
Machine
Hub
+6
-14
+20
Moveable
Machine
Hub
-6
-20
In the general case, the desired vertical and horizontal offsets may be computed in the same manner as equations (12-5) and (12-6). Expanding these fundamental expressions into the desired vertical and horizontal offsets at each dial
indicator plane yields the following four expressions.
Vd
Vd
Hd
Hd
where:
Vd-fix
Vd-mov
Hd-fix
Hd-mov
=
=
=
=
Bottom
= ---------------------
2
d fix
(12-24)
Bottom
= ---------------------
2
d mov
(12-25)
Right Left
= -----------------------------------
2
d fix
(12-26)
Right Left
= -----------------------------------
2
d mov
(12-27)
fix
mov
fix
mov
Desired Vertical Shaft Offset at Fixed Machine (Mils)

Desired Vertical Shaft Offset at Moveable Machine (Mils)
Desired Horizontal Shaft Offset at Fixed Machine (Mils)
Desired Horizontal Shaft Offset at Moveable Machine (Mils)
Hot Alignment Techniques
693
To verify the validity of the reverse indicator readings shown in Fig. 12-22,
equations (12-24) to (12-27) will be used to compute the desired vertical and horizontal offsets in the following manner:
Vd
Vd
Hd
Hd
mov
fix
mov
fix
Bottom
20
= ---------------------
= ------
= +10 Mils
2
d fix
2 d fix
Bottom
20
= ---------------------
= ---------
= -10 Mils
2
d mov
2 d mov
Right Left
6 14
= -----------------------------------
= ---------------
= -4 Mils
2
d fix
2 d fix
Right Left
( 6 ) ( 14 )
= -----------------------------------
= -------------------------------
= +4 Mils
2
2
d mov
d mov
Clearly the desired parallel offsets of 10 Mils vertically and 4 Mils horizontally are displayed in the above calculations. These desired offsets must now be
combined with the current or ambient offsets previously described by equations
(12-16) through (12-19). These two sets of offset equations may now be integrated
into the following general solutions for vertical and horizontal alignment moves
at the near foot and the far foot of the moveable machine.
B
Vert nf = ( V o V d + V o
V d ) ---- { V o V d }
D
fix
fix
mov
mov
fix
fix
(12-28)
C
Vert ff = ( V o V d + V o
V d ) ---- { V o V d }
D
fix
fix
mov
mov
fix
fix
(12-29)
B
Horiz nf = ( H o H d + H o
H d ) ---- { H o H d }
D
fix
f
mov
mov
fix
fix
(12-30)
C
Horiz ff = ( H o H d + H o
H d ) ---- { H o H d }
D
fix
f
mov
mov
fix
fix
(12-31)
Obviously, if the desired offsets in equations (12-28) through (12-31) are set
equal to zero (i.e., no process related or thermal offset), these expressions would
default back to the earlier simplistic set of equations (12-20) through (12-23).
Using the example values from the previous section on reverse indicators,
plus the desired offsets from Fig. 12-22, equations (12-28) through (12-31) may
now be used to compute the required vertical and horizontal corrections at the
near foot and far foot as follows:
694
Chapter-12
21
( 6 ) 21
Vert nf = ( 9 10 15 ( 10 ) ) ------ { 9 10 } = ----------------------- ( 1 ) = 14.8 Mils
8
8
40
( 6 ) 40
Vert ff = ( 9 10 15 ( 10 ) ) ------ { 9 10 } = ----------------------- ( 1 ) = 29 Mils
8
8
21
( 4 ) 21
Horiz nf = ( 1 ( 4 ) 3 4 ) ------ { 1 ( 4 ) } = ----------------------- ( 3 ) = 13.5 Mils
8
8
40
( 4 ) 40
Horiz ff = ( 1 ( 4 ) 3 4 ) ------ { 1 ( 4 ) } = ----------------------- ( 3 ) = 23 Mils
8
8
Clearly, these alignment calculations are simple to perform, but they are
prone to error due to the constant manipulation of plus (+) and minus (-) signs.
This may not be a problem at the beginning of an alignment job, but pluses and
minuses have a tendency to get crossed up in the middle of the night. The
machinery diagnostician must be fully aware of this potential for silly errors, and
he or she should implement procedures to prevent, or at least minimize, errors.
As a minimum, all calculations should be rechecked at least twice. In an ideal
situation, it is best if two people can acquire the data and independently perform
the calculations.
Based on the above calculations, approximately 15 Mils of shims must be
removed from the near foot, and 29 Mils of shims removed from the far foot of the
moveable machine. In addition, the moveable machine must be moved to the
right by 13.5 Mils at the near foot, and 23 Mils at the far foot. The same results
could be obtained by plotting the reverse indicator data. Case history 43 at the
end of this chapter will demonstrate the combination of this calculation technique, plus a graphical solution of a hot pump driven by an induction motor.
The previous example and associated discussion have assumed that the
desired offsets due to thermal expansion are known. In actuality, this is a presumptuous posture, since the determination of the thermal growth is often a difficult endeavor. In many instances, the physical alignment moves are easily
computed and accomplished. However, the majority of the technical effort is
directed at determining the positional changes of the rotating shafts between the
initial cold or ambient alignment conditions, and full load operation at normal
process temperatures.
Initially, the techniques used to calculate thermal growth (or contraction)
presented earlier in this chapter may be used to estimate the changes in
machine casing position based on differential temperatures. The issue of determining average operating pedestal temperature may be difficult on complex
machines installed in cramped quarters. In some cases, thermography may be
used to determine the temperature gradient across the machine support members. This used to be a difficult measurement, but the advent of electronic thermography based upon infrared video cameras and associated digital signal
695
processing has considerably simplified this task, and the associated cost. Equipment of this type may also be used to examine and document the temperature
profile of machine casings, couplings, bearings housings, and associated piping
and support structures. Hence, tools are available to accurately determine surface temperatures in a wide variety of situations.
The machinery diagnostician should not forget that many machines have a
simple temperature gradient on their support members. The use of a contact
thermometer can often fully describe the surface temperature profile, and provide the basis for calculation of the thermal growth. Certainly these thermal
growth calculations are useful and necessary, but one should not forget the old
adage of one good measurement is worth a thousand expert opinions.
The measurement of process machinery thermal growth may be accomplished in several different ways. Specifically, the optical measurements discussed earlier in this chapter represents one of the first techniques for accurate
measurement of machinery position changes. The 1973 article by Al Campbell6
describes the use of a jig transit combined with reverse dial indicator readings to
properly align rotating equipment. In the same year, and at the same symposium, Charles Jackson7 presented a paper that included optical alignment, plus
the use of proximity probes mounted on cold water stands to measure thermal
growth. The proximity probes added a new dimension to alignment measurements by providing an electronic output (i.e., probe DC gap voltage) that could be
directed to various types of voltage recorders. With this information, machine
position may be measured as a function of time, and the machinery movement
could be tracked during startup and thermal heat soak. The Jackson cold water
stands could be used to measure changes in casing position, or direct shaft position measurements on exposed shaft surfaces.
Since proximity probes are electronic micrometers, their use for alignment
growth measurements represents a direct and logical application. However, in
many machines, cold water stands supporting proximity probes cannot be used
due to space limitations on the baseplate. Even with Invar brackets, it may not
be possible to mount probes close to the shaft or a suitable external machinery
surface. In these situations, proximity probes and companion measurement surfaces may be mounted on Dodd bars8. These devices consist of two sets of bars
that span across the coupling and allow differential vertical and horizontal thermal growth measurements. Typically, two sets of X-Y probes are mounted on one
bar, and their respective targets are mounted on the second bar. This technique
provides a relative growth (or contraction) measurement across the coupling, but
it does not allow the determination of absolute position changes as obtained with
cold water stands or optics.
6 A. J. Campbell, Optical Alignment of Turbomachinery, Proceedings of the Second Turbomachinery Symposium, Turbomachinery Laboratories, Texas A&M University, College Station, Texas
(October 1973), pp 8-12.
7 Charles Jackson, Cold and Hot Alignment Techniques of Turbomachinery, Proceedings of
the Second Turbomachinery Symposium, Turbomachinery Laboratories, Texas A&M University,
College Station, Texas (October 1973), pp. 1-7.
8 V.R. Dodd, Total Alignment, (Tulsa, Oklahoma: The Petroleum Publishing Company, 1975).
696
Chapter-12
Another approach to the measurement of machinery growth was developed

by Jack Essinger9. This technique requires the installation of two tooling balls on
the machinery bearing housings, plus mating tooling balls on the baseplate or
foundation. Precision distance measurements are then acquired between these
tooling balls under ambient alignment conditions, and they are compared
against similar measurements with a full heat soak of the machinery. The distances between balls are measured with extended range dial indicators mounted
in spring loaded telescoping gauge columns. These devices allow accurate length
measurements and the differential machinery growth may be determined with
calculator programs or graphical solutions.
Certainly laser alignment systems have evolved to increasing levels of
sophistication and capability. In many applications, lasers have replaced the previously discussed techniques for measurement of position changes between the
ambient conditions and the fully heat soaked machinery operating at normal
process rates. Although this is wonderful technology, the diagnostician must
always remember the machinery fundamentals, and be able to recognize when
the alignment computer yields bogus or questionable results. In addition, the
diagnostician must keep an open mind, and use the best available measurement
and alignment technique for each specific application.
Finally, the shaft alignment discussions in this chapter have addressed only
a simple two case machinery train, where one unit is fixed or stationary and the
other case is moveable. Although this covers a large percentage of the installed
machinery, it does not specifically address the multicase trains used for critical
service. For these installations, one case is generally designated as the fixed
machine and the other cases are sequentially aligned to the fixed unit. It is
highly desirable to plot the graphic solutions for all rotors in the machinery train
on the same piece of paper. Actually, one piece of graph paper should be used for
the vertical solution and another for the horizontal alignment. This overall train
view is necessary from a visibility standpoint, and it soon becomes apparent
when the unit gets bolt bound or shim limited due to the consecutive stacking of
case to case offsets. In many ways field shaft alignment is similar to field rotor
balancing. That is, in most cases there are no ideal or universal solutions, but
there are generally more than one acceptable solution. It is always a constant
challenge to use the most appropriate techniques to satisfy the engineering
requirements in the minimum amount of time.
9 Jack N. Essinger, Benchmark Gauges for Hot Alignment of Turbomachinery, Proceedings of

the Ninth Turbomachinery Symposium, Turbomachinery Laboratories, Texas A&M University, College Station, Texas (October 1980), pp 127-133.
697
Case History 43: Motor To Hot Process Pump Alignment

Shaft alignment of most general purpose motor driven process pumps is a
fairly straightforward exercise. In many instances, the operating conditions do
not warrant thermal offsets or any running positions corrections. Hence, the
methods to achieve a straight ambient alignment condition discussed earlier in
this chapter may be directly applied. However, for machines that require an initial offset to compensate for a specific hot or cold running condition, the job
becomes somewhat more complicated.
For demonstration purposes, consider the pump-motor machinery train
described in Fig. 12-23. The two pole, 400 HP induction motor drives a multistage boiler feed water pump. The motor is equipped with roller bearings, and it
operates at 3,570 RPM under full load. The horizontally split pump supplies
boiler feed water to a package boiler. Pump suction is maintained at a constant
temperature of 150F, and the pump discharge pressure is 650 Psig at approximately 250F. The pump is equipped with sleeve bearings and a fluid film thrust
bearing. This machinery arrangement is the same configuration that has been
used for examples in the previous discussions of rim and face, plus reverse indicator shaft alignment.
From the pump coupling flange, a distance of 21 inches is noted to the near
foot of the motor, and 40 inches to the far foot. In the opposite direction, the
pump inboard, or near foot, is 12 inches from the pump coupling flange, and the
outboard, or far foot, is located 30 inches from the coupling flange. Once again,
the spacing between dial indicator planes is 8 inches. The major area of concern
during alignment of this unit is the anticipated thermal growth of the pump during operation. Minimal expansion of the motor is expected, but the elevated
pump operating temperatures suggest that thermal alignment offsets might be
appropriate. As shown in Fig. 12-23, the inboard pump pedestals experience a
52F rise, combined with an average 80F rise at the outboard supports. Due to
the rigidity of the pump suction and discharge piping, the pump is considered as
Discharge
at 250F
Suction
at 150F

Fixed to Moveable
D=8
Moveable
Motor
Fixed
Pump
T = 80F
T = 52F
12
30
Far
Foot
Near
Foot
Near
Foot
Far
Foot
20
B = 21
C = 40
Fig. 1223 Machine Arrangement Of A Hot Pump Driven By An Induction Motor
698
Chapter-12
the fixed unit, and the motor will be the moveable unit for alignment. Based on
the 20 inch height between the baseplate and the pump centerline, the expected
thermal growth may be calculated as follows:
L pum p
L pum p
nf
ff
= T L C = { 52 F } 20 Inches 0.0063 Mils/InchF = 6.5 Mils

= T L C = { 80 F } 20 Inches 0.0063 Mils/InchF = 10.1 Mils
Hence, the thermal growth at the pump near foot should be 6.5 Mils, combined with 10.1 Mils on the pump far foot. These values may be easily confirmed
with Fig. 12-1, describing the thermal growth of cast steel elements. These calculations could be supplemented by physical measurements of the actual thermal
growth. In either case, the diagnostician has thermal growth information at the
support feet of the stationary machine, and these values must be converted to
vertical offsets at the dial indicator planes. The easiest way to accomplish this
conversion is by plotting the thermal offsets at each pump support leg, and
extending this desired pump centerline back across the coupling. For instance,
on Fig. 12-27, the 10 Mil growth on the pump outboard was plotted at the pump
far foot location of 30 inches from the pump coupling flange. Similarly, the 6.5
Mil thermal growth was located 12 inches from the pump coupling flange. Connecting these two points, and extending the desired pump centerline across the
coupling provided the intercepts with both dial indicator planes.
From Fig. 12-27, the desired vertical offset at the pump coupling hub Vd-fix
may be read directly from the plot scale as -4 Mils. On the motor coupling hub,
the desired vertical offset Vd-mov is similarly determined to be +2.5 Mils. As
always, the polarity changes across the coupling, and the sign goes from negative
to positive. Assuming no appreciable horizontal offset, these desired vertical offsets may be easily converted to the reverse dial indicator readings in Fig. 12-24.
To state it another way, when the pump-motor set is correctly aligned to compensate for the pump thermal growth, a set of ambient reverse indicator readings
should be quite similar to the values displayed in Fig. 12-24.
0
Fig. 1224 Desired

