Vibration Intermediate

INTERMEDIATE
VIBRATION ANALYSIS
WORLD HEADQUARTERS
835 Innovation Drive
Knoxville, Tennessee 37932
Phone: (423) 675-3200
Fax: (423) 675-3205
DETROIT TRAINING CENTER

39830 Grand River Ave. Ste. B-1-A
Novi, Michigan 48376
Phone: (248) 426-0472
Fax: (248) 426-0574
HOUSTON TRAINING CENTER

15425 North Freeway, Ste. 160
Houston, TX 77090
Phone: (281) 873-6000
Fax: (281) 873-6633
SAN DIEGO TRAINING CENTER

4180 Ruffin Road, Ste. 115
San Diego, California 92123
Phone: (619) 571-8882
Fax: (619) 571-8887
ONE STEP IN YOUR JOURNEY TO BENCHMARK STATUS

Copyright 1999, Computational Systems Incorporated. All rights reserved.
Content for this manual provided by CSI Training Instructor(s).
Formatting and editing by M. Angela West.
0199
Intermediate Vibration Analysis
Table of Contents
SECTION TITLE
One Analytical Troubleshooting
Two Review of Vibration Basics
Three Effects of Amplitude and Phase
Four Unbalance
Five Misalignment
Six Looseness
Seven Antifriction Bearings
Eight Data Averaging
Nine Digital Signal Processing
Ten Waveform Characteristics
Eleven Peak/Phase Analysis
Twelve Gears
Thirteen Electrical Defects
Fourteen Belts
Fifteen Resonance
Sixteen Pump Vibration Basics
APPENDIX TITLE
A Glossary of Terms
Section One
Analytical Troubleshooting
Objectives
< Understand the importance of frequency analysis in determining machinery
defects.
< Recognize the importance of predetermining potential defect frequencies.
< Identify the difference between synchronous, nonsynchronous and

subsynchronous faults.
Preparing for Analysis

Gathering as much history and physical information as possible before
acquiring and analyzing data proves helpful. Knowing the physical
specifications of the machine provides the analyst with the information needed
to calculate potential fault frequencies. Bearing geometry, coupling type,
number of gear teeth, and process considerations all play a part in this process.
Take the following steps to properly prepare for analysis:
Ô Collect machine information.
Collecting the machine information with a form similar to the one in the
appendix provides a great deal of reference information for building a
measurement point, as well as a mental picture of the forces and defects
that could possibly occur.
Õ Determine appropriate measurement points.
Sketch the machine train and define the nomenclature for each of the
points. If multiple machines are in question, ensure the point
descriptions from one machine to the next are consistent and easily
understandable. If monitoring systems are already installed, use them
and their point descriptions for ease and faster setup.
1/99 © Copyright 1999, Computational Systems, Incorporated. All rights reserved. 1-1
Ö Calculate potential fault frequencies.

Every rotating or moving machine component has the potential of
failing. For this reason, the fault frequencies should be predetermined.
The manner in which the fault is expected to fail should also be taken
into account. If a bearing’s inner race is the fault condition being
defined, the analyst must consider the higher frequencies first. Also
consider other similar faults such as stator slot pass, rotor bar pass, and
gears.
× Determine alarm criteria.

Setting the alarm criteria for a machine is easier than it sounds. The
presence of fault frequencies indicates existing faults. With this in
mind, set the alarms without existing faults. If other machines of the
same type are accessible, draw comparisons to establish a mean value
for energy. Finally, alarms should also consider trends. After setting
the initial alarm levels, look at trends with respect to the rate of change
between readings.
Ø Set priority of potential faults.

Once the potential faults have been identified, establish the occurrence
probability for each fault. Then consider the severity of the fault
condition. Faults that may not happen frequently may be considerably
more serious, therefore, warranting a higher priority. Last of all,
consider the difficulty in detecting the fault condition. If the fault is
difficult to identify, the priority for the fault should be placed close to
the top of the list.
Ù Determine possible fault causes.

Most faults have a variety of possible causes. For example, unbalance
may be caused by material build up, wear, broken components, etc.
After each of the different causes have been identified, prioritize each
cause for each fault condition.
Ú Establish preventive actions.

Some preventive actions should be established to eliminate or at least
minimize the frequency of occurrence. Balance and alignment problems
can be minimized by precision balancing and alignment techniques.
1-2 © Copyright 1999, Computational Systems, Incorporated. All rights reserved. 1/99
Û Establish information feedback.

After all is said and done, a feedback loop must be established to refine
all the information above. Spectral and waveform data, trends, other
machine changes, operational speed and load, work and the reasons for
the work should all be a part of the feedback loop. The chart on the next
page illustrates the feedback loop. Other components and potential
faults should also be considered and prioritized.
Vibration Analysis Flow Chart
Potential Fault Analysis Diagnostic Analysis
Determine Collect
Potential Machine
Faults Vibration Data
Set Priority Define

on Potential Actual
Faults Problem
Determine Compare Actual

Potential with Potential
Causes Faults
Determine
Preventive Most Probable
Actions Cause/Causes
Prioritize Causes
Correct most Contingent

probable causes Action
Collect
Followup
Machine Data
Were Machine
Information Faults
Feedback Corrected?
Yes/No
Steps to Solving Vibration Problems

Ô To identify the problem causing the machine vibration, ask yourself
some questions.
< What part of the machine has the vibration problem?

< How was the vibration measured?
< Were good measurement procedures used?
< Is the data valid?
< Does the vibration problem occur at only certain loads, temperature
or power conditions?
< Is the machine also noisy?
Õ The machine geometry should be understood as completely as possible.
< Sketch the entire machine train.

< Identify all the major components: motor, pump, gears, etc.
< Identify specifications on all the bearings in the machine.
• Sleeve or antifriction
• Type
• Number
• Bearing geometry
< Determine belt information.
• Center-to-center distance
• Pulley pitch diameters
• Number of belts
< Couplings.
• Gears
- Spool length
- Lube practice
- Clearances
- How aligned
• Disc
- How aligned
- End clearances
- Maintenance practices
• Bun
- How aligned
- Bun compound
- Possible torsional problems
• Fluid
- How aligned
- Estimate slippage
< Drives: motors, engines, turbines

< Gears: types and reduction ratios
< Shaft diameters and lengths
< Rotor dimensions and weights
< Other information unique to the machine
Ö Gather available maintenance history and any previous vibration data.
< Has vibration data been collected on this machine before?

< Has the vibration fault been developing over time, or is it a new
problem?
< Determine the last thing done to the machine.
< Talk to the machine operator and get his/her input.
× Determine the best points to collect data.
< At the bearings

< In the problem area
< Consider possible resonances
< Consider forcing functions from other machines
< What type of instrumentation will the solution of this vibration
problem require?
• Tape recorder
• Impact data
• Noncontact probes
• Displacement, velocity, or acceleration probes
• Coastdown or startup data

• Single or multichannel data
• Reference transducer input
• Current transformer
• Special averaging methods
• Temperature data
Ø Determine as many as possible forcing frequencies before taking data.
< Determine any and all shaft rotational speeds. Most machine defect
frequencies are related to a shaft turning speed.
< Bearing fault frequencies - BPFO, BPFI, FTF, and BSF.
< Belt frequencies.
< Gear Mesh frequencies.
< Blade pass frequencies.
< Resonant frequencies.
Ù Take data.
< Frequency data, at least horizontal, vertical, and axial data at each
bearing. It may be helpful to compare normally averaged data with
synchronous time averaging to identify synchronous and
nonsynchronous components.
< Check for beats by watching the instantaneous spectrum or by

comparing the peak hold spectrum with the normally averaged
spectrum.
< Check the skirt width of the spectral components. Steady state
signals collected with a Hanning window will occupy three to four
cells. Wide skirt widths indicate the presence of signal modulation,
another component very close in frequency, or a component that is
varying in frequency during the sample time.
< Do not limit your data collection to the bearing locations. Data at the
bearings should be considered the minimum data to be collected.
Consider data collection on the machine case, foundation, piping,
etc.
< Once data has been collected, break the spectrum into three different
regions: subsynchronous, synchronous, and nonsynchronous.
FAULT GUIDE
Vibration Dominant Frequency Dominant Plane Phase Reading
Unbalance
Static . . . . . . . . . . . . . . . . . . . . . . 1xTS ..................... Radial . . . . . . . . . Radial in phase
Dynamic . . . . . . . . . . . . . . . . . . . . 1xTS ..................... Radial . . . . . . . . . Radial 0-180 out / 2 plane
Couple . . . . . . . . . . . . . . . . . . . . . 1xTS ..................... Radial/axial . . . . . Radial 180 out
Overhung rotor . . . . . . . . . . . . . . . 1xTS ..................... Radial/axial . . . . . Radial unsteady/Axial in
phase
Misalignment
Angular . . . . . . . . . . . . . . . . . . . . 1x, 2xTS . . . . . . . . . . . . . . . . . . Axial . . . . . . . . . . Axial 180 out
Offset . . . . . . . . . . . . . . . . . . . . . .
1x, 2x, 3xTS . . . . . . . . . . . . . . . Radial . . . . . . . . . Radial 180 out
Offset + angular . . . . . . . . . . . . . . 1x, 2xTS . . . . . . . . . . . . . . . . . . Radial/axial . . . . . Radial/axial 180 out
Sleeve bearing . . . . . . . . . . . . . . . 1x, 2xTS . . . . . . . . . . . . . . . . . . Radial/axial . . . . . Axial 180 out
Antifriction bearing . . . . . . . . . . . 1x, 2x, 3xTS . . . . . . . . . . . . . . . Axial . . . . . . . . . . Axial 180 out
# of balls x TS
Bent Shaft . . . . . . . . . . . . . . . . . . . 1x, 2xTS if on coupling end . . . Axial . . . . . . . . . . Axial 180 out
Mechanical Looseness
Non-rotating bearings . . . . . . . . . . 1 - 10xTS . . . . . . . . . . . . . . . . . Radial
Rotating impellers . . . . . . . . . . . . 1xTS predominant as high
as 10xTS
Antifriction ì - (early stages) . . Bearing frequency . . . . . . . . . . . Radial
Bearings - (late stages) . . . . 1xTS and harmonics . . . . . . . . . Axial on thrust bearing
Sleeve Bearings
Looseness . . . . . . . . . . . . . . . . . . . Multiples of TS . . . . . . . . . . . . . Radial
Oil whirl . . . . . . . . . . . . . . . . . . . . 0.43 x TS . . . . . . . . . . . . . . . . . . Radial
Belt Drives
Mismatched, worn . . . . . . . . . . . . 2 x belt frequency . . . . . . . . . . . Radial inline with belt
Eccentric sheave . . . . . . . . . . . . . . 1 x shaft speed . . . . . . . . . . . . . Radial
Misalignment . . . . . . . . . . . . . . . . 1xTS . . . . . . . . . . . . . . . . . . . . . Axial
Gears — GMF = Gear Mesh Frequency, SG = Spur Gears, HE = Helical Gears
Transmission error . . . . . . . . . . . . GMF 1 + harmonics . . . . . . . . . Radial SG / axial HE
Pitch line runout . . . . . . . . . . . . . . GMF + sidebands . . . . . . . . . . . Radial SG / axial HE
Unbalance . . . . . . . . . . . . . . . . . . 1xTS . . . . . . . . . . . . . . . . . . . . . Radial SG / axial HE
Misalignment . . . . . . . . . . . . . . . . 1x, 2xTS . . . . . . . . . . . . . . . . . . Radial SG / axial HE
Faulty tooth . . . . . . . . . . . . . . . . . GMF + sidebands . . . . . . . . . . . Radial SG / axial HE
Rotor Rub . . . . . . . . . . . . . . . . . . 0.5xTS and ½ multiples . . . . . . Radial
Electrical
Loose iron . . . . . . . . . . . . . . . . . . 2 x line frequency (LF) . . . . . . . Radial
Stator problems . . . . . . . . . . . . . . 2 x LF . . . . . . . . . . . . . . . . . . . . Radial
Phase unbalance . . . . . . . . . . . . . . 2 x LF . . . . . . . . . . . . . . . . . . . . Radial
Loose stator . . . . . . . . . . . . . . . . . 2 x LF . . . . . . . . . . . . . . . . . . . . Radial
Broken rotor bar . . . . . . . . . . . . . . 2 x LF at 1xTS with . . . . . . . . . Radial
with sidebands
Eccentric rotor . . . . . . . . . . . . . . . 2 x LF at 1xTS . . . . . . . . . . . . . Radial
with slip sidebands
Loose slot . . . . . . . . . . . . . . . . . . . 2 x LF, slot frequency + . . . . . . Radial
sidebands
Pole pass . . . . . . . . . . . . . . . . . . . . At 1xTS with sideband spacing =
to # of poles x slip frequency
Blade/Vane Pass . . . . . . . . . . . . . # of blades/vanes x TS . . . . . . . Radial
Subsynchronous Frequencies
Subsynchronous frequencies appear below the shaft turning frequency.
Oil Whirl
Oil whirl occurs at approximately 44 percent of shaft frequency in pressure fed
sleeve bearings. The frequency decreases as the shaft speed slows down. The
frequency usually drops out at about 75 percent of normal running frequency.
Oil Whip
Oil whip occurs when the shaft is running at a turning frequency above the
second critical frequency. The oil whirl frequency locks onto the first shaft
critical frequency turning into oil whip. The frequency does not drop away
until the shaft frequency drops below the second critical frequency.
Rub
Rub frequency is close to 50 percent of shaft frequency and 1½, 2½, etc., and
harmonics.
Antifriction Bearing Loose in Housing