Reverse Indicator Readings To Compensate For
Pump Thermal Growth
-4
Fixed
Pump
Hub
-8
-4
+2.5
Moveable
Motor
Hub
+2.5
+5
Since these are true offsets, indicator bracket sag has not been included.
However, bracket sag does exist in the initial sweep values depicted in Fig. 1225. Obviously, a true comparison of reverse indicator readings cannot be
addressed until the initial data has been corrected for bracket sag. Although
these calculations were performed earlier in this chapter, the exercise will be
699
Fig. 1225 Initial Reverse

Indicator Readings
Describing Pump To Motor
Cold Offset With Bracket
Sag Included
+7
Fixed
Pump
Hub
+5
-15
+12
Moveable
Motor
Hub
-21
-36
repeated for the sake of completeness. For both brackets, the total indicator reading was -6 Mils (= 2 x Sag) with the bracket inverted. Hence, the initial reverse
indicator readings shown in Fig. 12-25 may be sag corrected as follows:
Fixed Pump Hub Readings:
Moveable Motor Hub Readings:
Left = ( +7 ) ( -3 ) = +10
Right = ( +5 ) ( -3 ) = +8
Bottom = ( +12 ) ( -6 ) = +18
Left = ( -15 ) ( -3 ) = -12

Right = ( -21 ) ( -3 ) = -18
Bottom = ( -36 ) ( -6 ) = -30
These results may now be displayed as the sag corrected reverse indicator
readings shown in Fig. 12-26. This data reveals a minor horizontal offset, combined with a significant vertical deviation from the desired readings previously
developed and displayed in Fig. 12-24.
0
Fig. 1226 Initial Reverse

Indicator Readings
Describing Pump To Motor
Cold Offset With Bracket
Sag Removed
+10
Fixed
Pump
Hub
+18
+8
-12
Moveable
Motor
Hub
-18
-30
From Fig. 12-26, it is clear that the vertical offset Vo-fix at the pump coupling hub is +9 Mils (=+18/2). In a similar fashion, the vertical offset at the
moveable motor coupling hub Vo-mov is equal to -15 Mils (=-30/2). These values
may now be combined with the desired offsets of Vd-fix =-4 Mils, and Vd-mov =+2.5
Mils. Based upon these desired and actual vertical offsets, plus the machine
train dimensions, the required vertical corrections at the motor feet may be computed with equations (12-28) and (12-29) in the following manner:
21
( 4.5 ) 21
Vert nf = ( 9 ( 4 ) 15 2.5 ) ------ { 9 ( 4 ) } = --------------------------- 13 = 24.8 Mils
8
8
40
( 4.5 ) 40
Vert ff = ( 9 ( 4 ) 15 2.5 ) ------ { 9 ( 4 ) } = --------------------------- 13 = 34.5 Mils
8
8
700
Chapter-12
Vo
fix
Vd =
fix
-4 Mils
+
-
+
Vd
Desired
Pump
Centerline
=
mov
2.5 Mils
-20
Up
0
20
Distance (Inches)
40
Down
10
-40
-15 Mils
+9 Mils
10
=
mov
35 Mils Above
Desired Centerline
Move Down
20
Vo =
Motor Near
Foot
te
25 Mils Above
nd
C
Desired Centerline
en ed
te Mo
Move Down
rlin to
r
e
Motor Far Foot
Ex
Motor Indicator Plane
Pump Indicator Plane
Vertical Offset (Mils)
30
Pump Near Foot
40
Pump Far Foot
The computed results indicate that approximately 25 Mils of shims must be

removed from the inboard (near foot) of the motor. Additionally, 35 Mils of shims
must be extracted from the motor far foot. These are clearly shim reductions as
defined by the negative sign on both sets of calculations. The same information
may be displayed graphically as presented in Fig. 12-27.
60
Fig. 1227 Graphical Solution Of Vertical Alignment Of Induction Motor To Hot Boiler
Feed Water Pump With Calculated Thermal Offsets At Pump Support Feet
In this diagram, the previously discussed thermal offsets are shown along
with the desired pump centerline at the bottom of the plot. This desired pump
centerline is extended past the motor support feet, and it is the reference line
that the motor shaft attempts to meet. The initial vertical position of the motor
shaft is defined by the Vo-fix and Vo-mov points plotted at dial indicator planes. An
extension of this line produces the extended motor centerline. The actual shim
corrections required to bring the motor centerline back down to the desired pump
centerline are determined by the vertical difference between the two centerlines.
Hence, a 25 Mil correction is required at the motor near foot, and a 35 Mil
change is indicated at the motor outboard (far foot). It is also clear from Fig. 1227 that the motor shaft is sitting higher than the desired position of the pump
shaft. Hence, the motor must be lowered to become properly aligned with the
pump. These results are totally consistent with the previous alignment calculations.
701
Note that the complexity of Fig. 12-27 is significantly greater than the simple offset alignment previously discussed in Fig. 12-20. The inclusion of the thermal offset, plus the requirements to properly identify the new points has
substantially complicated the entire graphical solution. If this machinery train
consisted of three or four individual cases, and various physical offsets had to be
incorporated for each case, the complexity of the alignment diagram would be
exponentially increased. It is once more highly recommended that the diagnostician be very meticulous in this work.
The horizontal corrections on this machine were determined in accordance
with the previous horizontal alignment plot displayed in Fig. 12-21. As recommended earlier in this chapter, the vertical shim changes are performed first,
and then the couplings are re-swept to check the horizontal offsets. The changes
in horizontal motor position are achieved by loosening the outboard end motor
bolts, and one of the inboard bolts. The motor was then pivoted on the remaining
tight inboard bolt to achieve the proper centerline position.
The final aligned position of this induction motor and the hot boiler feed
water pump was within 1.5 Mils of the desired indicator readings in Fig. 12-24.
The slight remaining offset was considered to be well within the misalignment
tolerance of the coupling. After completion of this work, the subsequent startup
was quite smooth. The steady state vibration data at full load and heat soak
revealed a well-aligned machinery train, with no evidence of pre-loads at any
bearing.
In the overview, proper machinery alignment is mandatory for successful
long-term operation of the mechanical equipment. The tools and techniques discussed in this chapter are directed at providing a general introduction to this
subject, plus specific instructions on performing some basic alignment functions.
Some diagnosticians might sayIll probably never do a field alignment job, so
why bother learning about machinery alignment In actuality, so many mechanical problems are directly associated with misalignment and the resultant rotor
preloads, this topic cannot be ignored. Furthermore, it must be recognized that
other mechanisms besides misalignment may be actively engaged in destroying
a mechanical coupling. Poor lubrication, incorrect installation, overloading, plus
a variety of corrosion attacks, and fatigue failures may be responsible for coupling difficulties. For descriptions and photographs of a wide range of coupling
problems, the reader is encouraged to examine the array of failed couplings presented by Mike Calistrat10 in his text entitled Flexible Couplings.
10 Michael M. Calistrat, Flexible Couplings, Their Design Selection and Use, (Houston: Caroline Publishing, 1994), pp. 390-424.
702
Chapter-12
BIBLIOGRAPHY
1. Calistrat, Michael M., Flexible Couplings, Their Design Selection and Use, Houston:
Caroline Publishing, 1994.
2. Campbell, A. J., Optical Alignment of Turbomachinery, Proceedings of the Second
Turbomachinery Symposium, Turbomachinery Laboratories, Texas A&M University, College Station, Texas (October 1973), pp. 8-12.
3. Dodd, V. R., Total Alignment, Tulsa, Oklahoma: The Petroleum Publishing Company, 1975.
4. Essinger, Jack N., Benchmark Gauges for Hot Alignment of Turbomachinery, Proceedings of the Ninth Turbomachinery Symposium, Turbomachinery Laboratories,
Texas A&M University, College Station, Texas (October 1980), pp. 127-133.
5. Gas Engine and Compressor Field Service Manual - Section 16, Drawing R19307A
(Painted Post/Corning, New York: Dresser-Rand, 1986), pp. 1-12.
6. Jackson, Charles, Shaft Alignment Using Proximity Probes, ASME Paper 68-PET25, Dallas, Texas (September 1968).
7. Jackson, Charles, Cold and Hot Alignment Techniques of Turbomachinery, Proceedings of the Second Turbomachinery Symposium, Turbomachinery Laboratories,
Texas A&M University, College Station, Texas (October 1973), pp. 1-7.
8. Mancuso, Jon R., Couplings and Joints - Design, Selection, and Application, New
York: Marcel Dekker, Inc., 1986.
9. Piotrowski, John, Shaft Alignment Handbook, 2nd edition, New York: Marcel Dekker, Inc., 1995.
13
Applied Condition Monitoring 13
his chapter is based upon a presentation

for the Turbomachinery Symposium of Texas A&M University. The original document was part of a Short Course1 entitled Inspection and Overhaul of Major
Turbomachinery presented in conjunction with personnel from Chevron USA.
This course covered a variety of machinery subjects, and the three sections prepared by the senior author have been extracted and modified for this chapter.
The first portion provides an introduction to condition monitoring concepts, with
examples of diagnosed problems on operating machines. The second section covers calibration and verification checks on instrumentation systems that should
be performed during a turnaround. The third section addresses the application of
condition monitoring during machinery startup following an overhaul. Again,
illustrative examples of machinery case histories have been incorporated.
Since the original presentation of this short course, the instrumentation
systems have evolved to more sophisticated levels. However, the basic measurements and approaches associated with condition monitoring remain unchanged.
Hence, this document is still considered to be appropriate and representative of
the application of condition monitoring to turbomachinery systems.
MAINTENANCE PHILOSOPHIES
Maintenance philosophies concerning major machinery trains are many
and varied. Each company, geographic plant location, and individual operating
unit are subjected to numerous factors that influence the normal maintenance
approach. Items such as product demand, required machine availability, historical behavior, and management attitudes are integrated into an overall doctrine.
Generally, maintenance activities may be categorized as either reaction-based,
time-based, or condition-based maintenance
It is reasonable to briefly review each category, and identify the respective
merits of each approach. Since the most cost-effective approach resides within
1 Robert C. Eisenmann, John East, and Art Jensen, Short Course 1 - Inspection and Overhaul
of Major Turbomachinery, Proceedings of the Seventeenth Turbomachinery Symposium and Short
Courses, Turbomachinery Laboratory, Texas A&M University, Dallas, Texas (November 1988).
703
704
Chapter-13
the third group of condition maintenance, additional discussion will be provided.