This frequency is 50 percent of shaft frequency, but 1½, 2½, etc., not as
noticeable as rub.
Cage or Train Frequency of Antifriction Bearing

This frequency usually indicates advanced stage of bearing failure. Check for
outer race fault and its harmonics as well as sidebands at the cage frequency.
Primary Belt Frequency

Check by calculation and look for higher harmonics. Could be caused by belt
misalignment, worn sheaves, or defective belts.
Defective Tooth-to-tooth Repeat Frequency

Usually a very low frequency which can be better seen in the time domain.
Often referred to as a hunting tooth frequency.
Surge
Usually a high component from 10 percent to 50 percent of rotor frequency.
Check differential pressure across the fan or pump. Check the operating point
with the best efficiency point on the fan or pump curve.
Ignition or Fuel Problem on Four-Cycle Recip

Usually accompanied with higher ½ orders.
Synchronous Frequencies
Frequencies synchronous with the shaft in the bearing where the data is taken.
The frequencies are integer multiples of the reference pulse.
Unbalance
One times shaft speed. If any looseness exists in the system, there may be
several harmonics of the first order component.
Check horizontal and vertical amplitude and phase at each bearing. If

horizontal-to- vertical phase shift is approximately 90E on both sides and the
phase relationship is similar, then it is unbalance.
If the horizontal-to-vertical phase shift is not close to 90E, consider possible

pedestal resonance or shaft centerline misalignment.
If the phase relationships side-to-side in both the horizontal and vertical

directions are not similar, consider the possibility of misalignment.
If the readings indicate an unusually high unbalance, look for a possible

resonance, bent shaft, or fault in rotor supports. Use the weight of the rotor in
ounces times the vibration amplitude in inches to estimate the unbalance in
ounce inches. For example, a 20,000 ounce rotor times an amplitude of 0.005
inches (5 mils) = 100 ounce inches of unbalance.
If the problem is unbalance and impact data can be taken, impact both sides
and estimate the pivot point. If this is far outside the bearings, it may not be
possible to field balance the rotor. In addition, the impact data will indicate the
first critical at each bearing. If this is close to the running frequency, it may be
better to consider stiffening the system.
Misalignment
The first effect of shaft misalignment is an increase in the radial load on the
bearings. In most cases, the next effect is an increase in the first order of the
shaft frequency. As the condition worsens, the second order builds.
Check the axial vibration on the bearings on each side of the coupling. If in
phase, consider balance or gear coupling lock-up.
If the misalignment is primarily angular, the top-to-bottom and the side-to-side

phases on each bearing are normally out of phase.
Bent Shaft
High first order, if balance is attempted a bent shaft will usually require large
correction weights. Check the axial data at both bearings; 180E phase shift is
an indication of a bent shaft. If possible, check the shaft with proximity probes
or dial indicators (be sure to check low-speed runout).
Check the top-to-bottom and the side-to-side phase on each bearing. If they are
out of phase, it is probably a bent shaft. In most cases of a bent shaft, the end-
to-end phase readings of either the horizontal or the vertical radial vibration are
the same.
Looseness
Look for many harmonics of shaft frequency. Usually the second is almost as
high as or higher than the first order. However, the harmonics will be
predominantly odd order if it is pure looseness.
Any truncated function will produce harmonics in the spectra so that a

condition where the time domain waveform is nonsymmetrical will look like
looseness. Check the time domain.
One or more system resonances may be excited by one or more of these

harmonics so that the levels of the harmonics is magnified by the resonant
amplification.
Pitch Line Runout on Belt Sheaves

Often confused with unbalance. Check vibration in line with the belt drive.
The frequency of the sheave with the runout will usually appear at the other
sheave.
If the belts are removed and the first order vibration is significantly lower, it is
not unbalance, but more likely pitch line runout.
The vibration will be at the frequency of the sheave with the pitch line runout.
Cavitation
Look for the first order and higher harmonics up to the number of blades along
with very high frequencies. Check the time domain.
Cavitation is the implosion of a void or bubble in the intake fluid when it

reaches the pressure side of the pump or fan. Because this can be a very steep
waveform, it can usually be easily identified in the time domain. It results in
many higher harmonics in the frequency domain. Often the blade pass
frequency will appear.
Blade Pass Frequency

Blade Pass Frequency is defined as the number of blades, etc., times the
rotational frequency. Blade pass is always there; however, high levels can
indicate system resonances excited by the blade pass component or an
arrangement of supports which causes process flow variations in sync with the
blade pass frequency.
On axial flow units, the primary blade pass may be reduced by an out-of-pitch
or out- of-track blade, but other harmonics will be higher. Check bearings in
the axial direction for components from the first order up to the number of
blades as a possible indication of this condition.
Gear Mesh Frequency

Gear Mesh Frequency is calculated as the number of gear teeth times the
frequency of the gear. These can be very high frequency and many times must
be measured with an accelerometer. Frequencies as high as 8,000 to 10,000 Hz
are not unusual.
A magnet base cannot be used to collect this data. Because the levels can be as
high as 50 to 150 g’s, it may be difficult to use a hand-held probe. The
transducer should be stud mounted or glued to the gear box if the levels are
above 20 g’s. In addition to the gear mesh frequency and its harmonics, the
pinion shaft, gear shaft, gear wobble, hobbing ghost, tooth resonance, and
entrapped oil frequencies and their harmonics may also be present in the
spectra.
Nonsynchronous Frequencies
These frequencies are higher than the shaft turning speed, but they are not
integer multiples of the shaft turning frequency.
Another Component in the Machine

Identify and determine severity. If possible, attempt to simplify the system by
disconnecting or shutting down some part of the whole.
One or More System Resonances

When a system is excited by some energy in the machine or in the process; take
enough impact data to identify such resonances and their excitation.
Belts
Even though the primary belt frequencies are subsynchronous, multiples of this
frequency can be nonsynchronous. The largest components of belt vibration
are usually at higher frequencies. In most cases, the highest level is at the
primary belt frequency times the number of pulleys over which the belt passes.
In multiple belt drives, there may be frequencies from two times the primary
belt frequency up to twice the number of belts times the primary belt frequency
in the spectrum.
Any of these frequencies may excite a system resonance and cause the
vibration level to exceed the acceptable level.
Belt alignment, tension, and quality all play a part in the level of vibration from
the belt drive.
In some cases, it is impossible to reduce the belt vibration to an acceptable

level. Replacing the regular V-belts with Grip-Twist or Grip-Link V-Belts
usually reduces the vibration level from two to four times.
For high quality drives, it may be necessary to use flat belts, such as the Habasit
belt, in order to reduce the vibration to an acceptable level. These belts usually
reduce the belt vibration ten times or more.
Antifriction Bearings
The basic frequencies generated by antifriction bearings are the cage or carrier,
the ball spin, the outer race fault, and the inner race fault. One or more of the
primary frequencies along their harmonics and their sidebands may appear in
the spectra. Because the outer race is always in the load zone, however, the
outer race frequency and its harmonics are the most commonly detected.
The bearing fault frequencies can be calculated accurately from the bearing
physical data or estimated. Although the major causes of antifriction bearing
failure are improper handling, installation, and lubrication, vibration can give a
reliable and early indication of bearing failure.
For antifriction bearing analysis, data should be taken as acceleration using the
shortest solid path to the load zone of the bearing for the transducer mounting.
Carefully examine the spectra for a series of harmonically related peaks which
are not synchronous with the shaft in the bearing.
If the actual fault frequency values are not known, check to see if this series of
peaks falls near the approximate value. Operating the bearing above the rated
load and at higher than recommended shaft speeds will shorten the life of the
bearing under the best of conditions.
Electrical
Vibration can be caused by the dress of the conductors in the raceway, loose
lamination in a transformer or a motor, broken or cracked rotor bars, open
shoring rings, eccentric rotors, eccentric stator, etc. Conductors should be
redressed and loose lamination should be readjusted and tightened to reduce the
line frequency and the two times line frequency components.
Eccentric stators cause two times line frequency vibration, which is directional
to the largest air gap.
Magnetic misalignment results in two times line frequency plus the number of
poles times the slip frequency sidebands.
Eccentric rotors cause running speed and twice running speed components with
the side bands at the number of poles times the slip frequency. Broken or
cracked rotor bars or open or shorted rings may cause line frequency and twice
line frequency components at the number of poles times the slip frequency.
Broken or cracked rotor bars often show up at one times turning speed with
sidebands at the number of poles times the slip frequency.
This defect only shows up under load, so many motor repair shops are unable
to determine these faults. One or more of the faulty bars will cause unequal
heating in the rotor, which results in a thermal vector so that the balance will
only be good for one load condition. Such thermal vectors are very common in
motors and generators.
Noise
Chain drives, gears, the process, etc., can be the cause of high noise levels in
the system. It is important to separate the noise problem from the vibration
problems.
In general, noise levels are reduced by sound absorbing materials and vibration
levels are reduced by correction.
Often high noise levels do not indicate a serious vibration problem. Noisy
gearboxes, for example, may be the result of the gear frequencies exciting
natural frequencies in the gearbox cover which in turn radiates the acoustic
energy.
Unusual Sources
There are many cases of unusual sources of vibration: foreign material or
objects which move inside a rotor, water weeping in airfoil blades, entrapped
water in a rotor which turns to steam when the rotor reaches operating
temperature, axial clearance used up by thermal expansion, both bearings
locked on a shaft, uneven stress caused by dissimilar material, electrical
discharge through bearings, etc.
Suspect the unusual when the levels are very high, when phase and amplitude
readings do not repeat from one run to the next, or when the readings are
erratic.
• Tennis ball inside automobile tire

• Can of baked beans inside intake manifold
• Clean-up man dumping cold water in fan intake
Background Vibration
The background vibration levels should be checked with the machine shut
down. Much time has been spent trying to determine the machine cause of a
frequency in the spectrum which later turned out to be present in the
background and not part of the machine under study.
If both the background and the machine vibration are repetitive, the
background levels can be removed from the spectrum by linear averaging while
the machine is running and negative averaging with the machine shut off.
Summary
Remember, vibration analysis is not always as simple as we would like it to be.
However, when problems are approached logically, recognizing what is known
about the machine in question and its history, the solutions to the vibration
problems can be found.
Determine the turning speed frequency. Then the frequencies present in the
spectrum can be categorized as subsynchronous, synchronous and
nonsynchronous energy.
Don’t give up during the analysis process and remember to think about the
possible unusual causes.
Section Two
Review of Vibration Basics
Objectives
This section provides a review of the basics of vibration theory.
< Describe the fundamentals of vibration theory covered in Vibration