This will include the tools and techniques necessary for evaluating machine condition, plus case histories to illustrate the application of condition monitoring
towards operating machinery.
Reaction-based maintenance requires little if any advanced planning. It
typically takes the form of continuous operation until a failure occurs. Then it
becomes a purely reactionary mode to repair the damage and return the machinery train to operation as soon as possible. Maintenance of this type is expensive,
and the potential of an unexpected catastrophic failure is significant. In addition,
normal safety considerations may be compromised when a machine is allowed to
run until a failure occurs. Reaction maintenance might be appropriate in facilities where the installed machinery is minimal; and the plant is not totally dependent on the reliability of any individual machine. A realistic example would be a
unit that contains only small process pumps that are each 100% spared.
Expenditures for instrumentation, documentation, and personnel training
are low with reaction maintenance. However, the risks are often significant, and
most companies gradually moved away from this type of maintenance during the
1970s and 1980s. Unfortunately, the emphasis on corporate profitability during
the 1990s have driven many industrial plants into severely curtailing a variety
of preventative and/or predictive maintenance programs. These plants are
returning to reaction maintenance. Although the immediate financial implications may be impressive, the long-term prognosis of this approach is frightening.
The concept of time-based maintenance evolved due to the economic
impact and safety unsuitability of reaction maintenance. Within this category,
machinery maintenance is planned and scheduled on a periodic basis. Often an
annual cycle is used for shutdown, disassembly, and mechanical inspection of the
various machine elements. Although this approach is substantially more costeffective than simply reacting to failures, two obvious deficiencies are apparent.
First, a mechanical malfunction may occur during the normally scheduled
operating time period. This may result in an unexpected machine failure, with a
necessary default to reactive maintenance. At this stage, repair costs may
become exorbitant, and production losses may be irrecoverable. Second, timebased maintenance often results in the disassembly of perfectly good machinery.
This is not only expensive from a labor and downtime position, there is another
risk involved. That is, there may be a problem induced during machinery reassembly. Hence, the turbine/compressor train that was in excellent condition prior
to a turnaround, cannot be restarted due to an assembly error.
Critical mechanical items such as bearing clearances, seal installation, or
coupling hub position may be violated during an inspection, and the machinery
rendered inoperative. The solution to avoiding this situation is the direct
approach of: If its not broken, then dont fix it. This basic philosophy leads to the
evolution of condition maintenance. That is, machine repair is based upon a reasonable knowledge of the mechanical condition of the machinery, and it is not
governed by the calendar or a potentially destructive mechanical failure.
Condition-based maintenance is safer, and economically more attractive than either of the reaction or time-based methods. The superiority of this
Condition Monitoring
705
approach has been demonstrated in various ways. In many plants the maintenance dollar expenditures per installed horsepower have significantly decreased
following the implementation of condition maintenance. This is generally due to
a reduction in direct maintenance, an extension of the time period between overhauls, plus secondary effects such as the reduction in the quantity of warehouse
spare parts. The fundamental concept behind condition maintenance consists of
evaluating the process machinery from many aspects and assessing the current
mechanical condition. Maintenance plans and material requirements are then
driven by the anticipated work scope. This is certainly more advantageous than
either of the two previously mentioned maintenance categories.
Evaluation of machinery condition requires management support and commitment. Resources must be allocated for the devices to determine machinery
behavior via performance and vibration response measurements. In addition,
documentation and historical records must be established and maintained.
Finally, trained personnel are required to implement and sustain the program.
In an effectively run program, hardware and administration costs are more than
offset by the savings incurred from reduced maintenance. The success or failure
of this approach is dependent upon a good knowledge of the condition of the
installed machinery. More specifically, condition monitoring of the machinery is a
necessary prerequisite. The tools and techniques required for implementation of
condition monitoring are discussed in the following sections of this chapter.
CONDITION MONITORING
Detailed observation, plus monitoring of machinery behavior, provides the
database for condition monitoring. As discussed by John Mitchell2 there are
many technical aspects to consider in any condition monitoring program. However, for this text, the two major characteristics to be addressed are the machinery performance and the vibration response characteristics.
These characteristics typically include continuous monitoring plus periodic
samples obtained with greater detail/resolution. For example, machine performance is constantly monitored with the normal operating instrumentation. Specific performance tests may be periodically conducted to compute operating
points and the efficiency of individual machinery cases. During the course of
these precision measurements, the accuracy of the operating instruments would
also be verified. The frequency of the performance tests is a function of the
machinery service. Obviously, a clean refrigeration compressor might only be
subjected to a performance test on an annual basis. Conversely, a cracked gas
compressor with the potential for internal coke accumulation might be checked
on a monthly or even a weekly basis.
Similarly, the vibratory behavior of a major machinery train should be continually monitored and protected with an automated system. Overall radial
2 John S. Mitchell, Introduction to Machinery Analysis and Monitoring, second edition (Tulsa,
OK: Pennwell Publishing Company, 1993), pp. 291-345.
706
Chapter-13
vibration levels plus thrust position are used for automatic shutdown. Additionally, detailed vibration response data may be acquired on a periodic basis. These
dynamic vibration signals are viewed in a variety of formats, and compared with
previous results under similar operating conditions. During the course of this
routine data acquisition, accuracy of the monitoring instrumentation may also
be verified. Some computer systems allow digital storage of the dynamic data,
and detailed trending of all normal vibration parameters. However, it is comforting to routinely commit this data to a hard copy paper format for historical comparison and future reference purposes.
Other types of information, such as lube oil analysis, thermography, and
current analysis on electric machinery, might be incorporated into the condition
monitoring program. In all cases, the diagnostician must exercise common sense
when including additional technologies into the program. For instance, if a routine examination of an oil sample from a process pump revealed 10.0 parts per
million of a metallic bearing material, there would be cause for concern. However, on large turbomachinery, the main reservoir may contain several thousand
gallons of oil. If a routine oil analysis detected 1.0 part per million of bearing
metal, it would already be too late. In reality, by the time the oil analysis
detected the presence of trace quantities of the bearing metal, the bearing was
already destroyed. In this case, the small quantity of bearing metal is virtually
undetectable within the large volume of the lube and seal oil system. Hence,
many technologies are available in the marketplace, but the diagnostician must
carefully select only those measurements that are related to the direct determination of the machinery condition.
The machinery database should be optimized to quantify or characterize
normal behavior and highlight abnormal characteristics. Generally, the diagnostician should establish realistic limits that define normal variation of the measured parameters. For example, a rotational speed vector may exhibit minor
changes in amplitude and phase during routine machinery operation. It is necessary to define an appropriate window or envelope of normal behavior for each
parameter, or group of associated parameters. When these various windows of
normal behavior are exceeded, the diagnostician must then analyze the machine
to determine the probable cause. In this manner, maintenance options may be
reasonably discussed, and a suitable balance maintained between the economics
of continued operation versus various levels of corrective maintenance.
MACHINERY PERFORMANCE
A performance measurement and evaluation program will vary with the
sophistication of the machinery and the available process data. In an ideal situation, the process controls are interfaced through a Distributed Control System
(DCS). All necessary pressures, temperatures, flow rates, and molecular weights
are measured and scanned as part of the normal control scheme. Hopefully, a
direct shaft torque measurement is included, and this data is also directly available. In this type of system all of the necessary parameters are logged by the
Machinery Performance
707
computer. In addition, the performance calculations for head, flow, and efficiency
may also be stored within the DCS. Respective data points can be compared with
the OEM performance test curves, and deviations from expected conditions identified. The calculated data points may also be trended as a function of time.
With an analog control system, the available process information may not
have suitable resolution. In this case, a separate set of data may be obtained
with calibrated, high resolution gauges. This includes pressures, temperatures,
and flow rates. Fluid samples may be extracted and processed to determine fluid
properties. For condensing turbines the condensate flow may be measured, and
for electrical equipment the voltage and current may be obtained. Although
these indirect measurements are not as accurate as a direct shaft torque reading, they do differentiate between poor versus acceptable performance.
The machinery test data may be used to compute the performance parameters, and this data may be compared against the performance curves, and
trended with time. Detailed test procedures, measurements, and calculations are
published by the ASME. Standards such as the Power Test Code for Compressors
and Exhausters3 (PTC-10), and the Performance Test Code for Steam Turbines4
(PTC-6) are readily available. It should be recognized that a strict ASME performance test is often difficult to conduct in a field environment. In many cases, the
necessary hardware provisions are not available to obtain the required precision.
In addition, measurement of items such as transmitted torque across a coupling
is impossible if the proper transducers are not installed. However, for many
machines, an overall polytropic or isentropic efficiency may be sufficient for
trending purposes. Since the basic requirement is to consistently trend and identify any degradation in performance, a less sophisticated test may be acceptable.
In addition to the ASME standards, other references on the performance
characteristics of process machinery are available. This includes mechanical
engineering handbooks, OEM literature, and various textbooks. The diagnostician should design a performance determination system that is suitable and
compatible with the installed machinery. Factors to consider for performance
measurements are summarized in the following list:
Machine Types
Expected Degradation Mode(s)
Measurements Available
Accuracy of Measurements
Process or Test Gauges
Method of Performance Calculation
Method of Trending Results
Performance Limits
3 ASME Power Test Codes - Test Code for Compressors and Exhausters, PTC-10, The American Society of Mechanical Engineers, (New York: The American Society of Mechanical Engineers,
1965, reaffirmed 1986).
4 ASME Performance Test Codes - Code on Steam Turbines, PTC-6, The American Society of
Mechanical Engineers, (New York: The American Society of Mechanical Engineers, 1976, reaffirmed
1982).
708
Chapter-13
Considering these items will allow the implementation of a usable system

that will provide visibility of reduced performance on a machinery train. This
type of information is highly beneficial by itself, and it becomes even more meaningful when coupled with the detailed machinery vibration response data.
VIBRATION RESPONSE DATA

It has been recognized for many years that the mechanical integrity of a
machinery train can be evaluated by a detailed analysis of the vibratory motion.
This is achieved by using a variety of dynamic transducers that provide information on both the casing vibration and the relative shaft motion plus position.
From an evolutionary standpoint, the original vibration transducers were
seismic devices. As discussed in chapter 6, these transducers measure the vibration of the attached surface with respect to an inertial reference frame (free
space). They can be mounted directly on bearing housings, machine casings,
foundations, piping, or other stationary elements. For measurements in the
lower frequency domains (600 to 90,000 CPM), velocity coils may be used. Accelerometers are generally employed to obtain high frequency characteristics (typical maximum of 1,200,000 CPM). These external vibration transducers provide
considerable information about the motion of the surface on which they are
mounted. However, many industrial machines display minimal external casing
motion. For instance, a high speed rotor in a barrel compressor may experience
an attenuation of 100:1 when shaft and casing measurements are compared.
This is certainly reasonable when one considers a 600 pound rotor contained
within a 12,000 pound casing. Any casing vibration due to shaft motion is attenuated by the enormous mass of the casing, plus the viscous damping inherent
with oil film bearings and seals. It should also be mentioned that the mechanical
impedance between a rotor excitation and an external measurement point is not
constant. Impedance variations with speed and journal position are normal, and
should be expected. In short, there is no such thing as a constant amplitude ratio
between shaft and casing vibration.
By the early 1970s shaft sensing eddy current displacement probes evolved
into common usage. These proximity probes are noncontacting devices that allow
measurement of rotor motion relative to the probe mounting surface. Typically,
the probes are attached to the bearing housings. Thus, the shaft displacement
measurement provides the relative motion between the rotor and the bearing
housing. This type of measurement is very useful for diagnosis of machinery malfunctions. In addition, since displacement probes are not susceptible to external
(not machine related) excitations, they are excellent candidates for continuous
monitoring and trending of rotor vibration.
Today, most large machinery trains are equipped with shaft displacement
probes. In many cases, the probes are connected to permanent machinery protection systems. These systems include radial vibration monitors, thrust position
monitors, tachometers, differential expansion indicators, plus other machine
related information such as bearing temperatures from embedded thermocou-
Vibration Response Data
709
ples or resistance temperature detectors (RTDs).

Although definitive process measurements may vary considerably between
plants and machinery trains, the shaft displacement transducer suite will be
reasonably similar. That is, radial bearings should contain a pair of orthogonal
probes, and thrust bearings will usually be equipped with a pair of axial probes.
Standard configurations are addressed in documents such as the American
Petroleum Institute (API) standard for Vibration, Axial Position, and Bearing
Temperature Monitoring Systems5. In general, the API standards depict the
minimum instrumentation requirements for a piece of machinery. The actual
installed instruments may be increased as a function of unique machinery, or the
need for additional diagnostic capabilities.
As discussed earlier in this text, the coplanar radial probes are typically
located above the shaft at 45 from the true vertical centerline. The probes
mounted to the left of top dead center are referred to as the Y-Axis or vertical
probes. The transducers mounted to the right of the vertical centerline are called
the X-Axis or horizontal probes. For a machine with a single thrust collar, two
axial probes may be used to observe the end of the shaft. In other arrangements,
one axial probe might be inserted between the thrust pads to observe the thrust
collar, and a second axial probe might be mounted at the end of the turbine shaft.
Finally, a once-per-revolution timing probe is installed. This transducer is used
to phase reference the machinery train, and provide a tachometer input.
Typically, all the monitors from one machinery train will be located in one
rack of monitors. The location of monitors in the rack should follow the location
of transducers installed on the equipment. That is, the first monitor position
should be reserved for the turbine thrust, followed by the governor end radial
vibration monitor, and so forth throughout the machinery train. On large units,
the monitors may occupy two or more racks. In this situation, the left side of the
second rack begins where the right side of the first rack ends.
As previously mentioned, these types of systems are often armed for automatic shutdown. Certainly this is totally dependent on the company and their
maintenance philosophy. In organizations that fully support the concepts of
machinery protection and condition monitoring, the units will be fully armed for
radial vibration, and thrust position shutdown. Organizations that are not fully
convinced of the merits of these devices will probably not tie the monitors into
automatic trip circuits.
Shaft displacement probes are suitable for measurement of frequencies
between zero and approximately 90,000 CPM. At higher frequencies, displacement amplitudes are quite small, and respective changes would probably go
unnoticed. However, the low frequency capability allows the probe to measure
actual changes in the average rotor position. Typically, this is applied for the
thrust position measurements. However, it should be noted that both radial and
axial displacement probes do measure relative position (DC gap) as well as the
5 Vibration, Axial Position, and Bearing Temperature Monitoring Systems API Standard
670, Third Edition, American Petroleum Institute, (Washington, D.C.: American Petroleum Institute, November 1993).
710
Chapter-13
shaft dynamic motion (vibration). Hence, it is highly desirable to alarm, or even

shutdown, on excessive changes in radial shaft position as indicated by variations in probe DC gap voltages. As demonstrated throughout this book, the average radial or axial rotor position is significant information, and should always be
incorporated in an assessment of machinery vibration.
Output signals from shaft sensing proximity probes are typically presented
in units of Mils (0.001 Inches). For vibration readings, the actual measurement is
a peak-to-peak value that represents the total motion within the bearing. Position measurements (axial or radial) are normally identified as a plus or minus
position change from some defined reference location. Since the probes are
observing rotating shafts that may be subjected to a variety of forces and frequencies, the output signals are often complex waveforms. In order to consistently characterize these dynamic signals, the data is typically discussed in
terms of the following parameters:
Overall Vibration Amplitude

Running Speed (1X) Vibration Amplitude
Running Speed (1X) Vibration Phase Angle
Frequency Content
Relative Position Change
The overall vibration amplitude is the unfiltered peak to peak motion measured by the probe. Running speed (1X) vibration amplitude is the motion filtered precisely at rotational speed. Similarly, the phase angle at (1X) running
speed is the phase relationship between the rotational speed vibration signal,
and the timing probe. Although instrumentation manufacturers vary somewhat
in their phase circuitry, the most logical phase convention consists of the angular
relationship between the leading edge of the timing notch and the next positive
vibration peak. As discussed in previous chapters, this is expressed as a phase
lag, with engineering units of degrees. The frequency content of a signal
describes the mixture of frequencies and their respective amplitudes. Generally,
the steady state vibration data is presented in the following hard copy formats:
Unfiltered Orbit and Time Base

Running Speed (1X) Filtered Orbit and Time Base
Spectrum Plot
Shaft Position Plot
The unfiltered orbit and time base may be observed directly on an oscilloscope connected to the probe output signals. The running speed (1X) filtered data
is viewed in a similar manner, with the addition of a narrow band-pass filter on
each signal. Typically, this is accomplished with a synchronous tracking filter
that adjusts the filter center frequency to match the exact running speed of the
machine. The oscilloscope screen may be photographed with a camera to provide
permanent documentation of the orbital and time domain information. Various
computer-based systems also perform the same basic operation, and many examples of this type of data are presented throughout this text.
Bearing Temperature Data
711
Spectrum analysis is used to determine the frequency content of the signal.

This task is handled as a Fast Fourier Transform (FFT) with either a dedicated
instrument, or with a software routine and an appropriate analog to digital conversion. Data of this gender provides visibility of the frequencies present, and
their respective amplitudes. One more, examples of this type of data are shown
throughout this text. Naturally the shaft position plot is a summary of changes
in probe DC gap voltages (e.g., zero speed versus full operating load). This type of
information may be manually calculated and plotted. However, automated diagnostic and monitoring systems perform this task quite easily and accurately.
Although computer-based systems are in common use, it is still advisable to
provide analog backup for checking and verification of the output data. The software required to capture and present vibration data is inherently complex. In
most cases, a simple oscilloscope check of the raw waveforms helps to maintain
sanity and provide credibility to the resultant database.
BEARING TEMPERATURE DATA

Correlation of information to support a common conclusion is always desirable. Hence, measurement and observation of bearing temperatures via embedded Thermocouples or RTDs are highly useful, if not mandatory, for most large
process machines. The traditional bearing temperature measurements recommended in the previously referenced API 670 are summarized as follows:
Both Machine Journal Bearings
Active Thrust Shoes
The location and quantity of temperature sensors is dependent upon the

specific bearing configuration. In most instances, a total of six temperature
probes are installed on each machinery case. This would include temperature
sensors on each of two active and two inactive thrust pads. In addition, a single
temperature pickup would be installed on each journal bearing. Thus, a quantity
of six temperature sensors would be normal for many installations. The number
of radial bearing temperature sensors would be increased if the bearings are long
(e.g., L/D ratio of greater than 0.5). For a long journal bearing, two sensors might
be required to properly detect the maximum bearing temperatures. Specific temperature transducer arrangements and recommended locations are clearly illustrated in API-6706.
The mounting locations of the bearing temperature pickups are preferentially positioned to be close to the load zone on each bearing. This location allows
the measurement of the maximum bearing temperature. It must be mentioned
that the thermocouple is installed only into the base metal of the bearing. Typi6 Vibration, Axial Position, and Bearing Temperature Monitoring Systems API Standard
670, Third Edition, American Petroleum Institute, (Washington, D.C.: American Petroleum Institute, November 1993), pp 23 and 24.
712
Chapter-13
cally, the tip of the thermocouple remains 30 to 50 Mils (0.030 to 0.050 inches)
from the bearing babbitt. Under no circumstances should the thermocouple be
allowed to penetrate the babbitt. This type of installation would result in an
early, and totally unnecessary, bearing failure.
Furthermore, the angular location of the temperature sensors must be carefully controlled to be in the load zone, as determined by shaft rotation and specific bearing configuration. If the thermocouple is on the wrong side of a tilting
pad pivot point, the measured temperature will be less than the desired maximum bearing temperature. In addition, it should be noted that units such as
overhung machines will often have a temperature sensor mounted in the bottom
half of the coupling end bearing and the top half of the wheel end bearing. This is
correct placement of the thermocouples, since the shaft on the overhung rotor
will generally ride in the bottom half of the coupling end bearing and the top half
of the inboard wheel end bearing. When in doubt, the actual load zone of the
shaft within the bearing may be determined by measuring the shift in shaft centerline position with DC gap voltages from proximity probes.
Other machines such as large vertical units often have fully flooded bearings that contain one or more sensors installed in each bearing pad. Again, the
specific machine configuration dictates the quantity and location of bearing temperature probes. Redundant or spare temperature pickups are often installed to
backup the primary sensors. This is especially true for trains that are required to
operate over long intervals between overhauls. In the 1960s the common practice
was to perform disassembly and inspection of critical machines on a yearly cycle.
In the 1990s, the expected run time between major overhauls often exceeds five
years. Hence, redundant temperature sensors make good sense. It is also desirable to install and monitor additional temperature measurements such as the
lube and seal oil supply, plus the ambient temperature. The solution to some
mechanical malfunctions is often dependent upon proper correlation with oil or
atmospheric temperatures.
DATA TRENDING
It is highly recommended that two distinct types of documentation be
employed to chart machinery attrition versus operating time. As previously mentioned, the combination of continuous trend plots, plus a periodic detailed examination of machinery behavior, provides the most useful base for machinery
evaluation. In most cases, the data should consist of information regarding both
vibratory behavior plus performance data. The mechanical information is normally obtained by interfacing the machinery protection system with a data
trending system. In older plants, the vibration and thrust position monitor
recorder outputs were connected to analog strip chart recorders. Process measurements were recorded in a similar fashion with strip chart, circular chart, or
multipoint recorders. This type of instrumentation provided a good deal of information on the machinery, and many problems have been identified and solved
with this type of analog data.
Data Trending
713
Improved resolution, accuracy, and convenience may be obtained from the

digital storage of trend data. For modern process plants containing a Distributed
Control System (DCS) the process operating history is easily stored and
retrieved. In many cases, process and vibration data are archived on nonmagnetic media such as magneto optical disks that have a storage life of twenty or
more years. Hence, the immediate operating conditions, as well as all of the historical information, may be accessed and examined. As a cautionary note, many
digital data storage systems employ averaging algorithms to compress the measured data. This compressed data may be misleading, since a transient event
may be automatically averaged into digital oblivion.
The second important category of trend data consists of periodically acquiring detailed vibration and performance data at constant load conditions. As a
minimum, the vibration information should be processed in both the orbital/time
and frequency domains. In addition, relative shaft positions and bearing temperatures should be obtained for all radial and thrust bearings. The minimum performance data includes all applicable measurements of process pressures,
temperatures, flow rates, loads, and molecular weights. It is important to obtain
these field measurements in a consistent and repeatable manner. The resultant
hard copy data may then be compared from one sample period to the next. In a
typical situation, data is obtained during the initial startup of a new or rebuilt
train. Following temperature and load stabilization (nominally 12 to 48 hours), a
complete set of baseline data is acquired. This baseline information documents
the initial vibratory and performance characteristics of the machinery, and it
provides the reference point for future comparison.
This type of detailed machinery evaluation should be periodically repeated.
The actual time between samples is a function of the type of machinery, severity
of service, gross variations in load and/or speed, and the development of any specific problems. For example, a machinery train in good operating condition may
be examined on a monthly or quarterly schedule. However, if it appears that a
particular mechanical element is showing signs of distress, it would be appropriate to increase the frequency of the samples to weekly or even daily. This higher
sample rate would be maintained until the problem is resolved. It should also be
recognized that during the time periods between detailed data sets, the machinery monitoring system is fully operational, and continuous protection is available
via the automated trip system.
In all cases, the trend data should be directed at providing reliable documentation of the machinery historical behavior. In many cases, significant
changes can be recognized and addressed before a mechanical failure occurs.
Often, the degrading or failing mechanical part can be identified. This provides
the ability to make a reasonable judgment call regarding continued operation
versus the shutdown and repair of a specific item. In some situations it is possible to take a brief outage and repair or replace the offending piece of hardware.
To illustrate these points in greater detail, the following three field case histories
are offered for consideration and demonstration of the application of on-line
machinery condition monitoring.
714
Chapter-13
Case History 44: Four Pole Induction Motor Bearing Failure

A small catalytic cracking unit contains the main air blower machinery
train described in Fig. 13-1. Within this arrangement a 3,500 horsepower, four
pole, induction motor drives a five stage centrifugal air compressor through a
down mesh, double helical gear box. The vibration transducers installed on this
train are directed to a rack of control room monitors. Unfortunately, the monitors
for this machinery train are mounted about eighteen inches from the floor, and it
is reasonably difficult to obtain readings from these knee high monitors. Due to
the location of these monitors, it is understandable why the control room operators pay minimal attention to the vibration amplitudes.
CV
CH
DY
45 45
DH
45 45
Discharge
Normal
Thrust
145/54
Induction
Motor
3,500 HP
4 Pole
Suction
Gear
Box
Compressor
5 Stages
HAa,b
AV
AH
45 45
CCW
Rotation
BV
BH
FV
45 45
FH
45 45
EV
EH
45 45
GV
GH
45 45
HV
HH
45 45
CW
Rotation
Fig. 131 Machinery And Transducer Arrangement For Motor Driven Air Compressor
This particular machinery had been running for an extended time period
with no evidence of any abnormality. Motor shaft vibration response characteristics under typical operating conditions are shown in Fig. 13-2. This diagram documents the motion at the inboard, or coupling end, motor bearing. The top orbit
and time base plots are unfiltered, and the bottom set of traces are filtered at the
running speed of 1,790 RPM. From this data it is clear that the majority of the
motion occurs at motor rotational speed. Although the running speed amplitude
is approximately 2.5 Mils,p-p, the runout compensated levels are in the vicinity of
1.5 Mils,p-p. These amplitudes are historically typical, and mechanically reasonable for this class of motor. In addition, proximity probe DC gap voltages
revealed proper journal position within the sleeve bearing.
Late one Saturday night, a process upset occurred, and the coupled com-
Data Trending
Fig. 132 Motor Coupling End Bearing

Under Normal Condition
715
Fig. 133 Motor Coupling End Bearing

Following Process Upset
pressor surged for an undefined period of time. Vibration and temperature

alarms were acknowledged, but the machine was not shutdown (no automatic
trip system). After the process was back under control, it was noted that the
motor coupling end bearing vibration reading was off scale on the 5.0 Mil,p-p
vibration monitor. In addition, the bearing temperature strip chart showed a ten
minute excursion to 210F, followed by a leveling out to an unusually low journal
bearing temperature of 70F. Following this upset, the meter on the motor
inboard vibration monitor remained pegged. E&I personnel discovered that the
vertical channel vibration amplitude was considerably less than the horizontal
reading. At this time, the motor inboard horizontal probe was disconnected from
the monitor to allow visibility of the vertical probe amplitude. It was noted by
control room personnel that the peak meter reading had changed from 2.6 to
about 3.3 Mils,p-p. It was erroneously concluded that the shaft vibration had only
increased by 0.7 Mils,p-p, and that the horizontal probe had probably failed.
In actuality, the motor coupling end horizontal probe was still operational,
and emitting a proper signal. After reconnecting this transducer, the vibration
response characteristics at the motor inboard following the upset are displayed
on Fig. 13-3. Note that the vertical probe reveals a slight vector shift, but the
horizontal transducer has a rotational speed amplitude of 8.9 Mils,p-p. It should
also be mentioned that the shaft orbits shown in both Figs. 13-2 and 13-3 have
been rotated 45 to allow a presentation of the orbits in a true vertical and horizontal orientation. When the location of proximity probes at 45 from the vertical centerline are considered, it is clear that the vertical probe observes the
minor axis of the elliptical orbit. Simultaneously, the position of the installed
horizontal transducer is directly in line with the major axis of the shaft orbit.
This coincidence of the major and minor axes of the elliptical orbit with the
716
Chapter-13
physical location of the proximity probes reinforces the need for orthogonal (90)
transducers. In this case, if the machinery assessment was based totally upon
the vibration amplitude measured by the vertical probe, the conclusion would
probably favor continued operation. However, when the motion sensed by the
companion horizontal probe is considered, the extent of the machinery distress is
clearly evident in the data shown in Fig. 13-3.
From another perspective, the measured DC gap voltages on both proximity
probes revealed a substantial increase. Specifically, the left-hand probe had a
voltage change from -8.73 to -13.12 volts DC. The differential value of 4.39 volts
DC is equivalent to a distance of 22.0 Mils at a probe sensitivity of 200 mv/Mil.
Similarly, the right-hand probe exhibited a change of -8.94 to -12.66 volts DC.
This differential voltage of 3.72 is equivalent to a radial position shift of 18.6
Mils. Since the probes were mounted at 45 from vertical, the total shaft centerline position change may be determined by a vector summation of the two position shift vector amplitudes as follows:
Shaft Centerline Change =
( 22.0 ) + ( 18.6 ) =
830.0 = 28.8 Mils
Since the gap voltages increased, the distance between the probe tips and
the observed shaft surface must have increased. If it is assumed that the proximity probes remained in a fixed position, the measured changes in gap voltage
must be indicative of the shaft riding lower in the bearing. The only way that this
condition can occur is if the bottom half of the bearing has been damaged.
If that prognosis proved to be correct, then the expanded bearing clearances
will provide less restraint upon the journal (lower stiffness). Assuming a constant driving force (e.g., rotor unbalance), synchronous vibration levels should
increase. Also, it was discovered that the bearing temperature sensor was
improperly located in the top of the bearing assembly. In this position, the sensor
would observe a lower than normal temperature since the load zone was physically on the opposite side of the bearing from the installed thermocouple.
Finally, one last piece of evidence should be considered in this investigation.
Recall that the operating speed of this induction motor was measured at 1,790
RPM. For a normal USA line frequency of 60 Hz, the synchronous speed for a
four pole motor is 1,800 RPM. The slip frequency for a four pole induction motor
is computed from equation (10-44) in the following manner:
Poles
Slip = --------------- ( Synchronous Speed Rotor Speed )
2
4
Slip = --- ( 1, 800 1, 790 ) = 2 10 = 20 CPM
2
The extended time domain traces of the motor and the bull gear revealed an
amplitude modulation occurring every 1.5 seconds. The frequency of this motion
is determined from equation (2-1) as follows:
Data Trending
717
1
1 Cycle
Frequency = ------------------- = --------------------------Period
1.5 Seconds
Cycles
Second
Seconds
Minute
Frequency = 0.667 ----------------- 60 ------------------- = 40 CPM