Analysis I or other basic vibration courses.
< Apply this information as a foundation for later discussions on vibration

theory and fault types.
Fundamentals of Vibration Theory

Vibration is the movement of a body about its reference position.
• The vibration occurs as a result of an excitation force that causes the

motion.
• This exciting force may either be random or periodic.
The remainder of this chapter discusses forced periodic vibration such as

vibration found in rotating equipment. A classic example of vibration is a
weight supported by a spring. The weight has mass (M), and the spring has
stiffness (K). If a forcing function (FT) such as unbalance is present, the mass
will move back and forth across its reference or starting position.
1/99 © Copyright 1999, Computational Systems Incorporated. All rights reserved. 2-1
As long as there is a forcing function, the mass will continue to vibrate

forever. The diagram below shows a damper that adds resistance to the
motion. Some systems have dampers to isolate the vibration from other
structures and machines.
System Damping Examples
Light Damping
Moderate Damping
Heavy Damping
The amount of damping actually determines the resistance the system will have
to motion. As seen above, a lightly damped system offers little resistance to
motion; therefore, it takes a longer time for the signal to decay. Most rotating
mechanical systems have very little damping. As long as the forcing function
(unbalance, misalignment, gear/bearing impacts) remains, the vibration will
continue. This vibration is repetitive and lends itself well to FFT (Fast Fourier
Transform).
2-2 © Copyright 1999, Computational Systems Incorporated. All rights reserved. 1/99
Vibration Transducers
In order to measure and analyze a machine’s vibration signature, the
mechanical vibration must be converted into a voltage signal that can be
analyzed by the spectrum analyzer. The amplitude of the voltage signal varies
with time, which causes the time waveform on which the FFT is performed to
arrive at the spectrum.
Settle Time
A device called a transducer is needed to convert the mechanical vibration to

the electrical signal. In the world of vibration analysis, there are three primary
groups of transducers.
• Displacement Probes
• Seismic Velocity Probes
• Accelerometers
Displacement Probes
Displacement probes are noncontact probes that measure the relative motion
between the probe tip and the vibrating element. They convert mechanical
vibration into a time-varying voltage proportional to the displacement of the
vibrating surface. Usually displacement probes are permanently or semi-
permanently installed on the machinery. These probes have a limited
frequency range of about 1,000 Hertz.
Seismic Velocity Probes

Seismic velocity probes are not as commonly used in the vibration industry as
they once were. They are composed of a permanent magnet spring mounted
inside a coiled wire. These probes convert the mechanical vibration into a
time-varying voltage proportional to the velocity of the vibration. Due to the
magnet moving relative to the coil, the probe is self-generated and does not
require external power. These probes also have a limited frequency range of
about 1,000 to 2,000 Hertz.
Accelerometers
Accelerometers come in various types. One of the most common is the
piezoelectric accelerometer. These probes contain a piezoelectric crystal that
emits a small voltage proportional to the acceleration levels sensed by the
transducer. Accelerometers are available in many different designs with
widely varying frequencies. Special accelerometers are available that convert
acceleration to velocity before the voltage signal reaches the transducer cable.
Mounting Resonances
Mounting resonances occur more often in accelerometers with higher
frequency ranges available. A resonance is the frequency at which a
structure’s vibration is amplified due to structural considerations — low levels
of damping or low stiffness. Transducer mounting resonances are affected by
how the sensor is attached to the machine or vibrating surface. A very stiff
method of attachment, such as a stud mount, generally moves the mounting
resonance to a frequency above the frequency range of interest.
It is important when selecting a transducer to select the appropriate transducer

for your application and an attachment method that will place the mounting
resonance out of the frequency range of interest. Transducer selection and
mounting are covered later in the section on Transducer Selection.
Amplitude Units
A trace of the displacement over time reveals a sinusoidal waveform pattern.

The amount of vibration, the amplitude, varies depending on which units are
used to describe the vibration (displacement, velocity or acceleration).
The typical units for each of these amplitudes are:
Displacement = Mils, Peak-to-Peak

Velocity = Inches/sec, Peak
Acceleration = g’s, RMS (1g = 386 inches/sec/sec)
For most PDM applications, vibration data displays in units of velocity, which
treats high frequencies and low frequencies equally. Acceleration enhances
the amplitudes of high frequency data, and displacement enhances the
amplitudes of low-frequency data.
2120 Lab - Analyze/Monitor Mode
Ô Connect 2120 to communications cable.
Õ Start MasterTrend.
Ö Open Database INTMDVIB.DAT.
× Use Analyzer communications to download a route from the

station DIAGNOSTIC LAB 1 named INT. VIB EXMPL.
Ø Connect lab table BNC X output to model 728 voltage adapter volt
input (in this lab, the lab provides power).
Ù Monitor variable speed rotor per instructor’s directions on

following pages.
Displacement
Using the Analyze, Monitor Mode,

Monitor Waveform, set up the
analyzer to monitor the signal from
the variable speed rotor.
Let’s look at the simple rotor kit

running at around 25 Hz (1500
RPM) with some unbalance as the
predominant forcing function. We
know from earlier studies that an
unbalance force creates a 1x turning
speed peak in the spectrum. We will
start by looking at the displacement
of the rotor with respect to time (D),
— a digital recreation of the
movement from the at rest position
(centerline) over time.
The distance traveled by the rotating mass can be described by measuring the
maximum positive displacement and the maximum negative displacement —
the Peak-to-Peak Displacement. This is normally measured in mils (1 mil =
.001"). This could be measured mechanically with a dial indicator at slow
speeds, but it becomes impossible at higher speeds. A more convenient way is
to use an accelerometer to measure the acceleration and mathematically
integrate to obtain displacement.
The cursor on the waveform above

marks the peak positive value of the
time waveform. The value is
approximately 9.6 mils peak (your
values may differ). Industry standard
describes displacement in Mils Peak-
Peak. To get this value, we need to
find the peak negative value.
The cursor marks the peak negative

value at approximately -9.2 mils
peak. The value of this time
waveform is 18.8 mils pk-pk.
Velocity
Change the Waveform Units to

Velocity.
To find the speed at which the unbalance mass travels, divide the distance
traveled (displacement) by the time it took to travel this distance. This speed is
the velocity (V). Inches per second (in/sec) is the industry standard unit.
D / T = Velocity
Velocity is normally measured 0-

Peak or simply Peak. In the
example above, the value is 1.484
in/sec.
Acceleration
Change the Waveform Units to

Acceleration.
Circularly the mass is moving at a constant velocity. In the vertical direction

— what is being measured — the unbalance mass stops and changes direction
at the top and the bottom. Vertically, the velocity changes during one
complete cycle of motion. The rate that the velocity changes is called the
acceleration (A).
V / T = Acceleration
Acceleration is normally measured in

g’s RMS. In this example, the peak
value of 0.6921 is multiplied by .707
to obtain the value in g’s RMS units.
Time Units
Let’s look at the basics describing the units of the time waveform again. This
time, instead of the amplitude units (Y axis) we will look at the time units (X
axis).
+
0 to Peak
RMS
Average
Peak-to-Peak
Time
A C E
- D
The time from A to E is how long it takes the event to happen. We can
measure this time and relate it to a frequency (F).
We can see that at time 0.03144, the unbalance force was at its peak positive
position. At time 0.07031, the mass has completed the cycle. This time is
referred to as T or Period. Since the time it takes for the unbalanced mass to
return to its starting position is also equal to one revolution, we can calculate
the time it takes by subtracting the two values.
T = .07031 sec. - .03144 sec.
T = .03887
The following equation links event times in seconds to frequencies in Hertz:
1
F (Frequency) '
T (Time)
1 1
Frequency ' ' ' 25.73 Hz ' 1543 RPM
Time in Seconds 0.03887
This calculation is not very difficult for the given example, but a typical
vibration signal has a very complex time waveform with many different signals
combined into one. We cannot easily pick out frequencies or amplitudes so we
must convert the time waveform into a more usable format. The mathematical
process of converting from the time domain to the frequency domain is called
the Fast Fourier Transform (FFT).
Frequency Spectrum
The Fast Fourier Transform (FFT) converts time waveform information to
frequency-based information. The FFT has the graphical affect of rotating the
waveform 90E so that the waveform is coming out of the page toward the
reader. Many vibration faults are more easily diagnosed in the frequency
spectrum than in the time waveform, because the frequency spectrum can
separate closely-spaced frequencies and allow for the viewing of relative
amplitudes.
In order to get full benefit from a spectrum, the analyst must understand signal
processing, bandwidth calculations, lines of resolution and transducer
selection. These topics will be covered later in this manual.
In review, the frequency units used to identify frequencies in a spectrum are:
Cycle per Minute = CPM

Cycle per Second = Hertz (HZ)
Cycle per Revolution = Order
2120 Lab - Frequency Spectrum

Ô Collect data from variable speed rotor kit (IMBAL EX) for the
points labeled VSD, VSV, VSA.
Õ Dump the data to MasterTrend for viewing.
Machinery faults occur at specific frequencies that are often difficult to

recognize in the time waveform. This makes the spectrum a very useful tool.
The time waveform, although not as useful for fault detection, is valuable for
determining the character of the vibration such as impacting symmetry.
Summary
This section reviewed some basic vibration concepts. The main points to
remember are the definitions of time waveform, frequency spectrum,
amplitude units, and frequency units. It is also important to know the various
types of vibration transducers.
VIBRATION IN ANSWERS OUT
H ow ’d
they
do that?!
The goal of this course is to remove the black box mystique from the vibration
measurement process. It is important for an advanced vibration analyst to
understand the material in this manual and the analysis processes behind many
of the choices made on a daily basis in the vibration industry.
Section Three
Effects of Amplitude and Phase
Objectives
< Describe the effects of amplitude units on frequency and time domain
data.
< Explain what phase is.
< Explain the phase shifts associated with amplitude units.
< Discuss the importance of transducer polarity.
Amplitude Unit Characteristics

Effects in the Spectrum
Basic vibration training introduced three units used to measure amplitude.
Velocity, Displacement, and Acceleration apply to amplitude in both the
frequency domain and the time domain. The most common amplitude unit for
vibration trending is Velocity. Displacement may be used when relative
motion or slow speed is a consideration. Acceleration is often used in gearbox
and high speed machinery as well as bearing troubleshooting.
Effects in the Time Domain

As expected, the amplitude units affect the time domain much in the same
manner as the frequency domain. The next few examples show the effects of
amplitude in both the spectrum and waveform.
Displacement
The spectrum and waveform display the attenuation of the high frequencies.
The spectral plot clearly shows no high frequency data. This is evident in the
waveform by the lack of noise riding on the sinusoidal shape.
Velocity
Viewing data linearly across the frequency range is a characteristic of Velocity.
Some energy is present around 15xTS and 30xTS in the spectrum. This
energy is reflected in the waveform as the noise riding in on the sinusoidal
shape. This high-frequency energy was not present in the displacement
waveform. Again, the dominant shape is sinusoidal due to the characteristics
of velocity.
Acceleration
Accentuating high frequencies and attenuation of low frequencies are
characteristic of Acceleration. Notice the raised noise floor in the spectrum
beginning at approximately 12xTS. The waveform has an underlying
sinusoidal shape, but the high frequencies are very prominent. These
amplitude units also affect the relationship of phase data.
Phase Relationships in the Time Domain