This amplitude modulation at 40 CPM occurs at twice the motor slip frequency of 20 CPM. It is reasoned that the bull gear was responding to this motor
excitation. The vibration severity of this low frequency modulation is summarized in Table 13-1. This data suggests that the motor excitation at twice slip frequency was transmitted to the bull gear in a much stronger manner after the
process upset. One explanation for this behavior would be an ineffective motor
inboard bearing, and the requirement for the gear bearings to carry more load
through a reduced flexibility coupling.
Table 131 Variation Of Low Frequency Amplitude Modulation
Initial Shaft
Vibration
Shaft Vibration
After Upset
Motor Coupling End Bearing
2.0 Mils,p-p
3.0 Mils,p-p
Bull Gear Coupling End Bearing
0.9 Mils,p-p
2.0 Mils,p-p
Bull Gear Outboard End Bearing
1.1 Mils,p-p
Based on this overall data, the unit was shutdown in an orderly manner for
replacement of the inboard motor bearing. To minimize production loss, the plant
was placed in recycle. This plan allowed the cracking furnace to remain in hot
standby, and permit a plant restart with minimal delay.
Prior to any disassembly, a lift check was performed at the motor inboard.
The dial indicator revealed a total diametrical clearance of 31 Mils. This value is
consistent with the previously calculated changes in probe gap voltages (28.8
Mils). Since the correct bearing clearance should be between 8 to 10 Mils, it was
clear that all measurements pointed towards excessive bearing clearance.
Following disassembly, inspection revealed severe babbitt damage to the
lower half of the inboard bearing liner. Fortunately, the motor journal was not
marred, the bearing liner was replaced, and clearances checked. The new bearing provided correct diametrical clearances, but air gap measurements (rotor to
stator) revealed that the motor rotor was not concentric with the stator. These
variations in air gap exceeded the OEM specifications, and were logically responsible for the previously discussed modulation at twice slip frequency. Due to a
lack of available resources, this problem was not corrected at this point in time.
The train was restarted, and the plant suffered minimal impact due to the brevity of this machine train shutdown.
After a 24 hour thermal soak, the steady state vibration and temperature
data returned to the originally documented levels. This was considered to be
indicative of normal behavior for this four pole induction motor.
718
Chapter-13
Case History 45: Cracked Gas Compressor Intermittent Instability

The four case machinery train described in Fig. 13-4 operates in a large ethylene plant. This equipment is used to compress cracked furnace gas prior to drying and fractionation. The machinery train consists of a steam turbine driving a
low pressure compressor that is coupled to the bull gear of a down mesh, double
helical gear box. The pinion output shaft is coupled to a high stage compressor
that operates between 10,700 and 11,500 RPM. The suction end of this high
stage is positioned at the coupling end. The final discharge nozzle and the thrust
assembly are located at the outboard end of the high pressure compressor.
HV
HH
45 45
Inlet
Exhaust
Discharge
Suction
Suction
Normal
Thrust
CW
Rotation
Normal
Thrust
Normal
Thrust
K
hs
High Stage
Compressor
Gear
Box
DA/a,b
AA/a,b
K
Steam
Turbine
JA/a,b
FA/a,b
Low Stage
Compressor
ls
Discharge
110/47
AV
AH
CV
45 45
CH
FV
45 45
JH
FH
45 45
JV
BV
CCW
Rotation
BH
45 45
60
30
IH
DV
IV
85
60
30
DV
Fig. 134 Machinery And Transducer Arrangement For Cracked Gas Compressor Train
This compressor contains eight impellers and a large balance piston. The
rotor was supported with five pad, tilting pad journal bearings. Both radial bearings are contained within a housing configuration that includes oil control rings
on each end. These control rings were designed by the OEM to float within a circumferential groove cut into each respective bearing housing. The inner diameter of the four control rings contains a babbitt coating and the diametrical
clearances of the control rings are greater than the adjacent tilting pad bearing
assemblies. The high stage compressor was also equipped with a traditional double acting Kingsbury type thrust bearing. A single thrust disk was keyed to the
shaft, and restrained axially by the stationary thrust bearing assembly. The coupling between the pinion and the high pressure compressor was a gear type, and
the hubs were keyed to their respective shafts.
Since this type of service is susceptible to internal coke buildup on the
wheels and diaphragms, performance data is acquired on a regular basis. In
some instances, injection or wash oil flow rates are increased to maintain head
and efficiency across each of the cracked gas compressors.
Following eight months of successful operation, it was noted that efficiency
Data Trending
719
Fig. 135 Orbit And Time

Base Of High Stage Compressor Discharge Bearing
Under Normal Operating
Conditions
was decreasing across the high stage compressor. It was anticipated that coke
buildup was responsible, and injection oil rates were increased. Shortly thereafter, the outboard radial bearing began to display intermittent vibration alarm
conditions. During periods of normal vibration levels, the average shaft motion
was in the vicinity of 1.0 Mil,p-p as shown in Fig. 13-5. During this normal condition dominant motion occurred at rotational speed, and the shaft exhibited a
slightly elliptical orbit, with a normal forward precession (i.e., with rotation).
Initially, the vibration alarm events occurred sporadically. Eventually, the
alarm conditions occurred with reasonable regularity early every morning. It
was observed that vibration levels at this bearing would begin to increase at
Fig. 136 Orbit &Time Base Of Discharge

Bearing With Subsynchronous Motion
Fig. 137 Spectrum Plots Of Discharge

Bearing With Subsynchronous Motion
720
Chapter-13
approximately 4:00 AM. On most days, by 5:00 to 5:30 AM the first alert setpoint
of 2.5 Mils,p-p was exceeded. Typically, by 7:00 AM the vibration levels were
decreasing. By 9:00 AM, the high stage compressor was back to normal behavior,
and it generally remained that way until the next morning. From a management
standpoint, this behavior was viewed as a non-problem. That conclusion was
attributed to the fact that by the time the morning meetings were concluded, the
high vibration condition had subsided for the remainder of the day.
During one of the early morning excursions, dynamic shaft vibration was
observed and documented on magnetic tape. The orbital and time domain information during this period is presented on Fig. 13-6. The upper data set in this
diagram displays the unfiltered vibration characteristics. The bottom orbit and
time base traces documents the shaft motion at all frequencies except rotational
speed (i.e., 1X notch filter). From this data, it is evident that the major vibratory
activity occurs at a frequency located below running speed. This subsynchronous
motion is forward (with rotation) with a slightly elliptical shape.
Precise frequency identification of the subsynchronous component is made
in the spectrum plot of the vertical and horizontal probes, as shown in Fig. 13-7.
This data reveals that the subsynchronous component occurs at a frequency of
3,888 CPM, with a shaft rotative speed of 11,252 RPM. From a percentage standpoint, this low frequency motion is at 35% of running speed. It is noteworthy that
this is the initiation frequency, and that continued operation resulted in a gradual increase in the subsynchronous frequency. In fact the final frequency was at
42% of rotative speed (4,720 CPM). Shortly after attaining the 42% condition,
the subsynchronous component decayed on schedule at 9:00 AM.
It was also noted that bearing temperatures were lower prior to the appearance of the subsynchronous excitation. Lube oil supply temperatures were logged
at 98F at the beginning of the cycle, and 106F at the end of the daily excursion.
Based upon the acquired data, the subsynchronous vibration was characterized
in the following manner:
1.
2.
3.
4.
5.
6.
Forward Precession
Nearly Circular Motion
Frequency Varies from 35% to 42% of Speed
Frequency Appears Below 1st Critical
Sensitive to Oil Supply Temperature
Repeatable, but Not Self-Sustaining
From this evidence, the self-excited vibration was considered to be an oil

whirl. Another theory that was suggested by one of the plant engineers was an
aerodynamic whirl between the rotor and diaphragms as coke buildup occurs.
However, the machine had no change in balance, and the potential for coke accumulation was discounted. In reality, the triggering mechanism was oil supply
temperature, and a problem within the outboard bearing was suspected.
It is often stated that tilt pad bearings cannot exhibit oil whirl. This statement is generically true since vertical to horizontal cross-coupling does not exist
Data Trending
721
in a tilt pad configuration (i.e., Kxy=Kyx=0). Without cross-coupling, the rotating

oil wedge cannot be sustained, and the shaft cannot whirl. However, the oil control rings on either side of the tilt pads provide an annular surface that could
sustain a self-excited whirl. This is possible if the clearance of the tilt pad assembly expanded, and the journal became supported on the control rings. Under normal conditions, the diametral bearing clearance should be 5 to 7 Mils. At
clearances approaching 20 Mils, the control rings could act as bearing surfaces.
It was concluded that this was the most plausible explanation for the whirl.
As mentioned earlier, the lube oil supply temperature was directly related
to the appearance of the whirl. As the reservoir cooled down each evening, a minimum oil temperature of 98F was observed. As the sun came up each morning,
the reservoir temperature increased and the whirl disappeared. Since this
behavior was the apparent triggering mechanism, it also provided the control
tool. Specifically, the reservoir heating coils were used to maintain a minimum
oil supply temperature of 110F throughout the night.
This approach was successful in controlling the whirl, and preventing a
reoccurrence. From a process standpoint, the ethylene plant remained in full
operation for the next few weeks until a short shutdown could be scheduled. A
one shift outage was planned to coincide with high product storage inventory.
The ensuing inspection revealed a lift check at this outboard compressor bearing
in excess of 0.014 Inches (i.e., twice the maximum allowable clearance). Although
the babbitted tilt pads were not damaged, the pad seats were worn, and overall
clearance was excessive. A replacement bearing assembly was installed with a
0.006 Inch overall diametral clearance. In addition, the oil control rings were
modified with axially milled slots every 60 to prevent these rings from allowing
any future whirl mechanisms. The train was successfully restarted, and the
plant experienced minimum downtime. Again, the application of condition monitoring allowed the selection of the most favorable options to maximize production
rates, and minimize maintenance and material costs.
Case History 46: High Stage Compressor Loose Thrust Collar
This problem addresses a four case machinery train operating in ammonia
plant syngas service. The train consists of a condensing steam turbine coupled to
a high pressure topping turbine, plus two barrel compressors. The low pressure
compressor is coupled to the topping turbine, and it drives the high pressure barrel. Normal train operating speeds vary from 10,400 to 10,800 RPM.
The high stage compressor is oriented with the suction nozzle, and thrust
assembly at the outboard end of the machine. The two radial bearings are five
pad, tilting pad assemblies. A double acting Kingsbury type thrust bearing is
installed around the single thrust disk. This thrust disk is keyed into position,
and secured with an outboard thrust lock nut.
Following an overhaul, this train operated at full load for several days.
Unexpectedly, radial vibration levels increased at the thrust (suction) end of the
high stage. The vibration monitor drifted up and down over a 2.4 Mil,p-p span for
722
Chapter-13
several hours before the abnormality disappeared. Although axial position

remained reasonably constant, the axial 1X vibration revealed a 0.9 Mil,p-p
swing. The high stage discharge bearing exhibited similar characteristics, but at
lower amplitude fluctuations of 0.8 Mils,p-p. The low stage compressor was
slightly influenced and both turbines remained unchanged.
Orbits at the high stage suction end bearing revealed a significant phase
change in addition to the variation in synchronous vibration. This is depicted on
Fig. 13-8 that shows the decay and initiation of this mechanism. This computergenerated four hour trend plot has a horizontal scale in minutes. The top diagram depicts 1X phase from the X-Axis probe at the compressor suction, with 1X
amplitude on the bottom trend plot from the same probe.
It is noted that 1X amplitudes vary from a minimum of 0.5, to a maximum
of 2.8 Mils,p-p. Simultaneously, the phase lag angle changed from 90 to 7. In
essence, the rotational speed vectors are constantly changing while this mechanism is active. Operating under constant process loads with only minor speed
changes, the fluctuating 1X vectors would come and go in a random manner.
Fig. 138 High Stage Syngas Suction Bearing - Four Hour Trend Plot Of 1X Vector
Thrust and radial bearing temperatures remained constant, and no particular

event could be identified to initiate or stop this unusual behavior. The time duration of this fluctuating vector also varied. In some instances the mechanism
would be active for twenty minutes, and at other times the fluctuation would persist for five to six hours. The only consistency was the amount of cycling once the
mechanism was active.
Since the observed motion was primarily a change in 1X vectors, it is reasonable to develop a summary vector diagram of the measured response characteristics. Fig. 13-9 shows such a plot for the suction end horizontal probe. Three
running speed vectors were plotted to develop this diagram. First, the normal
steady state vector of 1.7 Mils,p-p at 45 was drawn. Next, the maximum amplitude vector of 2.8 Mils,p-p at an angle of 32 was drawn. Finally, the minimum
amplitude vector was included at 0.5 Mils,p-p, at 60. It should be mentioned that
Data Trending
723
the respective phase angles associated with the maximum and minimum amplitudes were read off the trend plot in Fig. 13-8. Hence, some minor phase errors
are inherent with this estimation of the charted phase angles.
Fig. 139 High Stage

Syngas Compressor
Suction Bearing Vector
Summary Diagram
Nevertheless, the resultant plot reveals that the initial vector is positioned
between the minimum and maximum vectors. Drawing a circle from the head of
the initial vector through the heads of the identified minimum and maximum
vectors reveals a reasonably constant amplitude of 1.2 Mils,p-p. This is interpreted as a 1.2 Mil,p-p rotating vector that either adds or subtracts from the initial steady state vector. Actually, the circle represents the locus of resultant
vectors (i.e., vector summation between the initial vector of 1.7 Mils,p-p at 45
plus and minus the 1.2 Mil rotating vector). It is also noted that the phase angle
exhibits a minimum phase of 0 and a maximum phase angle of 92 that coincides with the trend plot on Fig. 13-8.
It was concluded that the measured behavior was driven by a 1.2 Mil,p-p
rotating vector. Since the compressor performance remained constant, the gas
flow characteristics were not influenced. Thus, it is reasonable to eliminate the
high stage compressor impellers as a potential problem source. Consideration
must then be given to some other rotating element. Obviously, the thrust disk at
the suction end of the high stage compressor is a potential candidate. This is
even more reasonable when it is recalled that axial vibration was also fluctuating by 0.9 Mils,p-p. Hence, any wobble of the thrust disk would result in a change
in axial vibration, plus a balance and/or gyroscopic response on the lateral
(radial) vibration characteristics. It was also discovered that the V groove on the
outer diameter of the thrust collar had been used for balance corrections.
Shutdown and inspection of this thrust assembly revealed a loose thrust
collar and a worn thrust lock nut. The thrust assembly was rebuilt, and the train
restarted within a single shift. Again, condition monitoring identified a specific
problem area, and repairs were executed in an orderly and cost-effective manner.
Clearly, condition monitoring can reduce maintenance labor and materials.
It is also reasonable to expect increased production stream factors due to more
724
Chapter-13
efficient and effective maintenance. Furthermore, proper machinery analysis

provides benefits for turnaround planning. In situations where a process unit is
shutdown for pressure vessel inspections, heat exchanger cleaning or re-tubing,
piping replacement, etc. condition monitoring can identify the machinery
items that require attention, as well as those inspections that are not necessary.
Overall, this is an important maintenance planning tool, and these techniques
can extend the time period between major overhauls. Today, there are many
machinery trains that are not subjected to annual overhauls. Some of these
machines have operated for five to ten years without a major disassembly. In an
ideal situation, a machine might continue to display parameters indicative of
good mechanical condition. Over the years, the performance of a machine may
degenerate due to widening interstage labyrinth clearances. Based on condition
monitoring, the machine would be opened and overhauled for a specific reason
(i.e., to restore the efficiency) rather than the old and expensive approach of:
while were down, why dont we open it up, and look inside?
PRE-STARTUP INSPECTION AND TESTING

After a machinery overhaul, many of the systems have been inoperative for
a period of time, or they have been disassembled. A successful startup, followed
by a trouble free extended run is dependent on the integrity of all devices. Thus,
it is necessary to methodically check each of the following machinery systems:
Process Alarms and Shutdowns

Lube and Seal Oil System
Speed Control System
Surge Control System
API 670 Machinery Protection System
It is the intent of this section to review each category, present some guidelines for system verification, and identify some of the potential pitfalls. For the
sake of consistency in addressing this topic, a generalized machinery train will
be discussed. Fig. 13-10 depicts a typical steam turbine driving a multistage centrifugal compressor. The turbine is not equipped with extraction or injection nozzles, and the compressor contains a single suction and a single discharge nozzle.
To facilitate discussion of the API 670 instrumentation, all traditional shaft sensing displacement probes are shown on this sketch.
Each of the five subsystems contains measurement functions, display and
monitoring capabilities, plus a myriad of alarms and shutdown features. In addition, the first four systems are engaged in active control functions. It is difficult
to rank the importance of one system over another, since malfunctions in any one
of the categories can cause a false shutdown. In all cases, it is desirable to check
each system independently, and develop a suitable testing and verification plan.
Before addressing the individual subsystems, it should be recognized that
the only safe way to stop the turbine and compressor set is to cut off the steam
Pre-Startup Inspection and Testing
Steam Inlet
T&T
725
Exhaust
Discharge
Anti
Surge
Suction
CW
Normal
Thrust
Normal
Thrust
K
1Aa
Multi Stage
Centrifugal
Compressor
Condensing
Steam Turbine
1Ab
1Y
1X
45 45
Rotation and
Probe Orientation
Viewed From
The Turbine
Governor
2Y
2X
45 45
3Y
3X
45 45
4Y
4Aa
4Ab
4X
45 45
CCW
Fig. 1310 Typical Steam Turbine Driven Centrifugal Compressor Machinery Train
supply. The ability to do this rapidly is provided by the trip and throttle valve
(T&T). This valve is used as a hand throttle valve for admitting steam to the turbine when bringing it up to minimum governor. The other, and perhaps the most
important function, is to rapidly terminate the steam inlet to the turbine during
a trip condition (typically in less than 0.3 seconds). The closure of the T&T valve
can generally be accomplished by any one of the following conditions:
Manual Trip
Turbine Overspeed Trip
Low Governor Oil Pressure Trip
Opening the Solenoid Dump Valve
Irrespective of the type of trip, the result is the same; that is, the virtually
instantaneous closure of the T&T valve. This valve contains a hydraulically operated mechanism that is activated by governor oil pressure. Thus, any of the
above four actions will relieve this pressure, and the valve will close. It must be
manually reset before it can be reopened. If governor oil pressure is not available, the valve remains closed. The various trips discussed in the following sections have one thing in common. That is, all of the trip signals have their final
termination point at this solenoid operated dump valve (SOV).
Furthermore, it is very important to consider the overall system time delay
between trip initiation and the closure of the T&T valve. Although the T&T valve
closes rapidly after the oil is relieved, the time delay in getting the signal from
the sensor to the trip solenoid may be intolerable. Based on the particular instruments installed, there may be a group of three or five second delays in series. In
some cases, system time delays of over thirty seconds have been measured. Generally, overall system delays in excess of five seconds are unacceptable.
Many of the newer steam turbine control systems include sophisticated
726
Chapter-13
electronic governor controls. In some installations this may be a dedicated Process Loop Controller (PLC) within the Distributed Control System (DCS), or it
may be a completely separate free standing console dedicated to controlling
machine speed. A separate machine console is even more likely for complex drivers such as gas turbines that must incorporate a variety of firing controls and
safety backup systems into the control scheme. In any configuration, the turbine
control system must be capable of rapidly shutting down the turbine in case of
emergency and this capability should be thoroughly check and physically verified during each maintenance turnaround.
In addition to the normal process control loops, many devices are installed
with the specific purpose of preventing damage to the machinery trains. These
process related alarms and shutdowns typically provide two independent
setpoints. The first point is an alarm that may annunciate locally, or in conjunction with a parallel control room alarm. If action is not taken to correct the
impending problem, the second setpoint is usually an automated trip of the
machinery train. In other cases, the second setpoint may not be an instrument,
but it may be a relief valve or a rupture disk that protects the machinery from an
overpressure condition. Process problems can cause severe damage to gas handling equipment. This includes the ingestion of liquids, plus excessive high or
low compressor casing temperatures. The following list represents typical process instruments that are wired for alarm and machinery train shutdown:
Compressor Suction Drum Level
Low Level Alarm
High Level Alarm
High Level Shutdown
Compressor Suction Pressure
High Pressure Alarm
High Pressure Relief and Shutdown
Compressor Suction Temperature
High Temperature Alarm
High Temperature Shutdown
Compressor Discharge Pressure
High Pressure Alarm
High Pressure Relief and Shutdown
Compressor Discharge Temperature
High Temperature Shutdown
Condensate Hotwell Level
High Level Alarm
Spare Hotwell Pump On Alarm
High Level Shutdown
Prior to startup of a machinery train, the level transmitters should be

checked by building a suitable high or low level within the respective suction
727
drum or turbine hotwell. This level can usually be verified by direct observation
in the attached gauge glass. For the shutdown points, it is mandatory to verify
that the T&T valve does shut. It is convenient to just check the transmitter output. However, one is never really sure of a trip until the T&T valve slams shut.
With a high level in the hotwell, the spare condensate pump should be
allowed to auto-start, and verification should be made that the alarm does function. In some installations the main condensate pump may be steam turbine
driven, and the spare unit motor driven. In other plants, both the main and the
spare condensate pumps are motor driven. In all cases, each unit that is
equipped for automatic startup based on high condensate level should be tested
for proper operation with a high water level in the hotwell.
The compressor temperature alarms and shutdowns are usually simulated
electronically at the thermocouple junction. Again, it must be verified that the
T&T valve actually does close at the proper trip temperatures. Pressure transmitters may be tested with dead weight or other hydraulic pressure devices.
Rupture disks cannot be field tested, but relief valves should be removed and
bench tested on a scheduled basis.
Due to the complexity of the turbine trip signals, it is necessary to have an
updated ladder diagram of all trip relays and any associated interposing relays.
This relay sequencing and logic generally cannot be deduced from the field wiring. An updated logic or ladder drawing is mandatory for proper verification of
the system. In addition, the logic diagram does not do any good locked away in
somebodys office or file cabinet. In all cases, the logic diagram should be available in the control room with twenty-four hour a day accessibility. In many
instances, it is highly desirable to mount a current copy of this diagram on the
wall or inside the cabinet door of the trip instrumentation (e.g., relay cabinet).
The lube and seal oil system generally provides three basic functions for
the process machinery. First, it supplies lubricating oil to the machine train
bearings. For the example train shown in Fig. 13-10, this would include the four
journal bearings, the two double acting thrust bearings, and the gear coupling.
Second, the oil system delivers governor oil to the T&T valve. Third, the oil console provides seal oil to the overhead seal oil tank to maintain a nominal fifteen
foot head above the compressor shaft seals. Some operating companies prefer to
divide these functions into separate lube oil and seal oil consoles for each train.
Other end users elect to combine these functions for several machinery trains
into one large lube and seal oil system.
Regardless of the specific console configuration, the required system checks
are in accordance with normal maintenance practices. This includes traditional
level, pressure, and temperature tests to insure that the various alarms and
trips are functioning properly. All control valves should be stroked and packing
checked. Filters should be changed, and the oil reservoir should be verified for
cleanliness. Oil coolers should also be checked to insure that both the shell and
the tube side are clean and suitable for continued service. Also, the level floats in
the sour oil traps should be verified for proper condition and operation. Sometimes it is easy to ignore the oil console due to the built-in redundancy of spare
pumps, coolers, filters, etc. However, a thorough flushing and system operational
728
Chapter-13
check is quite comforting, and it may pay big dividends in terms of extended unit
operation. For reference purposes, the normal alarms and shutdowns associated
with a typical lube and seal oil system are summarized as follows:
Lube Oil Pressure
Low Pressure Alarm
Start Auxiliary Oil Pump Alarm
Low Pressure Shutdown
Seal Oil Pressure
Low Pressure Alarm
Start Auxiliary Seal Oil Pump Alarm
Seal Oil Level in Overhead Tank
High Level Alarm
Low Level Alarm
Start Auxiliary Seal Oil Pump Alarm
Low Level Shutdown
Lube & Seal Oil Reservoir
Low Level Alarm
Low Temperature Alarm
Governor Oil Pressure
Low Pressure Alarm
The turbine speed control system is designed in accordance with the

requirements of the particular process. In many instances, speed is regulated by
compressor suction or discharge pressure. The turbine governor accepts this
input signal, and regulates the angular position of the rack (camshaft), which
controls the sequential opening and closing of the individual steam inlet valves.
Mechanical linkages should be checked for proper positions and pin
arrangements. It is not unusual to miss a linkage by one hole, and not be able to
run the train above minimum governor speed. The older oil relay governors are
being replaced by electronic governors. These units have excellent control capabilities, but they do require accurate speed sensing. It is customary to use a magnetic probe observing a multiple tooth gear. These sensors can be damaged
during an overhaul and not noticed until the turbine is ready to run. If possible,
install one or more redundant probes, and check the transducer outputs with a
portable test rig, or during initial slow roll.
The last or final speed control device is the mechanical turbine overspeed
trip. The setting and repeatability of this device should be verified during the
turbine solo run. Generally, three overspeed trips are made to establish confidence in the mechanism. It is usually desirable to measure the actual trip speed
with a digital tachometer set to capture the maximum RPM. Since the actual
overspeed trip occurs in a very short period of time, the assistance provided by
the peak speed sensing capability of a digital tach is highly desirable.
729
Surge control systems are installed on many centrifugal compressors.

These machines are susceptible to a low flow, high head phenomena known as
surge. On a standard head capacity curve, the surge point is located at the lefthand edge of the curve. When various operating speeds are considered, a family
of curves are produced, and the surge line forms the left-hand, or low flow, limit
for these curves. If the compressor is operated past this surge line, the system
back pressure exceeds the compressor discharge pressure. At this point, flow is
momentarily reversed through the machine. This action lowers the system back
pressure, and forward flow is re-established. Unless some flow or pressure
changes are made, this cycle will continue until some mechanical element fails
and aborts the cycle. Some common techniques for surge control are as follows:
Vent to Atmosphere
Vent to Flare
Inlet Control or Variable Guide Vanes
Speed Control or Reduction
Bypass to Lower Pressure (e.g., Suction)
The atmospheric vent is generally unacceptable for anything but air compressors. A vent to flare is a waste of resources and is generally frowned upon by
the local residents. Inlet control or variable guide vanes are expensive from an
initial procurement standpoint and from the costs associated with periodic maintenance. Speed reduction is a reasonable candidate, assuming that the control
and mechanical systems can respond within an effective time frame. The last
option of bypassing gas to a lower pressure is the technique generally applied on
machinery trains such as the example in Fig. 13-10. In this simple system, gas
from the compressor discharge is returned to the suction. Flow control opens the
anti-surge valve, and higher flow rates are maintained through the machine.
Thus, the compressor operates at a comfortable distance from the surge line. In
most instances, cooling is required on the final discharge stream to prevent
excessive heat buildup during operation of this kick back loop.
In this traditional control scheme, a flow or anti-surge controller is used to
activate the bypass valve. Often these devices are field calibrated at several
speeds. Specifically, the compressor is briefly placed in surge, and the associated
flow rates and speeds are documented. The surge occurrence is verified by differential pressure measurements and shaft vibration data. Following the acquisition of three to five data points the controller is set to open the anti-surge valve
well in advance of the empirically determined surge curve.
Following an overhaul, the surge controller can be calibrated, and verified
electronically to duplicate the previous test data. In addition, the valve stroke
and response time can be checked. Due to the variety of system configurations,
and instruments, the specific test details should be developed for each individual
system. As more complicated machine trains are considered with multiple cases,
and multiple side streams per case, the surge control problem increases exponentially. The resultant anti-surge system may be a combination of two or more of
the previously discussed options. Again, the calibration and checkout procedures
730
Chapter-13
are often unique to each system.