What is Phase?
Photo Tach T
CL
Time
Phase = t •2 (radians)
T
x
Phase = t •360 (degrees)
T
t
Single-Channel Phase Measurement (@ 1XTS)
Phase is one of the most important single-channel analysis tools for a vibration
analyst. Phase is a relationship — simply the difference in time between two
events. We talk about two different types of phase in vibration terms —
absolute phase which is really the relationship of the peak vibration as it
relates to some reference signal and relative phase. The most common
frequencies for single-channel phase analysis are 1x and 2x RPM signals. The
peak of this signal is compared to a reference mark (reflective tape, keyway,
stroboscope). The time relationship of the reference mark and the peak
vibration is converted to an angular value from 0 to 360E (one revolution). In
later chapters we will look at relative phase relationships between two
vibration points using the two-channel capabilities of the 2120 (Cross Channel
Phase). The analyst uses phase when trying to balance an imbalanced rotor to
locate the heavy spot. Phase is also a useful tool to determine types of
unbalance, misalignment, looseness, soft foot, bearing misalignment and
resonance which we will look at in the appropriate chapters.
2120 Lab - Peak Phase

Press Analyze, select 6) MONITOR MODE, and then select 5) MONITOR
PEAK/PHASE.
We will monitor 1X RPM or 1 Order. The defaults will be good for the rest
of the choices.
Notice the phase reading. We will now show how the phase is calculated.
Press the Reset button.
Let’s take a look at how this Peak/Phase reading is developed. Select the
Analyze function and then MONITOR MODE and MONITOR
WAVEFORM. We will turn the TRIGGER MODE to TACH.
When you have a stable signal, press Enter to suspend the measurement.
Use the Exp and Dec keys to adjust the display as shown. We need to know
the total time for one revolution. Use the left and right arrow keys to move
the cursor until it lines up with the tach marker near the center of the screen.
Write this time at 1. below. Use the left arrow to cursor to the previous tach
marker. Write the time at 2. below. To calculate the time for one revolution
(360E), subtract 2. from 1. to get T.
1. Time 1
2. - Time 2
Time
T
Next we need to find when the peak

is reached. Use the arrows to move
the cursor to the peak positive
vibration. Write this time at 3.
below. We will use a proportion to
calculate the time event relationship
to 360E.
3. Time 3
2. - Time 2
Time
Your answer should be close to the reading you had monitoring the
peak/phase.
t
Phase ' @ 2 (radians)
T
t
Phase ' @ 360 (degrees)
T
2120 Lab - Phase vs. Amplitude Units

We will look at our rotor kit time waveform signal again — this time with a
tach signal. We have placed a piece of reflective tape on the rotor and an
accelerometer on the unit.
Press the Analyze button and then 2) MONITOR WAVEFORM. We will

use the following setup:
Make sure TRIG MODE is set to TACH.
You may need to use the Exp and Dec keys to adjust the display for 2 to 3
revolutions (vertical dashed lines).
The displacement is at its peak about halfway between the two tach pulses.
Remember the tach pulses represent one revolution or 360E. The peak
displacement in the signal above is close to halfway between the two markers
or approximately 180E.
Now change the waveform units to

velocity. Press Enter.
Again, adjust the Exp and Dec keys for the same display.
The peak velocity above occurs about 1/4 of the distance between the tach
markers. This equal to approximately 90E.
Now change the waveform units to

acceleration. Press the Enter key.
You may need to adjust the Exp and Dec keys again.
The peak acceleration takes place on the tach marker. This is equal to
approximately 0E.
Phase vs. Amplitude Units

What we have just seen is that there is a significant difference between the
phase relationships of the three different amplitude units — acceleration,
velocity and displacement. The laws of physics and relationship of vectors
govern this topic. The figure using displacement as the base unit type,
illustrates that velocity leads displacement in phase by 90E. Also, acceleration
leads velocity by 90E or leads displacement by 180E. This is important when
comparing phase readings collected with accelerometers and noncontact
displacement probes. The figure shows the phase shift caused by selecting
different amplitude units.
Velocity Sinewave
Displacement Sinewave
Acceleration Sinewave
90E 90E
time
The sinewaves above show the phase

relationship between the three most typically
used transducer types.
The selection of different amplitude units is just one source of hardware-

induced phase shifts.
Transducer Polarity
An indication of a phase shift across a machine may or may not be caused by
the machine. If the phase data was collected with a multichannel analyzer,
then the issue of transducer polarity must be addressed. The internal wiring of
vibration transducers will vary with model and manufacturer but can also vary
between the same model. Therefore, a phase shift on a piece of equipment
may be due to the transducers being used rather than a machine fault. For this
reason, it is extremely important to check the polarity of any transducers used
for phase collection.
Transducer A
Transducer B
The cross-channel phase data was collected from the machine pictured above.
Both accelerometers were the same model number and mounted in the same
orientation on the motor. The phase data being compared is associated with
turning speed of the motor. The expected phase shift would be 5 or 10E
between the transducers. Notice the actual phase shift is 174.74E. The
induced phase shift between the transducers was caused by the internal
connections being reversed. If this difference in polarity was not recognized
prior to data collection, an incorrect analysis may result.
Section Four
Unbalance
Objectives
< Describe the spectral and waveform characteristics of unbalance.
< Identify the different types of unbalance using spectral data and phase
data.
< Associate the special conditions of vertically mounted equipment with

the balancing process.
What is Unbalance?
Unbalance, sometimes referred to as imbalance, is very simply the result of a
shaft’s center of mass not rotating at the center of rotation. This results in a
heavy spot on the rotor. This heavy spot produces a centrifugal force that
forces the rotor to rotate off center and causes a high vibration amplitude at the
1xTS frequency in the spectral data and sinusoidal waveform in the time
domain data. Phase data is used to locate the heavy spot on the rotor.
Unbalance
It is important to realize that a vibration sensor and a tachometer signal cannot

directly identify the heavy spot on the rotor. The tach signal and the sensor
actually identify the high spot. The high spot is defined as the measured
response to the heavy spot on the rotor. This is the point where the measured
vibration is the highest.
Unbalance
Phase Lag
A phase lag angle exists between the actual heavy spot on the rotor and the
measured high spot. Remember that CSI equipment measures true phase,
which means that the phase angle increases opposite to the direction of rotation
from the reference mark. A factor that affects the angle of the phase lag is the
frequency of the first shaft critical frequency. If the turning speed is much less
than the shaft critical frequency, the phase lag will be close to 0E. As the shaft
speed approaches the shaft critical, the phase lag will approach 90E. At
resonance the phase lag will be 90E. Above resonance, the phase will be above
90E. As the shaft speed increases significantly above the resonant frequency,
the phase lag will approach 180E.
The lag angle exists because during the time required for the vibration due to
the heavy spot to be transmitted through the machine to the vibration sensor,
the rotor is still turning, allowing the phase angle to continue to increase. One
way of determining the phase lag angle is to place a known amount of weight at
a known angular location on the rotor. This is known as a trial weight. The lag
angle can be determined from the resulting amplitude and phase data, which is
the effect of the trial weight.
Unbalance
Most computerized balancing programs will determine the lag angle during the
balancing process. The concepts of heavy spot, high spot, and lag angle are
important when balancing a rotor. The high 1xTS vibration occurs because the
heavy spot passes the vibration sensor once per revolution, producing the
measured vibration at the turning speed frequency.
FAULT GUIDE
Vibration Dominant Frequency Dominant Plane Phase Reading

Unbalance
Static . . . . . . . . . . . . . . . . . . . . . . 1xTS ..................... Radial . . . . . . . . . Radial in phase
Dynamic . . . . . . . . . . . . . . . . . . . . 1xTS ..................... Radial . . . . . . . . . Radial 0-180 out / 2 plane
Couple . . . . . . . . . . . . . . . . . . . . . 1xTS ..................... Radial/axial . . . . . Radial 180 out
Overhung rotor . . . . . . . . . . . . . . . 1xTS ..................... Radial/axial . . . . . Radial unsteady/Axial in phase
Unbalance
Spectral Data
When unbalance is the primary problem, a 1xTS will appear in the spectral
data, and it will generally be at least 80 percent of the overall amplitude. The
harmonics of 1xTS should be very low in amplitude, with low axial vibration
levels. Also, if the turning speed frequency is below the first shaft resonant
frequency, then the unbalance amplitude will increase proportionally with the
square of the speed change. For example, if the speed doubles from 1000 RPM
to 2000 RPM, then the 1xTS amplitude can be expected to increase by a factor
of four.
Unbalance
Waveform Data
Since the spectral data comes from the time waveform, we can expect the time
waveform to have a strong sinusoidal pattern. The display units of the time
waveform will affect the clarity of the sine wave. A displacement waveform
will show a clean sine wave while an acceleration waveform will show a lot of
high-frequency data riding on top of the sine wave.
Unbalance
Types of Unbalance
Unbalance can be divided into two primary types — static and dynamic.
Static Unbalance
Static unbalance is often referred to as force unbalance. A simple force
unbalance can be corrected with a single-plane balance correction. In a
frictionless stand, the heavy spot will roll to the bottom and can be corrected
with one weight 180E from the heavy spot. A statically balanced rotor should
stay at any position in a frictionless stand.
The radial vibration readings are the highest amplitude with the axial vibration
generally much lower in amplitude. Phase measurements can provide the data
needed to identify static unbalance. Static unbalance will show a 0E phase shift
across the rotor (vertical to vertical or horizontal to horizontal) and a 90E phase
shift from vertical to horizontal at the same bearing location. Also, the
measured phase angle will change the same amount that the heavy spot
changes, if the system is linear.
B A
D C
90 E
Horizontal to Vertical
If a 0E or a 180E phase shift exists from the vertical to the horizontal

measurement direction, then the machine is most likely running at resonance or
is misaligned. The resonance or misalignment will need to be corrected before
attempting to balance.
Unbalance
Dynamic Unbalance
Dynamic unbalance is actually every case other than static unbalance. This
includes couple unbalance, overhung rotors, and vertically mounted machines.
Dynamic unbalance will require two or more correction weight planes to
balance the machine. A good rule of thumb to follow is that for a single rotor,
two-plane balancing may be required if the rotor diameter, D, is less than four
times the rotor width, W.
If D / W < 4, then two-plane balance is required.
The vibration data for a dynamically unbalanced rotor will not differ
significantly from static unbalance with the exception of the phase data. The
high 1xTS radial vibration will still dominate the spectral data. The phase data
is controlled by the location of the heavy spots on the rotor. To understand
how the phase will differ from static to dynamic unbalance, we need to
understand how the location of the heavy spots vary from static to dynamic
unbalance.
Unbalance
Static vs Dynamic vs Couple Unbalance

Dynamic unbalance is the result of:
< Two unequal heavy spots 180E apart in separate planes on the same
rotor. Phase is likely to be 180E across the rotor, but the amplitudes
will differ since the amount of the heavy spots differ.
< Two equal or unequal heavy spots in separate planes on the same
rotor located at some spacing other than 180E apart. The amplitude
and phase will be related to the location and the amounts of the
heavy spots.
< If the heavy spots are equal and 180E apart, then the rotor will be
dynamically unbalanced but statically balanced. The amplitudes
across the rotor should be similar and the measured phase readings
will be 180E apart. This special case of dynamic unbalance is called
a couple unbalance. Remember that although this rotor will stay at
any position in a frictionless balance stand and is statically balanced,
each end is unbalanced by the same amount, 180E apart. This
requires correction in two planes.
Heavy Spot Unequal Heavy Spots Equal Heavy Spots
STATIC DYNAMIC COUPLE
Unbalance
Other Considerations
Overhung Rotor
This subject always seems to get treated separately. However, if an overhung
rotor is statically unbalanced, then it could be balanced in one plane. If it is
dynamically unbalanced, then two or more planes will be necessary to correct
the unbalance. The vibration will be high in the radial directions as seen with
center hung rotors, but over hung rotors often show a high 1xTS axial vibration
due to axial thrusting resulting from the unbalance.
The phase readings will provide the needed information to determine whether a
single or multiplane solution is needed. If the phase readings from each
bearing are the same (vertical to vertical, horizontal to horizontal) then the
rotor could be balanced in one plane. This is provided the analyst has access to
the center line of the rotor. If the rotor diameter is less than four times the rotor
width, then two-plane balancing may be required.
Vertical Equipment
Vertical equipment really can be tricky because the unbalanced rotor is
sometimes inaccessible. Usually, weights can be placed in a plane at the top of
the motor so that a centrifugal force is created that pulls, counteracting the
vibration at the top of the motor.
Summary
Unbalance is the simplest defect to identify in a vibration spectrum provided
the turning speed of the machine is known. Be aware that, although unbalance
shows up at 1xTS, many other defects also appear at 1xTS. This is why
checking the phase can be helpful when confirming unbalance.
Unbalance
2120 Peak/Phase Balance Lab