The highest concentration of alarm and shutdown points typically resides
within a fully armed API 670 machinery protection system. For the example
shown in Fig. 13-10, a total of thirteen displacement proximity probes are
installed. Based on this array of transducers, the following list summarizes the
potential alarms and shutdowns on the steam turbine driver:
Turbine Speed Alarm
Low Slow Roll Speed Alarm
High Operating Speed Alarm
Electronic Overspeed Shutdown
Turbine Thrust Position
Active Thrust Alarm
Inactive Thrust Alarm
Active Thrust Shutdown
Inactive Thrust Shutdown
Turbine Inlet Bearing
Radial Vibration Alarm
Radial Vibration Shutdown
Turbine Exhaust Bearing
A similar set of alarms and trips may be employed on the centrifugal compressor as summarized in the next list:
Compressor Discharge Bearing
Compressor Suction Bearing
Compressor Thrust Position
Active Thrust Alarm
Inactive Thrust Alarm
Active Thrust Shutdown
Inactive Thrust Shutdown
The four journal and two thrust bearings are equipped with embedded thermocouples or RTDs. In addition, the ambient and lube oil supply temperature
may also be monitored with permanently installed temperature sensors. These
transducers are often terminated in the same rack as the digital tachometers,
plus the vibration and thrust position monitors. A typical array of temperature
alarms are described in the following list:
731
Turbine Thrust Bearing Temperature

Active High Temperature Alarm
Inactive High Temperature Alarm
Turbine Inlet Bearing Temperature
Turbine Exhaust Bearing Temperature
Compressor Discharge Bearing Temperature
Compressor Suction Bearing Temperature
Compressor Thrust Bearing Temperature
Active High Temperature Alarm
Inactive High Temperature Alarm
Auxiliary Temperature Measurements
Low Lube Oil Temperature Alarm
High Lube Oil Temperature Alarm
Low and/or High Ambient Temperature Alarm
Due to the quantity of transducers involved, and the various possibilities

for automatically tripping the machinery, considerable care must be exercised in
system checkout and verification. It is presumed that the original system installation was documented with the following minimum descriptive information:
Probe Calibration Data

Probe Installation Diagrams
Machine Train Wiring Diagram
Plot Plan of Cable Runs
Machine to Monitor Wiring Diagram
Monitor Arrangement Diagram
Monitor Back Plane Wiring Diagram
Shutdown Logic and Hardware Diagram
Calibration should be verified for all the thrust probes, and a reasonable
sampling of the radial probes. Transducer systems that do not meet the minimum API requirements for a linear range of 80 Mils, or sensitivity of 200 mv/Mil
5% should be discarded. Whenever possible, the calibration checks should be
run on the shaft material. Thrust probes should be installed and gapped in accordance with predetermined position values. In all cases, the thrust probes shall be
installed with concurrent agreement between the distance measured by probe
gap voltages, thrust monitor readings, and axial dial indicator readings. The
three values must agree as the rotor is moved between active and inactive thrust
shoes. It is also a good idea to rotate the rotor 180 and repeat the thrust position
measurements. There should be little if any difference between the 0 and the
180 thrust position readings.
732
Chapter-13
Continuity between each probe and the assigned monitor location shall be
verified. In addition, a dynamic AC vibration signal should be generated at each
radial probe location, and the respective vibration monitor checked for proper
values, and minimal signal noise. Where required, calibration shall be performed
to meet normal tolerances. All vibration transducers shall be checked for proper
field isolation, and single point grounding at the intrinsic safety barriers or the
monitor rack (whichever is appropriate). Also, the probe installation hardware
shall be reviewed for suitability, durability, and functional rigidity.
The associated trip circuitry shall be physically as well as functionally
checked. Particular attention shall be paid to relay connections. The standard
nomenclature of NC (Normally Closed) and NO (Normally Open) can be quite
confusing unless the intended relay state (energized or de-energized) is also
defined. The functional checks shall include the generation of AC and DC signals
to active alarms and trips on each channel. All alert (alarm) and danger (shutdown) setpoints on each monitor shall be verified. A shutdown (trip) indication is
not completed until the T&T valve is closed. Also, any voting logic between transducers should be checked for proper and consistent operation.
All probe to pigtail connectors should be securely wrapped in a suitable
insulating material to prevent any possibility of stray ground loops between the
probes and monitors. Insulating materials such as Scotch 70 Self-Fusing Silicone Rubber Electrical Tape is far superior to fragile materials such as teflon
tape. Finally, all changes in hardware, configurations, or calibration shall be permanently documented in the system file. Ideally, this would occur before startup
of the machinery train.
From these discussions, it is apparent that the five auxiliary systems are
critical to the operation of the main machinery train. If these systems are
ignored, the best possible situation would be a series of nuisance trips. In the
worst possible situation, the protective trips would malfunction during a real
emergency condition. The required automated trip would not occur, and a catastrophic machine failure might be the final result.
STARTUP INSPECTION AND TESTING

At the completion of an overhaul, and the conclusion of the inspections and
system tests, the moment of truth eventually arrives. That is, the time has come
to startup the unit. With motor or turbine driven equipment, the first phase is to
run the driver solo. Even though the driver has no load on it, the spin up does
verify that the unit is capable of attaining operating speed. For turbine drives,
the solo run provides an opportunity to check overspeed trips. In addition, various types of data can be obtained, and operational checks may be performed on
the auxiliary systems. For motor drives, this solo run provides the opportunity to
mark the axial magnetic center of the rotor running within the stationary field.
The solo run is followed by the second phase that consists of a coupled train
startup. The load carrying capability of the driver is tested for the first time, and
the driven piece of machinery is subjected to an initial trial run. More behavioral
Startup Inspection and Testing
733
data can be collected during this startup and an initial assessment made of the
machinery condition. If the equipment was subjected to an overhaul this startup
vibration data is vital for future documentation, and for providing the transient
speed base for future condition monitoring.
Assuming that the machinery has successfully passed through the first two
phases, the final test is imposed. This third phase addresses the thermal heat
soak, combined with process and load stabilization. The time duration of this last
startup phase is difficult to quantify due to the number of potential problems
that may be encountered. In very general terms, a twenty-four hour period might
be considered a reasonable norm. At the end of this third phase, a good evaluation of the machinery condition and behavior should be possible. Hence, a new
data point of performance and vibration response would be established for continued steady state condition monitoring.
It is the intent of the following sections to review the vibration measurements made during variable speed and load transients. The different data acquisition and documentation requirements will be discussed, and each of the three
startup phases will be reviewed in the form of a case history. Performance measurements as discussed earlier in this chapter would not be applied until constant speed and load has been attained. Hence, the majority of the following
discussion will center around the transient vibration measurements.
Various discussions of diagnostic vibration hardware have been presented
throughout this text. Clearly, the basic tools required for constant speed, full
load, data observation and acquisition include an oscilloscope, Digital Vector Filter (DVF), Digital Signal Analyzer (DSA), plotter, and a multimeter. Supplemental filters, meters, interface devices, and various levels of computer-assisted data
handling are useful additions. However, this basic hardware provides 95% of the
data documentation capability required for steady state conditions. Using these
tools, the information may be formatted as orbit and time base data, spectrum
plots, plus axial and radial rotor position plots.
As noted in previous chapters, the oscilloscope may be used to document
unfiltered shaft orbits and time base information. Inserting a tracking filter
between the transducer signals and the oscilloscope provides the capability for
1X filtered data. The spectrum analyzer allows an examination of the frequency
content of the signals. The multimeter is used to measure probe gap voltages.
From this DC voltage data, shaft position plots may be generated.
Generally, this class of instrumentation will only handle one or two data
channels at a time. This is acceptable when machine conditions are constant.
However, during a startup there are many transducers to simultaneously examine. Even the simple turbine compressor set shown in Fig. 13-10 requires 13 data
channels. In these situations, it is impossible to properly observe all channels
during the startup sequence. Assistance is required to handle this task, and the
multichannel tape recorder or direct digital data storage is the proper device for
this type of transient information.
From a data integrity standpoint, the tape or digital recorder should be considered as just another tool. It does not replace other instruments, it just provides multichannel recording and playback capabilities. In some situations this
734
Chapter-13
can be extremely helpful. However, if the entire database is committed to a single recorder, and the recording device malfunctions, the data may be lost forever.
The safest approach is to use a combination digital and analog data acquisition
system (e.g., Fig. 8-5). The recording may be supplemented by on-line hard copy
data, or appropriate log sheets. This provides some level of backup, and it also
improves visibility of machine behavior since the diagnostician is looking at the
dynamic signals. That is certainly more productive than sitting and watching the
reels spin. In addition, the hard copy data can be compared against the reproduced recorder signals and calibration verified. Besides, if recording redundancy
is combined with on-line hard copy data, the recorders will probably never fail.
It should also be mentioned that during a startup, the instrumentation
array can be used in a variety of manners. For example, the oscilloscope may be
used to observe output signals from monitors by moving the leads from jack to
jack. The scope may also be used to verify signals in and out of the tape recorder
to verify proper recorder operation. The DVF tracking filter may be used to log
1X vectors from different probes at different hold points, or it may be dedicated
to plotting data continuously from one specific transducer. Similarly, the DSA
spectrum analyzer may be used for examining inputs from live transducers.
It is almost mandatory to produce hard copy documentation of the slow roll
characteristics from each pair of proximity probes. Since these transducers
detect shaft surface imperfections, changes in permeability of the observed surface, localized magnetism, etc., it is necessary to define the slow roll, or runout
characteristics. Ideally this would be a combination of orbit/time base plots combined with logs of rotational speed vector amplitudes and angles.
Also, DC gap voltages from all probes should be logged at identifiable conditions. For example, the log should include gap voltages at stop before oil circulation and with oil circulating. The next point would be during slow roll on the
machinery. Then there may be two or three plateau regions on the way up to
speed. Certainly, DC gaps at minimum governor and full speed data should be
obtained, and a final set acquired under full load and full heat soak.
The recorded information may be reproduced into a variety of analog
instruments or a computed-based data processing system (e.g., Fig. 8-7). If the
data was initially stored in a digital format, it may be subjected to additional
manipulation without the need for analog instruments. Once again, synchronous
vectors are accommodated by the Bode plot, where rotational speed vibration
amplitude and phase angle are simultaneously plotted as a function of machine
speed. The Bode is combined with a polar plot describing the locus of rotational
speed vectors during variable speed operation. Although both of these plots provide the same basic data array, the Bode provides excellent visibility of changes
with respect to speed, and the polar plot yields improved resolution of phase
variation. Data of this type is essential for identifying rotor critical speeds and
the influence of various resonances. Under machine conditions where significant
subsynchronous or supersynchronous vibration components are generated, it is
desirable to generate a cascade plot of individual spectra at incremental operating speeds. This type of data presentation provides an excellent overview of the
frequency content of the vibration signals as a function of speed. Spectrum data
735
may also be processed as a function of time in a waterfall plot. In addition, individual frequency components or orders may be tracked with respect to speed or
time, as required.
The reader is referenced back to chapter 8 for a detailed discussion of the
types of analog and digital instrumentation systems, and the traditional data
presentation formats. For situations where high resolution, rapid data processing, significant post processing requirements, or extensive data manipulation is
anticipated, a combination between analog and digital systems may be mandatory. Based on this brief review of transient data formats, it is reasonable to discuss three case histories and observe actual machinery vibration response
characteristics under various normal and abnormal conditions.
Case History 47: Turbine Solo Operation with Tapered Journal
This case study addresses a twenty-year-old steam turbine solo run. In this
example, a four stage turbine was subjected to a thorough overhaul. The stationary diaphragms were replaced, a spare rotor was installed, and new seals and
bearings were installed. This unit had a good operating history, and minimal if
any problems were anticipated. During the turbine solo, the machine was operated from a slow roll of 550 to an overspeed trip at 9,510 RPM. The response data
from the governor end horizontal probe is presented in Fig. 13-11. The bottom
portion of this Bode plot displays rotational speed (1X) amplitude, and the top
plot exhibits phase angle as a function of speed. An unusually large runout vector of 2.32 Mils,p-p at 67 was documented. This vector was subtracted from the
sampled data, and dual plots of the runout corrected (or compensated), and the
uncompensated (or direct) traces are displayed in Fig. 13-11.
Fig. 1311 Turbine Solo Run - Lateral

Response At Governor End Radial Bearing
Fig. 1312 Turbine Solo Run - Axial

Response At Governor End Thrust Bearing
736
Chapter-13
Prior to startup, the governor end diametrical bearing clearance was measured to be approximately 6.5 Mils. That value was exceeded as the turbine
passed through the translational first critical at a speed of 3,500 RPM with an
amplitude of 7.5 Mils,p-p. At operating speeds above 7,000 RPM, a steadily
degenerating situation was quite evident as 1X radial amplitudes continued to
increase. Observation of rotor axial vibration during a solo run is normally
uneventful. However, the axial response characteristics of this turbine were
quite unusual. The direct and runout compensated Bodes are shown on Fig. 1312. It is evident that the sharp peak at 3.8 Mils,p-p occurs at 3,500 RPM, which is
coincident with the 7.5 Mil,p-p peak observed on the lateral Bode plot.
In most cases, it is normal to detect a minor level of lateral to axial crosscoupling in a rotor system. However, the observed ratio of almost 2:1 is unusual.
In addition, the high amplification factor of the axial response was judged to be
abnormal. Fortunately, the turbine was tripped by the overspeed trip assembly
and was not restarted. As always, some individuals wanted to couple up the turbine to the compressor on the basis that the compressor would calm down the
turbine. This is nothing but wishful thinking, and the machinery diagnostician
should always remember that you dont go on to the next step until the current
step or test has been successfully completed.
In this case, the turbine was shutdown following this single uncoupled
(solo) run. A subsequent bearing inspection revealed a wiped governor end radial
bearing, and a polished exhaust end bearing. The rotor was pulled for further
inspection, and it was determined that the governor end journal was tapered. In
fact, the diameter difference across the width of the bearing was in excess of 3.0
Mils. It was postulated that the bearing could not develop a proper oil wedge
across the axial length of the bearing. This journal taper plus the lack of a proper
radial support evidently resulted in the abnormal lateral and axial behavior.
This was a case where a machining oversight could have manifested into a
major turbine failure. Fortunately, the peculiarities of the vibration data called
attention to the mechanical problem during the uncoupled solo run of the turbine, and a suitable correction was achieved by re-grinding the turbine journal.
Case History 48: Coupled Turbine Generator Startup
The next case history considers a steam turbine driven generator set. In
this train, the turbine had a very acceptable solo run, but the generator had a
slightly rougher time during the coupled startup. The cascade plot presented in
Fig. 13-13 was acquired from the horizontal probe mounted at the exciter (outboard) end of the generator. Note that the individual spectra are reasonably
clean, with the dominant motion occurring at rotational speed (1X). The minor
harmonics of this fundamental frequency are primarily attributable to shaft surface imperfections rather than actual rotor vibration. Since the rotor surface
observed by the proximity probe was external to the bearing housing, and
exposed to the atmosphere, this type of signal noise should be expected.
Further visibility of the horizontal 1X motion is obtained from the Bode plot
737

Of Generator Startup On
Outboard Exciter End
Journal Bearing
shown in Fig. 13-14, and the polar plot of Fig. 13-15. Notice that the Bode displays two distinct amplitude peaks and associated phase excursions between
1,000 and 2,000 RPM. In this case, the observed response may be representative
of a split critical. That is, there may be a discernible difference in the vertical
versus horizontal rotor support stiffness. Since the horizontal stiffness is generally lower, the peak at 1,350 RPM might be the horizontal first critical speed,
and the response at 1,650 RPM might be representative of a vertical first critical.
Thus, the 300 RPM difference between peaks might indicate a split critical.
The same type of behavior could also be due to two closely spaced reso-
Fig. 1314 Bode Plot Of Generator Startup On Outboard Exciter End Bearing
Fig. 1315 Polar Plot Of Generator Startup On Outboard Exciter End Bearing
738
Chapter-13
nances that are representative of two independent modes. In some cases, there
may be a modal coupling between modes and the motion from one resonance
may excite or drive the other resonance. This type of modal coupling usually
appears on higher order criticals (e.g., 2nd and 3rd).
The additional phase angle resolution offered by the polar plot (Fig. 13-15)
reveals a slightly different perspective of the same vector data. Note that the
large polar loop represents the horizontal resonance at 1,350 RPM, followed by
the smaller inner loop that is the critical at 1,650 RPM. Since the polar loops are
coincident, this information begins to look like a pair of rotor resonances. However, another interpretation of this data might be a major rotor resonance followed by a minor secondary or structural resonance. In essence, examination of
the Bode and polar plots from one probe results in three possibilities for the
peaks at 1,350 and 1,650 RPM. As discussed, this could be attributed to:
Split Horizontal and Vertical Critical
Independent but Closely Coupled Shaft Resonant Modes
Shaft Critical followed by Secondary or Structural Resonance
The available data is insufficient to provide any additional clarification of

the origin of the two resonances. However, when the Bode and polar plots for the
other three radial probes on the generator are examined, it becomes clear that
the behavior is due to a split (horizontal then vertical) translational critical
speed. The presented plots (Figs. 13-14 and 13-15) are representative of the general behavior. However, examination of the overall machinery response from the
X-Y probes at both the coupling and the exciter end is necessary to fully identify
the modal behavior of this generator. This type of data is certainly beneficial for
condition monitoring, and it must be considered as necessary information for
evaluation of any future problems on this generator.
It should also be mentioned that this turbine generator set was installed on
an adequate foundation. The casing motion was minimal, and shaft relative
proximity probes were sufficient to fully describe the machinery response characteristics. However, if the foundation was softer (i.e., flexible or compliant), the
casing motion would have to be measured and considered. The inclusion of casing absolute with the shaft relative measurements allows the determination of
shaft absolute vibration as previously discussed in chapter 6. This does complicate the data, but sometimes the ability to accurately evaluate the behavior of a
machine is directly related to the thoroughness of the database.
On this generator, it is clear that the unit was fairly active through the resonant region. A trim balance correction would improve these synchronous characteristics through the criticals. However, the 1X vectors at full speed were all in
the vicinity of 1.0 Mil,p-p. These observed shaft displacement amplitudes are
quite acceptable when compared against the 18 Mil diametrical bearing clearance of both generator bearings. Hence, the unit remained on-line, and the generator was producing full rated power in less than four hours.
739
Case History 49: Heat Soak and Load Stabilization

The final phase in most machinery startups consists of a heat soak (twelve
to forty-eight hours) combined with process and load stabilization. Small
machines with thin cases and lower operating temperatures may stabilize in
twelve hours or less. Conversely, larger units with thick machinery cases, and
higher operating temperatures may easily require up to forty-eight hours to
achieve an acceptable operating equilibrium. During this condition, machine
speeds are fairly constant. Speed changes are generally performed a few RPM at
a time, and the rapid data sampling rates associated with the startup conditions
are no longer required. In many cases, the machinery behavior may be adequately examined and documented by reverting back to steady state data acquisition and processing techniques.
For instance, consider a steam turbine driven compressor following a routine startup. After attaining an initial speed of 5,054 RPM, the orbit and time
domain plots across the coupling are presented in Figs. 13-16 and 13-17. This
data was consistent with previous measurements, and it was considered to be
indicative of normal mechanical behavior. As noted in Fig. 13-16, the turbine
exhaust orbit was elliptical and predominantly horizontal. The maximum unfiltered horizontal probe amplitude was 1.85 Mils,p-p. This value was identical with
previous startups. The predominantly flat orbit was indicative of a vertical preload, and historically this had been attributed to cold misalignment across the
turbine. From previous startups, the turbine exhaust end orbit would always
evolve into a circular pattern as the front of the turbine warmed up to a normal
operating elevation.
Across the flexible coupling, the compressor inboard bearing had vibration
amplitudes of slightly less than 1.0 Mil,p-p. The motion was forward, circular, and
Fig. 1316 Turbine Coupling End Bearing

Immediately After Train Startup
Fig. 1317 Compressor Coupling End

Bearing Immediately After Train Startup
740
Chapter-13
quite stable. This documented orbital and time domain data was directly comparable with previous information. As the turbine achieved a uniform heat soak,
the lower temperature compressor casing exhibited minimal thermal growth. In
fact, as time progressed, the compressor coupling end vibration normally did not
change appreciably from the patterns shown in Fig. 13-17.
Many machinery startups are achieved after long hours of correcting
numerous instrumentation and system problems. Usually, when the equipment
is up and running on the governor, the problems are generally over. However, on
this particular machinery train, symptoms of substantial coupling misalignment
appeared 25 hours after startup. The turbine exhaust shaft vibration increased
somewhat to maximum amplitudes of 3.4 Mils,p-p as shown on Fig. 13-18. Concurrently, the horizontal probe on the coupling end of the compressor increased
from slightly less than 1.0 to 6.1 Mils,p-p as documented in Fig. 13-19.
Fig. 1318 Turbine Coupling End Bearing

25 Hours After Initial Train Startup
Fig. 1319 Compressor Coupling End

Bearing 25 Hours After Initial Train Startup
Operating at a slightly lower speed of 5,010 RPM, both rotors exhibited

severely pre-loaded (flat) orbits. In addition, the two orbits are essentially out-ofphase with each other. This behavior is most graphically illustrated by a direct
comparison of the 1X filtered orbits. The turbine exhaust displays a Keyphasor
mark at the upper right-hand corner of the orbit. At the same time, the compressor inboard reveals a Keyphasor dot at the lower left side of the orbit. This
behavior is representative of a pivotal motion across the coupling, and it is probably associated with a locked up coupling.
In this abnormal situation, the culprit was traced to the large diameter
compressor suction line. A pipe shoe somehow worked its way off a support, and
this section of the suction line dropped several inches. This overhung weight provided a tremendous torque on the compressor casing, and produced the observed
misalignment. Fortunately, the problem was detected rapidly, and the machinery
741
train was shutdown in an orderly fashion.

Mechanical inspection revealed that the gear coupling and the compressor
inboard bearing required replacement. Both turbine bearings and the compressor outboard bearing were not damaged. The thrust bearing on each case had
experienced some elevated temperatures, but the shoes were found to be in good
shape. Following completion of these repairs, the unit was successfully restarted,
and placed on-line. In the final analysis, it was concluded that scheduled condition monitoring performed one day after startup averted any further damage to
the machinery or the associated facility.
Based on these discussions, and the case histories, it is apparent that condition monitoring is a valuable and viable tool during startup situations. The fundamental logic is similar to the observation of machinery characteristics at
constant speed and load. Specifically, machinery in normal condition does
respond in a fairly repeatable and consistent manner. Mechanical problems or
developing failures will alter these typical response characteristics. Since these
variations are detectable, they are also identifiable, and therefore correctable.
742
Chapter-13
BIBLIOGRAPHY
1. ASME Performance Test Codes Code on Steam Turbines, PTC-6, The American
Society of Mechanical Engineers, (New York: The American Society of Mechanical
Engineers, 1976, reaffirmed 1982).
2. ASME Power Test Codes Test Code for Compressors and Exhausters, PTC-10,
The American Society of Mechanical Engineers, (New York: The American Society of
Mechanical Engineers, 1965, reaffirmed 1986).
3. Eisenmann, Robert C., John East, and Art Jensen Short Course 1 - Inspection and
Overhaul of Major Turbomachinery. Proceedings of the Seventeenth Turbomachinery Symposium and Short Courses, Turbomachinery Laboratory, Texas A&M University, Dallas, Texas (November 1988).
4. Mitchell, John S., Introduction to Machinery Analysis and Monitoring, second edition, pp. 291-345, Tulsa, OK: Pennwell Publishing Company, 1993.

Machinery Malfuntion Diagnosis and Corretion

Uploaded by

Copyright:

Available Formats

Machinery Malfuntion Diagnosis and Corretion

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Machinery Malfuntion Diagnosis and Corretion

Uploaded by

Copyright:

Available Formats

Machinery Malfunction

Diagnosis and Correction

Robert C. Eisenmann, Sr., P.E.

Robert C. Eisenmann, Jr.

Library of Congress Cataloging-in-Publication Data

Author: This space is reserved for Library of Congress

Original Library of Congress Cataloging in Publication Data

1997 PTR Prentice Hall

Prentice-Hall International (UK) Limited, London

To Mary Rawson Eisenmann,

Chapter 2 - Dynamic Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Chapter 3 - Rotor Mode Shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

Chapter 4 - Bearings and Supports . . . . . . . . . . . . . . . . . . . . . . . . 149

Bearing Supports Measurements and Calculations 179

Chapter 5 - Analytical Rotor Modeling . . . . . . . . . . . . . . . . . . . . 199

Chapter 6 - Transducer Characteristics . . . . . . . . . . . . . . . . . . . 243

Chapter 7 - Dynamic Signal Characteristics . . . . . . . . . . . . . . . 303

Chapter 8 - Data Acquisition and Processing . . . . . . . . . . . . . . . 365

Chapter 9 - Common Malfunctions . . . . . . . . . . . . . . . . . . . . . . . 395

Chapter 10 - Unique Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459

Chapter 11 - Rotor Balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535

Chapter 12 - Machinery Alignment . . . . . . . . . . . . . . . . . . . . . . . . 637

Chapter 13 - Applied Condition Monitoring . . . . . . . . . . . . . . . . 703

Chapter 14 - Machinery Diagnostic Methodology . . . . . . . . . . . 743

Chapter 15 - Closing Thoughts and Comments . . . . . . . . . . . . . 771

hen my son graduated from Texas

Competence in these three areas is only achieved by a combination of

describe and understand the actual behavior of the mechanical equipment.

machinery. The topics include synchronous (rotational speed) excitations such as

any mechanical problems are initially

MALFUNCTION CONSIDERATIONS AND CLASSIFICATIONS

The economic impact is directly associated with the criticality of the

Forced Vibration Mechanisms

Free Vibration Mechanisms

Forced vibration problems are generally solved by removing or reducing the

The box around this equation identifies this expression as a significant or

Fig. 21 Oscillating Pendulum Displaying Simple Harmonic Motion

Max. Neg. Displ.

In a rotating system, such as a centrifugal machine, frequency is normally

Combining (2-1) and (2-2), the rotational frequency may be expressed in

If the period is represented in terms of the pendulum length L, the above

eration characteristics of this pendulum would be a time domain examination.

In a rotating system, such as a centrifugal machine, this expression can be

The instantaneous velocity of this periodic motion is the time derivative of

By converting the cosine to a sine function, expression (2-12) is derived:

Note that velocity leads displacement by /2 or 90. Another way to state

Acceleration leads displacement by or 180, and it leads velocity by 90. It

mechanical systems in motion do display a consistent and definable relationship

Fig. 23 Traditional Angle Designation On A Rotating Disk

Phase displacement = Phase acceleration + 180

Phase velocity = Phase acceleration + 90

If a velocity phase angle occurs at 225, it is determined from (2-14) that

terms (i.e., sin=1), equation (2-12) may be restated as follows:

Next, consider the relationship between acceleration and displacement as

The relationship between acceleration and velocity may be stated as:

Simplifying this expression, the following common equation is derived:

The last three equations allow conversion between displacement, velocity,

0.3 IPS o-p 100 Hz

Vibration Amplitude (Mils, IPS, G's)

Fig. 24 Equivalent Displacement, Velocity, and Acceleration Amplitudes V. Frequency