In this lab, we will use the Analyze function and Monitor Peak/Phase to
evaluate an out of balance rotor.
Ô Setup the analyzer and associated equipment to monitor the rotor kit.
Õ Evaluate spectral data for unbalance.
Ö Monitor and record Peak/Phase data.
× Evaluate the results.
Unbalance
Case Histories - Unbalance
Unbalance Case #1 - Vent Blower
Ô 450-HP, 3600-RPM Motor.
Õ Fan is direct drive, center hung.
Ö Fan has 16 reverse curved blades.
× Material buildup has been a problem.
Unbalance
The spectrum above is from the Fan Outboard Horizontal (FOH) position.
Notice the 1xTS peak with some harmonics of turning speed. The amplitude of
the 1xTS vibration had exceeded the warning level on the monitoring panel.
What problem should be corrected first, the unbalance or the looseness?
Unbalance
The Fan Outboard Vertical (FOV) plot is shown above. As expected, the 1xTS
amplitude is lower than the FOH amplitude. Again, looseness is present in the
data. It was determined that the looseness was being driven by the unbalance
in the fan. The next page shows measurement points FOH and FOV after the
fan was balanced.
Unbalance
Notice the decrease in the amplitude of the 1xTS peak and the looseness on
both measurement points. This clearly shows what happens to vibration when
its driving force is removed. The unbalance appears to have damaged the
bearing.
Unbalance
Unbalance Case #2 - West Desuperheater Pump
Ô 75-HP, 3600-RPM motor.
Õ Unit had frequent pump failures.
Ö Motor had very few hours of run time.
× Vibration analysis was used as a last resort to determine the problem.
Ø Data was collected with the motor uncoupled.
Unbalance
The motor points are displayed on the multiple plot above. The running speed
peak is marked and shows an amplitude of 1.3 in/sec (IPS). It appears that the
highest amplitudes are in the horizontal direction. Remember, the motor is
running uncoupled from the pump.
Look at the amplitude of the 1xTS peak on the Motor Outboard Horizontal
(MOH) point. As discussed, there are no harmonics of motor turning speed.
What should the waveform look like?
Unbalance
As expected, the waveform is repetitive, sinusoidal and has a strong 1xTS

pattern. Notice the amplitude is fairly high, but lacks any impacting.
The spectrum from the Motor Outboard Vertical (MOV) measurement point is
shown above. Its amplitude is 6.5 times less than that of the horizontal reading.
This is a good indication that unbalance is not the problem that needs to be
corrected. When the motor was removed for balancing, it was found the
foundation grouting was cracked. Anytime you find a 3:1 ratio when
comparing the horizontal to vertical reading, suspect another problem other
than unbalance.
Unbalance
Unbalance Case #3 - Reactor Fan #6
Õ Fan is center-hung.
Ö Highest 1xTS vibration on horizontal readings.
Unbalance
The Fan Inboard Horizontal (FIH) measurement point is shown above. Once
again, the 1xTS peak is the dominant peak in the spectrum with little or no
harmonics. The Fan Inboard Vertical (FIV) spectrum is displayed.
This plot also shows the 1xTS peak as the dominant peak in the spectrum. The
amplitude of the FIH 1xTS peak is almost 10 times higher than the FIV 1xTS
peak. This should initiate the collection of more data to assist in making the
proper diagnosis. The next page shows a measurement point that should be
considered.
Unbalance
The fan axial measurement point is displayed above. Notice the 1xTS peak is
higher than the vertical 1xTS peak, however, it is still much lower than the
horizontal 1xTS peak. Further investigation should continue to determine the
cause of the high ratios horizontal to vertical.
Could this problem have been misalignment?
Without the benefit of phase data or the time waveform this example is more
difficult to make a call on.
Use all the tools that are available to you as an Analyst. Store the time
waveforms and collect phase data to verify any suspected problem.
Unbalance
The waveform does help us to make a call on this example. We can be more
confident now that the problem was due to unbalance.
Again phase data would have helped us verify this problem.
Unbalance
Unbalance Case #4 - Office Chair Shaper
Õ Unit is driven by a flat belt.
Ö The Sheave Ratio is 2:1.
× The shaper had excessive vibration.
Unbalance
The multiple-point plot is displayed above for the shaper measurement

locations. Measurement point CX has the highest amplitude reading over 1.0
IPS at shaper running speed.
The spectrum for measurement point CX is displayed above. The

characteristics for unbalance are again seen in this plot, which is a dominant
1xTS peak with little or no turning speed harmonics.
Unbalance
The spectrum for measurement point CY is displayed above. Its amplitude is

consistent with what would be expected when it is compared to the horizontal
reading. Remember, belt-driven systems can have high ratios when the
horizontal and vertical amplitudes are compared.
Unbalance
Unbalance Case #5 - Dunson Blower
Displayed below is a multiple points plot of the Blower. Where is most of the
vibration? In what direction? This plot also includes the motor inboard
measurement point.
Unbalance
Displayed below is a parameter profile trend of the BIH measurement point of

just the 1 x rpm band.
Note that the blower had been running well above the alert level that was
defined for this unit.
Unbalance
This single full scale spectrum also reveals almost all of the vibration is at 1 x
rpm.
The waveform for this spectrum displays a sinusoidal 1 x rpm per revolution
type pattern. Note that this is in the radial (horizontal direction). Armed with
both the spectral and waveform data the analyst can feel confident in making
the correct call on this machine. When the machine was taken down and
balanced it ran at less than .1 in/sec.
Section Five
Misalignment
Objectives
< Identify misalignment based on spectral and waveform characteristics.
< Explain how to use phase to help distinguish misalignment from

unbalance.
< Describe analysis techniques using single-channel data acquisition.
What is Misalignment?
Misalignment is another major cause of machine component damage. At one
time, it was considered the second most prevalent vibration source. Most
industries with predictive maintenance programs, however, have found that
misalignment occurs more often than unbalance.
Maintenance personnel too often rely on manufacturers’ claims that their self-
aligning bearings and flexible couplings compensate for shaft misalignment.
Nevertheless, major vibration problems can be present even when couplings
are aligned within manufacturers’ tolerances. It is important for those
responsible for machine installation and maintenance to understand both the
importance of shaft alignment and the proper methods of aligning shafts.
Misalignment occurs because of poor alignment between mating pieces such as

coupling halves, bearings, shafts and pulleys. Misalignment of couplings,
bearings and pulleys results in high axial and radial vibration readings. The
axial readings amplitude can be as much as 0.5 to 2 times the amplitude of the
radial readings. The radial readings can appear at 1x, 2x, 3x, and even higher
multiples of shaft turning speed. This is especially true when a coupling defect
is present. On certain types of equipment, however, the predominant vibration
occurs at 1xTS and can be confused with unbalance. Phase data is a valuable
tool in distinguishing between an unbalance or a misalignment problem.
Misalignment
Characteristics of Misalignment
< High axial levels
< High radial vibration at 1x, 2x, 3x, 4x, or higher multiples of turning
speed
< 180E phase difference in axial or radial directions across the
coupling
< Repeatable periodic time waveform with 1, 2, 3, or 4 clear peaks per
revolution
< Excessive bearing temperatures
< Premature bearing failures
< Coupling failures
< Oil leakage at the seal
< Excessive amounts of grease on the coupling guard
< Shaft breakage at or close to the inboard bearings
Shaft Center Line Misalignment

Two types of shaft center line misalignment are discussed below — angular
and offset or parallel misalignment. Most misalignment problems are a
combination of both.
Misalignment
Angular Misalignment
Angular misalignment occurs when the center lines of two shafts meet at an
angle. The shafts can be angled in the vertical or horizontal planes. The
spectrum typically has a dominant peak at 1xTS in the axial direction. Radial
readings in the spectrum will show peaks at 1xTS and 2xTS if offset
misalignment is also present. The highest vibration usually occurs in the
direction of the angularity, which means that if the angularity occurs in the
horizontal direction, the highest vibration should also appear in the horizontal
direction.
Misalignment
The waveform characteristically is sinusoidal with one or two clear cycles per
revolution. The waveform amplitude in g’s is low and does not show much
impacting.
Misalignment
The typical phase relationships for angular misalignment are:
< The horizontal phase

relationship from A to B is 180E
± 30E.
< The vertical phase relationship

from A to B is 180E ± 30E.
< The axial phase relationship

from B to C is 180E ± 30E.
Note: Do not forget to take the

transducer orientation into
consideration on axial readings.
Confusion here could lead to
making the wrong call.
Misalignment
Offset Misalignment
With offset misalignment, the shafts are parallel but displaced from each other.
The shafts can be offset in the vertical or horizontal planes. The spectrum
shows a predominant 2xTS peak in the radial direction. The highest vibration
usually occurs in the direction of the offset — if the offset occurs in the vertical
direction, the highest vibration should also appear in the vertical direction.
Since the coupling forces the two offset shafts to meet, it introduces some
angular misalignment. This can cause a 1xTS peak in the radial and axial
direction that exceeds the 2xTS peak.
Misalignment
The waveform is characteristically sinusoidal with one or two clear cycles per
revolution. The waveform amplitude in g’s is low and does not show much
impacting.
Misalignment
The typical phase relationships for Offset Misalignment are:
B A
D C
< The horizontal phase
relationship from B to
C is equal to 180E ±
30E. 0E 180E
< The vertical phase

C is equal to 180E ±
30E.
Misalignment
Shaft Component Misalignment

A partial list of possibly misaligned components are listed below:
Ô Pulleys Ö Couplings
Õ Gears × Bearings
This section will cover only couplings and bearings. Pulleys and gears are
discussed in detail later in this course.
Coupling Misalignment
Coupling misalignment occurs when two shafts are misaligned, the coupling is
worn, or the coupling was installed improperly. The coupling can be
misaligned vertically or horizontally. The spectrum can have multiples of 1x to
6xTS and harmonics in both the axial and radial directions depending on the
type of coupling used. Once again, the highest vibration usually occurs in the
direction of the defect.
The appearance of the waveform will be dependent on the fault type and the
coupling type. It can vary from very periodic and sinusoidal to very random
with impacting.
Misalignment
Phase readings will typically be a combination of the offset and angular

characteristics.
Bearing Misalignment
Bearing
Center Line
Induced Bearing
Angle
Shaft
Center Line
Bearing misalignment occurs when the bearings are not mounted in the same
plane; they are cocked relative to the shaft or if the machine distorts due to
thermal growth or soft foot. The spectrum normally has the highest peak at
1xTS, 2xTS, or the number of rollers xTS in the axial direction.
Misalignment
The waveform often shows a truncated or flattened pattern indicating a rub. It

may also appear periodic and sinusoidal with low amplitude.
The typical phase relationships of bearing misalignment are:
< The axial phase

relationship from A to
C is equal to 180E±
20E.
< The axial phase

D is equal to 180E ±
20E.
Bent Shaft
Vibration can vary depending on where the shaft is bent. If the shaft is bent
near the center, the spectrum has a dominant peak at 1xTS in the radial and
axial direction. It can, however, appear at 2xTS if the shaft is bent near the
coupling. When the rotor is spun up past the first critical, the amplitude will
continue to rise exponentially. When speed is reduced, the amplitude will
decrease drastically. If this was due to imbalance, the amplitude would not
decrease as fast. Remember, this is above the first critical.
The waveform will resemble a mixture of misalignment and imbalance. Keep

in mind that the shaft is actually out of balance.
Misalignment
The phase characteristics of a bent shaft are:
0E 0E
B A
< The axial phase D C
relationship from A to B
is equal to 180E ± 30E.
0E 180E
B A
D C
B A
D C
or
< The horizontal-to-vertical

90E or 270E
phase relationship on A Horizontal to Vertical
and B is equal to 90E or
270E ± 30E.
A1
D1 B1
< The axial end phase C1

relationship from A1 to Axial End Phase
C1 and B1 to D1 is equal
to 180E ± 30E.
Remember that the phase data will be 180E out across the bend, not across the
coupling.
Misalignment
Case Histories - Misalignment

Misalignment Case #1 - Quench Fan
Ô 100-HP, variable-speed Motor.
Õ Direct-driven, overhung fan.
Ö High axial readings on Fan Outboard bearing point.
Misalignment
The multiple plot above came from a case history involving a fan. The data
displayed is from the outboard fan bearing. Notice that the highest amplitude is
in the axial direction at fan running speed. The spectrum for the fan axial point
appears below.
A full-screen plot of the fan axial data is shown above. The cursor marks the
running speed peak of the fan. Note that the axial vibration amplitude is six
times the amplitude of the radial readings. If the axial measurement point was
not taken it would have been difficult to diagnose this problem as angular
misalignment.
Misalignment
The time waveform for the fan outboard axial data helps to confirm the
problem on this machine. Vertical lines on the plot show the time required for
the fan shaft to complete one revolution. Evidence that the vibration is being
driven by a repetitive force is demonstrated by the presence of one or two
peaks per shaft revolution. As discussed, the waveform is repetitive and lacks
complexity.
Misalignment
Misalignment Case #2 - Main Cake Mixer
Ô 75-HP, four-pole motor.
Õ The motor drives a double reduction gearbox.
Ö High vibration at motor running speed was present on all

measurement points.
× The 1xTS vibration was primarily in the horizontal direction.
Misalignment
High 1xTS vibration amplitudes exist on the motor and gearbox in this
example. The gearbox is direct-driven from the motor. Notice that the 1xTS
vibration appears primarily in the horizontal direction with amplitudes
exceeding 1.0 IPS.
The full-screen plot of the motor inboard horizontal spectrum is shown above.
Severity of the defect is easy to determine based on the 1xTS amplitude
exceeding 1.3 IPS.
Misalignment
A horizontal and an axial reading from the gearbox input shaft are displayed.
Again, the axial reading confirms that the defect is angular misalignment. In
what direction is the unit angularly misaligned? The amplitude in the axial
direction does not have to “exceed” amplitudes in the radial directions for
angular misalignment. Just compare radial to axial ratio, if the axial amplitude
is 50 percent of the radial direction, possible angular misalignment.
Misalignment
Misalignment Case #3 - Process Fan
Ô 250-HP Motor.
Õ High horizontal amplitude on fan bearing one.
Ö 0.025 to 0.030" runout was measured at the FOH point
× 1xTS in the horizontal direction.
Misalignment
In the example above, peaks at running speed indicate a bent fan shaft. Fan
Inboard Horizontal (F1H) appears to have the most energy. Note that the 1xTS
vibration is primarily showing in the horizontal direction. Looks like
unbalance. Phase data was a key to solving the problem. 180 degrees shift
from top to bottom of bearing at FOH. 180 degrees side to side.
The spectrum and waveform above are from measurement point F1H. Notice
the lack of any multiples of shaft turning speed in the spectrum. The waveform
displays a repeatable, sinusoidal pattern.
Misalignment
Misalignment Case #4 - Separator Water Pump #1
Ô 15-HP motor, 1800 RPM.
Õ Flexible type coupling. Direct Drive, Centerhung Pump
Ö High 3xTS vibration.
The trend plot for motor’s 3 to 4xTS band is shown above. Look at how the
amplitude has increased over the past four collection periods. This increase in
vibration is from angular misalignment.
Misalignment
The spectrum for Motor Inboard Horizontal (M2H) and Motor Inboard Vertical
(M2V) are shown above. Notice the 3xTS amplitude is consistent between the
horizontal and vertical reading.
Misalignment
The axial reading also shows the angular misalignment defect. In fact, its
amplitude is more than half the amplitude of the radial readings. The
waveform shows the sinusoidal pattern with three distinct peaks per one-shaft
revolution. This repetitive shape, however, lacks complexity, and the
amplitudes of the peaks remain relatively low.
Misalignment
Misalignment Case #5 - Separator Water Pump #2
Ô 15-HP, 1800-RPM AC Induction Motor.
Õ High 3xTS vibration on the motor and pump.
Ö The unit has a three-jaw coupling.
Misalignment
The multiple plot spectrum above shows the measurement points most effected.
A cursor marks the 3xTS peak on the motor inboard axial point. Notice the
amplitudes from the 3xTS vibration exceeds 1.0 IPS.
The motor inboard measurement points are shown above. Some looseness is
now seen in the spectra. The presence of looseness helps distinguish this
coupling defect from simple misalignment. The next page shows the motor
inboard horizontal spectrum.
Misalignment
The cursor marks the 3xTS peak measuring an amplitude of over 1.3 IPS.
Notice the presence of 3xTS harmonics. The amplitudes of this coupling
defect dictates a high priority of concern.
The waveform from the previous spectrum is displayed above. The waveform
does show a dominant 3xTS pattern. However, the looseness drives the
randomness and complexity.
Misalignment
Misalignment Case #6 - Soft Foot on Motor
Ô Rolled steel frame motor vs. cast iron
Õ Motor just installed and laser aligned
Ö High 1x RPM in radial directions
Misalignment
The multi-point plot above shows the motor points. The radial positions show
1x TS dominant with many harmonics.
The pump points also show high radial vibration with 1x TS predominant.
Misalignment
The MOH position shows the severity of the defect. A running soft foot check
was performed by using the monitor mode to monitor the spectrum while
loosening the motor hold down bolts one at a time. When the front motor bolt
was loosened, the vibration dropped to less than .1 in/sec. The soft foot was
binding the motor bearings and causing misalignment. The harmonics are the
result of forced looseness.
Section Six
Looseness
Objectives
< Identify looseness in machinery by examining both spectral and
waveform data.
< Understand that looseness can include several different variations of

mechanical and structural looseness.
What is Looseness?
Two Types, Structural and Component Looseness
Mechanical looseness occurs when structural or rolling element components

do not fit properly. A large number of TS harmonics characterizes looseness in
the spectrum. In some cases and stages, looseness exhibits subharmonic
multiples of ½xTS. The highest amplitude typically appears radially in the
vertical direction. If a coupling is worn, however, axial looseness becomes a
possibility.
The waveform shows a great deal of energy and impacting in a random, high
frequency pattern. You may also see truncation of the time waveform on rare
occasions.
The phase associated with looseness varies from one measurement point to
another, and it is unsteady. If the rotor is loose on the shaft, it may slip each
time the machine starts, causing the phase to shift from one startup to the next.
Mechanical looseness has several stages, as shown on the following page.
Looseness
Stages of Looseness
Stage Stage Stage

1 4 7
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0
Frequency in "Orders"
Frequency in "Orders" Frequency in "Orders"
Stage Stage
2 5
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0
Stage Stage
3 6
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0
Looseness
Loose Bearing Housing

Loose bearing housing is caused by looseness of the bolts holding the
pillowblock bearing down. This is also called a wobble effect because of its
physical and spectral characteristics. Cracks in the bearing pedestal or frame
structure can cause this vibration as well.
The spectrum demonstrates a pattern of synchronous harmonic peaks similar to

mechanical looseness. The dominant frequency, however, often appears at
4xTS radially as shown below.
The waveform looks like mechanical looseness because of its random and
nonperiodic nature, and it displays significant impacting and energy. The
phase is unstable.
Looseness
Loose Machine Feet

Loose machine feet, also called soft foot, is caused by a structural weakness or
looseness in the foundation, baseplate, or machine feet. The base or frame may
have become distorted, grouting may have deteriorated, or the hold-down bolts
may have become loose at the base.
The dominant frequency in the spectrum is at 1xTS radially. The waveform is

periodic and should demonstrate a strong cycle-per-shaft revolution.
The phase measurement provides the determining factor to distinguish this

condition from imbalance and other 1xTS problems. Measure the phase of the
frame or base, baseplate and machine foot vertically. If the phase shifts 180E
between those measurements, soft foot is the source.
Looseness
Looseness
Rotor Rub Shaft Rotation
1xTS
2xTS
3xTS
4xTS
Rotor Rub
Rotor rub is often a significant problem in rotating machinery and is frequently
caused by Looseness or Misalignment. Rubbing occurs when rotating elements
come in contact with stationary components of a machine.
The pattern in the spectrum demonstrates fractional subharmonics of TS

radially, usually harmonics of DxTS. The nature of rubbing and impacting
components would explain the common excitation of rotor and/or component
natural frequencies.
The pattern in the time waveform is quite distinct. The waveform appears
truncated or flattened on one side. The truncation demonstrates the lack of
freedom of movement incurred by the rotor as it rubs the stationary component.
The phase shows constant change. Whether or not the rotor rubs throughout
the entire revolution or just part of it determines how significantly the phase
changes.
Looseness
Looseness
Looseness Case Histories

Looseness Case #1 - Stack Fan
Looseness
This data shows a trend plot for a fan inboard bearing from the 3 to 12 xTS
band. Notice the plot shows an increase for the data collected on 10-APR-93.
The figure below shows a multiple plot for the previous month’s data.
The multiple-spectrum plot for the fan point F1V is displayed. The data shows
an increase in vibration energy over the three collection periods. Notice the
1xTS peak begins to increase in amplitude in March.
Looseness
The plot above shows the April spectrum for the fan inboard bearing. This plot
shows many harmonics of shaft turning speed. These harmonics are whole
number multiples of shaft turning speed and are classified as synchronous
peaks. The data is an illustration of a looseness problem.
The waveform data for the F1V measurement point is displayed above. Notice
the energy level is low, only 0.2 g’s. The low amplitudes and little impacting
indicate that the looseness is not yet severe. The typical looseness waveform is
shown above by its erratic and nonrepetitive shape.
Looseness
Looseness Case #2 - Combustion Air Fan
Looseness
The multiple-points plot shows the spectra for the radial motor points. The
cursor marks the 1xTS peak on the motor inboard vertical (M2V) measurement
point. Most of the energy appears to be near motor turning speed.
The single spectrum showing measurement point M2V is displayed above. The
cursor marks the motor running speed at 29.84 Hz or 1 order. The harmonic
cursor marks multiples of turning speed past 15 orders. This spectral pattern
signifies machine or component looseness.
Looseness
Looseness Case #3
Õ Overhung fan, direct driven.
Looseness
The above multiple-point plot shows all the motor points. Most of the activity
is occurring at the inboard points.
The above single spectrum is shown for point motor inboard vertical (M2V).
Several multiples of turning speed are identified by the harmonic cursor. This
is an indication of looseness.
Looseness
The above spectrum is an expanded plot of the previous spectrum. Notice the
pattern shows an increase in amplitude of the harmonic peaks as you move
away from shaft turning speed.
The waveform shows the impacting and ringing down pattern. Also, notice the
level of g’s indicated in the waveform. Why is the waveform showing this
pattern instead of a random looseness pattern? This is an indication of bearing
looseness showing impacting in the waveform, as opposed to structural
looseness. Structural Looseness typically shows more of a random pattern
in the waveform.
Looseness
The above spectrum was taken after the bearing was replaced. Notice the drop
in overall energy as well as the decrease in harmonic activity.
The waveform is showing very little activity after the bearing was replaced.
The impacting and ringing down pattern has disappeared.
Section Seven
Objectives
< Calculate the rolling element bearing defect frequencies.
< Recognize bearing defect frequencies in spectral data.
< Use the time waveform when diagnosing bearing defects.
< Determine defect severity based on spectral frequency patterns instead

of spectral amplitude alone.
Introduction
Rolling element bearings have specific bearing failure modes that can be
observed in the spectral and the waveform data. Bearing frequencies are
different than some other vibration frequencies present in a frequency spectrum
in that bearing frequencies are truly defect frequencies. These frequencies will
not appear unless the bearings are actually defective. Each failure mode centers
around a particular defect site and consists of defect frequencies that appear
because of defects in the bearing components.
Ball Pass Frequency
Outter Race (BPFO)
Ball Pass Frequency

Inner Race (BPFI)
Fundamental Train
Frequency (FTF)
Ball Spin Frequency

(BSF)
The four primary rolling element defect frequencies are classified by their
location.
C FTF ( Fundamental Train Frequency ) or Cage Frequency
C BSF ( Ball Spin Frequency )
C BPFO ( Ball Pass Frequency Outer Race )
C BPFI ( Ball Pass Frequency Inner Race )
Note: The BPFI will always be the highest calculated frequency for rolling
element bearings.
Displayed below is a page from the CSI bearing file.

C The BPFI is the highest calculated bearing frequency for each bearing.
C A rule to apply when selecting an F max on machines with rolling element bearings is:
BPFI x # of balls/rollers = F max
Calculation of Fundamental Rolling Element Bearing

Defect Frequencies
Case I Inner race is rotating and outer race is stationary (most common
industrial application).
S Bd
FTF ' @ (1 & @ cos )
2 Pd
Nb Bd
BPFI ' @ S @ (1 % @ cos )
2 Pd
Nb Bd
BPFO ' @ S @ (1 & @ cos )
2 Pd
Pd Bd
BSF ' @ S @ [1 & ( )2 @ (cos )2]
2Bd Pd
where:
RPM = revolutions per minute

S = speed, revolutions per second
FTF = fundamental train (cage) frequency
BPFI = ball pass frequency of the inner race
BPFO = ball pass frequency of the outer race
BSF = ball spin frequency
Bd = ball or roller diameter
Nb = number of balls or rollers
Pd = pitch diameter
= contact angle
Case II Inner race is stationary and outer race is rotating (e.g., front wheels
of some cars).
S Bd
FTF ' @ (1 % @ cos )
2 Pd
Nb Bd
BPFI ' @ S @ (1 % @ cos )
2 Pd
Nb Bd
BPFO ' @ S @ (1 & @ cos )
2 Pd
Pd Bd
BSF ' @ S @ [1 % ( )2 @ (cos )2]
2Bd Pd
where:
RPM = revolutions per minute

S = speed, revolutions per second
FTF = fundamental train (cage) frequency
BPFI = ball pass frequency of the inner race
BPFO = ball pass frequency of the outer race
BSF = ball spin frequency
Bd = ball or roller diameter
Nb = number of balls or rollers
Pd = pitch diameter
= contact angle
Two other relationships are also important:
Ô The sum of the BPFO and the BPFI is equal to the number of
rolling elements multiplied by the shaft turning speed.
BPFO + BPFI = # balls * TS
Õ For the Inner Race rotating, the FTF is equal to the BPFO divided
by the number of rolling elements.
FTF = BPFO / # balls
Ö For the Outer Race rotating, the FTF is equal to the BPFI divided
by the number of rolling elements.
FTF = BPFI / # balls
Obviously, each different bearing type has different bearing defect frequencies.
The frequencies depend on the bearing geometry. These frequencies should not
appear in the spectral and waveform data unless a defect actually exists.
The defect frequency amplitudes depend on the defect severity and the bearing
failure mode.
Vibration Data
Spectral Data
Measuring the spectral data can detect these defect frequencies. Not only are
the primary defect frequencies of interest when diagnosing bearing failures, but
also the frequency harmonics. The maximum analysis frequency (Fmax) during
data collection should be high enough to capture the higher frequency
harmonics. An Fmax set at or above a frequency equal to the BPFI multiplied by
the number of rolling elements allows all the bearing defect frequencies to be
captured. It is the detection of the higher frequency harmonics of the BPFI and
the BPFO defect frequencies that often appear during the first failure stages of
bearing failure. It is the initial detection of these defect frequencies that allows
an analyst to track the bearing defect.
Each of the four primary bearing defects can be considered to be a particular

mode of failure.
Ball Pass Frequency Outer Race Defect
The spectrum above shows a bearing with outer race defect frequencies
marked.
Ball Pass Frequency Inner Race Defect
The spectral data above is from a bearing with inner race defect frequencies
marked.
Rolling Element Defect (Ball Spin Frequency)
The spectrum above shows defect frequencies related to a rolling element

defect.
Cage Defect
The spectrum shown above shows the cage defect frequency appearing at less
than 1 x turn speed. At about .3 xTS. This is normally where a cage defect
appears on a new bearing with a cage defect.
On a bearing that has been in service for some time a cage defect normally
appears as a sideband about a BPFO or a BPFI defect frequency.
Time Waveform
The time waveform should not be neglected when analyzing rolling element
bearing defects. The waveform amplitude pattern assists the analyst in
confirming some bearing defects. The time waveform shows an impact and
ring down pattern as the rolling elements pass a defect site.
The waveform above shows classic impact and ring down due to the rolling
element impacting the defect site.
Understanding Amplitude Levels

It is helpful to know what bearing types are in the machines in order to more
accurately identify the bearing fault frequencies displayed in the data.
However, once the defects are identified, it is not always possible to only trend
the amplitude change over time. Some bearings may fail when no defect
frequency exceeds 0.07 IPS, while others may continue to run at amplitudes of
0.25 or more IPS.
One reason for amplitude variations is the vibration energy transmission path.
Amplitudes for the same defect severity will be lower if the vibration has
traveled through several feet of the machine than if the sensor were mounted
just a few inches from the bearing. Also, inner race defects may appear at
lower amplitudes than outer race defects due to the different transmission path.
Another reason that amplitudes alone cannot be relied upon to provide accurate
defect severity is the differences in each machine’s mass, stiffness, and
damping properties. Even side-by-side machines with the same basic design
could have slightly different system properties that could affect the measured
amplitudes of similar sized bearing defects on each machine.
For example, the lower frequency amplitudes are affected primarily by the
machine’s mass. As frequencies increase, the machine’s mass has a smaller
effect on the vibration and the machine’s stiffness begins to have a controlling
effect on the vibration amplitude. Therefore, at the lower frequencies closer to
turning speed, Sir Isaac Newton's second law is more applicable. This law
states that a force is equal to its mass multiplied by the acceleration of the
mass. The equation is written:
F = M*A
For vibration analysis, we can use this equation in a slightly different form to
better understand one of the reasons bearings fail at different amplitudes for
different machines. This equation can be rearranged:
F/M = A
Examining this equation, assume the same force is present on two different
machines. Machine B is 10 times the mass of Machine A. If the equation is
rewritten:
MA * AA = F A,B = MB * AB
Since MB = 10 MA, the equation can be rewritten:
MA * AA = 10MA * AB
(MA * AA) / 10MA = AB
(1 / 10) AA = AB
Machine B’s acceleration level will be one-tenth of Machine A if all other

factors, including the force are held constant.
Bearing Life
The life of the bearing is inversely proportional to the load cubed and the shaft
turning speed based on the following equation:
H = (C3 / L3) * (16667 / RPM)
Where,
H = life in hours
C = bearing capacity in pounds
L = bearing load in pounds
RPM = shaft turning speed
An increase in the vibration levels increases the effective load, which therefore,
decreases the life of the bearing. Increasing the speed decreases the bearing
life. The increased shaft speed raises the bearing defect frequencies which
results in higher vibration amplitudes as the acceleration levels are amplified at
the higher frequencies. This results in higher forces experienced by the bearing
since the force is related to the mass and acceleration as discussed previously.
Slower shaft speeds have the opposite effect on bearing life. The faults develop
over longer time periods at lower amplitudes.
This is why vibration alarm levels that are adequate for some machines do not
work well for other machines. Also, this is the reason that some machines fail
at low amplitude defects and other machines tolerate higher amplitude defect
related vibration for the same level of defect severity.
Importance of Pattern Recognition

The bearing defect frequency pattern is really the key in determining the defect
severity for each bearing failure mode. The number of bearing-related
harmonic frequencies present, possible sidebands, and time waveform data can
be a much more reliable method of projecting needed bearing repairs.
Viewing time waveform data in acceleration units can be very helpful when
diagnosing bearing defects. Inner and outer race defects typically show
repeatable impacting patterns and ring down as the rolling elements are passing
over the race defect.
The acceleration amplitude in the time waveform is also a valuable tool to use
for diagnosing bearing severity. The acceleration amplitude is directly
proportional to the forces affecting the bearing vibration. Remember, F = M *
A. However, also remember that the machine’s mass affects the amplitude
variations from machine to machine. Usually time waveform acceleration
levels in excess of 2 G's, peak-to-peak is significant enough to warrant the
analyst’s attention.
Although vibration amplitude is helpful in determining defect severity, the

frequency pattern in the spectral data is often more indicative of progressive
bearing damage than changes in spectral amplitude alone. Now the discussion
that follows is the classic failure progression that applies to about 80% of all
bearings. You should be aware that other defect scenarios do exist. However,
the basic idea that amplitude alone cannot always be used to correctly identify
bearings ready for replacement is still true.
Failure Progression Pattern

The classical bearing failure progression pattern for most bearing race defects
can be described as follows:
Mode 1
The first signs of a defective bearing will appear in the spectral data, typically
at the higher frequency multiples of the fundamental bearing defect
frequencies. One of the reasons that these high frequencies show up first is
that the impacting of the rolling elements through the defect site creates high-
frequency broadband noise that excites the high-frequency natural bearing
resonances. The fundamental frequencies do not tend to show up at this point.
D = BPFI
(1xBPFI 7.88 Orders)
6 x BPFI
46.28 48.28
5 x BPFI
38.40 40.40 45.28 49.28
1.0 7.88 15.76 23.64 31.52 39.40 47.28 55.16 63.04

Frequency In "Orders"
High frequency multiples will typically appear first with a race defect.
Mode 2
More harmonics of the defect frequency appear in the spectral data.
Continuing degradation often means modulation of the defect frequencies by
the shaft turning speed. The amplitude of the sidebands relative to the defect
frequency harmonics is very important at this point. Sideband amplitudes
exceeding the defect frequency harmonic amplitude can indicate significant
damage. Please note, that as the bearings degrade, the actual defect frequencies
are not going to exactly match the precalculated defect frequencies. This is
because degradation of the bearing has caused the internal bearing geometry to
deviate from the original design.
D = BPFI
1 x RPM
6 x BPFI
5 x BPFI
4 x BPFI
3 x BPFI
2 x BPFI
1.0 7.88 15.76 23.64 31.52 39.40 47.28 55.16 63.04

Defect progression includes multiple sidebands.
Mode 3
Mode 3 can be identified by the appearance of the fundamental bearing defect
frequency in the spectral data in addition to the defect harmonics and sideband
frequencies. Also, sideband frequencies may be present related to the BSF and
the FTF frequencies. The BSF and the FTF usually appear as sidebands of the
BPFI or the BPFO. This is found in the later stages of bearing degradation as
the rolling elements and the train have become damaged from constant contact
and contamination by material breaking loose from the defective raceways.
The remaining bearing life has certainly been limited at this point, depending
on the speed and the load.
D = BPFI
1 x RPM
6 x BPFI
5 x BPFI
4 x BPFI
3 x BPFI
2 x BPFI
1.0 7.88 15.76 23.64 31.52 39.40 47.28 55.16 63.04

The fundamental defect frequency now appears.
Mode 4
As the bearing continues to degrade, the internal clearances may begin to
increase, which accelerates the deterioration of the bearing components. This
allows more impacting to occur within the bearing, increasing the broadband
noise. The increased impacting appears in the spectral data as a raised noise
floor. The peaks in the data may seem to decrease in amplitude and become
less distinct as the noise floor increases.
D = BPFI
1 x RPM
6 x BPFI
5 x BPFI
4 x BPFI
3 x BPFI
2 x BPFI
1.0 7.88 15.76 23.64 31.52 39.40 47.28 55.16 63.04

The noise floor has increased, making the peaks appear to have decreased.
Failure is eminent at this point. Less than 2 percent of the bearing’s life
remains. If a crack develops in the inner or outer race, the internal clearances
may decrease. This could lead to a seizure of the bearing. Steps to schedule
the bearing's replacement should have been taken by this time. These steps
may include ordering a new bearing and scheduling a suitable repair time.
It is important to understand that each bearing may progress through the failure
modes at different rates. When a bearing defect mode is detected, data should
be collected more often so all the modes are captured in the database. Every
bearing failure is somewhat different. Often an analyst fails to capture some of
the failure modes because they occur between data collection intervals.
Lubrication Problems
A significant contributor to a defective bearing is improper lubrication. These
defects often excite the bearing natural frequencies. The defect frequencies
appear between 800 and 1200 Hz spaced 75 to 90 Hz apart as shown in the
spectrum below.
Summary
Relying on amplitude alone is a difficult way to predict bearing failure. Often
the bearing defect frequencies do not increase in amplitude as much as the
content of the spectrum changes, including more of the defect frequencies.
Provided data is collected correctly to a suitable maximum frequency range,

observing the failure progression pattern is a much more reliable method for
recommending bearing replacement than trying to rely on amplitude alone.
Using the analysis techniques discussed can provide a vibration analyst with a
better guide for determining when to replace defective bearings.
Bearing Defect Chart
(page one of three)
Frequency Defect Remarks
1-8 x BPFI inner race amplitude of harmonics often

defect exceeds that of fundamental
frequency; defects frequently
caused by forces from the rotating
element; 1xTS sidebands often
modulate it as bearing degrades
1-8 x BPFO outer race amplitude of harmonics often

defect exceeds that of fundamental
frequency
1-N x BSF ball or roller usually accompanied by race

defects; defects; the strongest multiple
sometimes frequently equals the number of
result from a defective rolling elements
broken cage (N
= number of
rolling
elements)
(page two of three)
FTF cage defect usually accompanied by other

defective components; may appear
as a difference frequency
1xTS modulation advance defects energy centers around initiating

(sidebands) or resulting in defect frequency, although this
significant alterations of frequency may disappear in
broadband energy bearing advanced degradation
geometry
sum and difference multiple extensive damage

frequencies defective parts
(sidebands)
involving RPM,
BPFI, BPFO, BSF,
FTF
(page three of three)
1-6xTS (1) excessive excessive clearances normally

internal accompanied by FTF modulating
clearances other frequencies; can also
significantly affect balance
sensitivity
(2) bearing frequently 3xTS or greater is the

turning on shaft predominant multiple
(3) bearing shows strong 1xTS and 4xTS

loose in housing
(4) misaligned generated frequency equals

bearing number of rolling elements xTS
900-1600 Hz inadequate amplitudes may grow to 0.1-0.2

region with 3-4 lubrication IPS; may result from excitement of
peaks separated by installed natural frequencies; can
80-130Hz also be caused by preloading or
heavy thrust load if lubrication does
not correct
Case Histories
Bearing Case #1
MOA, MOV, MOH
Ô 250 HP, 1200 RPM motor.
Õ Center Hung - double suction pump.
Ö Most of the energy in the axial direction.
Bearing Case #1
The multiple-point plot shows the outboard points on the motor. This is the
first time data was collected on this machine. The highest amplitudes are seen
on the Motor Outboard Axial (MOA) point. Notice the appearance of the
higher frequency peaks.
Bearing Case #1
The above plot shows the data from April to June on the MOA point. Notice
the amplitudes of the higher frequency peaks are increasing as the lower
frequency peaks are decreasing in amplitude. The next page shows a full
screen of the data taken in June.
Bearing Case #1
The above plot shows a full screen spectrum of the MOA point. The fault
frequency for the BPFO is displayed. Notice the primary BPFO is present in the
above spectrum. Typically, primary BPFO is not seen until the final stages of
the bearing defect, if it is seen at all. What should the waveform look like?
Bearing Case #1
The above plot shows the waveform from the MOA point. Notice the typical
waveform pattern of impacting and ringing down. However, the amplitude is
only about 0.9 g’s. This case illustrates why severity determination cannot be
based solely on amplitude levels.
Bearing Case #2
Õ These readings were taken for certification.
Ö Spectral data now shows a new problem.
Bearing Case #2
The above plot shows a multiple spectra plot of all of the Fan points. Notice
the higher frequency activity on the inboard points.
Bearing Case #2
The above plot shows only the inboard points. Notice the harmonics of 1xTS
and the high frequency nonsynchronous groups of peaks.
Bearing Case #2
The plot above shows a single spectrum of the FIV point. The bearing ID is
unknown, however, characteristics in the spectrum indicate an inner race
defect. Based on the spectral pattern, the waveform should be examined to
confirm the suspected inner race defect.
Bearing Case #2
The waveform for the FIV point shows significant impacting of ± 60 g’s.
Production required the unit to run through the weekend before any work could
be performed. The bearing failed within 24 hours of data collection.
Bearing Case #3
Õ Vibration increased from last month.
Ö Vertical mounted centrifugal pump with 7 blades.
× Motor is attached to pump by a spline connection.
Bearing Case #3
The above plot shows a trend plot of the overall parameter value for the Motor
Upper radial X (MUX) measurement point. The cursor is located at the March
5, 1993, point. The overall trend plot shows that the overall value increased on
March 5. A multiple parameter plot is needed to determine where the increase
occurred.
Bearing Case #3
The above plot shows the multiple parameter plot for the MUX point.
According to the plot, the parameter that caused the sudden increase in overall
value is the 3 to 4xTS parameter band.
Bearing Case #3
The trend plot above shows the 3 to 4xTS parameter band. This sudden
increase is cause for concern. The spectrum and waveform need to be analyzed
to determine the cause.
Bearing Case #3
The single spectrum above shows most of the energy in the spectrum is located
at 3xTS. The source of the 3xTS needs to be determined. Based on
information known (7 blades, no coupling), the analyst determined the 3xTS
must be a result of bearing housing looseness. He was determined to find out
what was causing this peak at 3.097 Hz.
This is a good example of using acceleration to help solve a possible bearing

problem. The units are now displayed in acceleration. Acceleration reveals
peaks at higher frequencies. What are these peaks?
Fault frequencies can sometimes be indispensable. The spectrum below

displays the fault frequencies for the BPFO.
The Analyst also used the peak list to help find the problem.
Note the percentage of nonsynchronous energy in this peak list. What was the
problem? This was an unusual case. When the machine was taken down for
repairs the motor outboard bearing did have BPFO’s and the bearing was also
loose in the housing. This would explain why the peak at 3.087 Hz was at such
a high amplitude (which was the fundamental frequency for the BPFO.
Unusual because of the amplitude of the fundamental BPFO compared to the
harmonics of that frequency.
Bearing Case #3
The waveform shows the sinusoidal pattern with three distinct peaks per shaft
revolution. This repetitive shape, however, lacks complexity and the
amplitudes of peaks remain relatively low.
This is a display of the spectrum after repairs were made. The vibration has
been reduced by more than ½ .
We have a completely different type waveform pattern also.
Bearing Case #4
Õ Overhung fan - direct driven.
Bearing Case #4
The above multiple-point plot shows all the motor points. Most of the activity
is occurring at the inboard points.
Bearing Case #4
The above single spectrum is shown for point Motor Inboard Vertical (M2V).
Several multiples of turning speed are identified by the harmonic cursor. This
is an indication of looseness.
Bearing Case #4
The above spectrum is an expanded plot of the previous spectrum. Notice the
pattern shows an increase in amplitude of the harmonic peaks as you move
away from shaft turning speed.
Bearing Case #4
The waveform shows the impacting and ringing down pattern. Also, notice the
level of g’s indicated in the waveform. Why is the waveform showing this
pattern instead of a random looseness pattern? This is an indication of bearing
looseness showing impacting in the waveform as opposed to structural
looseness which typically shows more of a random pattern in the waveform.
Bearing Case #4
The above spectrum was taken after the bearing was replaced. Notice the drop
in overall energy as well as the decrease in harmonic activity.
Bearing Case #4
The waveform is showing very little activity after the bearing was replaced.
The impacting and ringing down pattern has disappeared.
Bearing Case #5
Drive roll
Breast roll
Bearing Case #5
The above plot shows a multiple measurement point plot. Notice the high
frequency groups of peaks on point B1R.
Bearing Case #5
The bearing frequencies were calculated as FTF = .45, BSF = 4.36, BPFO =
10.25, and BPFI = 12.75. The single spectrum shown above has the fault
frequency for BPFO displayed. Six multiples of BPFO are present in the
spectrum.
Bearing Case #5
The single spectrum shown above has the BPFI fault frequency displayed.
Notice that both the BPFO and BPFI frequencies are present. What could the
spacing of the sidebands be equal to?
Bearing Case #5
The above spectrum is an expanded view of the spectrum from the previous
page. The sidebands are spaced around BPFO at 0.47 orders. The sidebands
are spaced at cage frequency. This is an indication of extensive bearing
damage.
Bearing Case #6
Õ Motor drives 2 hydraulic pumps.
Bearing Case #6
The above spectrum shows several harmonics present of 1xTS. If this

spectrum was not analyzed more closely, it could be misdiagnosed as a
looseness problem. The next page uses fault frequency information to identify
the real defect.
Bearing Case #6
The spectrum shown above has displayed the fault frequencies of BPFO on the
spectrum. Also, harmonics of BPFO have been shown with the harmonic
cursor markings. Notice that 16 harmonics of BPFO are present. This bearing
defect could be overlooked because of the absence of the typical bearing defect
pattern.
Bearing Case #6
Even the waveform does not show the typical pattern of bearing defect. This
waveform has fairly low energy levels and a random nonrepetitive pattern.
However, this waveform does show impacting of ± 1 G.
Bearing Case #7
Õ Center hung pump.
Bearing Case #7
The above multiple point plot shows all the pump points. Notice the high-
frequency mounds of energy between 800 and 1200 Hz. A full screen of point
P1V is shown on the following pages.
Bearing Case #7
The above full-screen spectrum shows a classic example of inadequate

lubrication. Most of the energy is located between 800 and 1200 Hz.
Bearing Case #7
The separation of the peaks is spaced at around 93 Hz. Typically the spacing
between the peaks are around 80 - 130 Hz.
Bearing Case #7
The waveform displayed above shows significant impacting in the waveform.
Bearing Case #8
Bearing Case #8
The multiple spectra plot shown above shows data taken on the same point, one
day apart. It appears that sidebands are present around the higher frequency
peaks. The following pages examine the single spectrum closer.
Bearing Case #8
The spectrum shown above is a full screen of the May 12, 1993, data. The
peak list on the following page shows that most of the energy in the spectrum is
nonsynchronous energy.
Bearing Case #8 Peak List

Machine: (BRG ) PICK-UP SECTION - DRIVE SIDE
Meas. Point: CRACKED IR-13A --> DRIVE ROLL #13 (DS) - AXIAL
Date/Time: 12-MAY-93 13:37:34 Amplitude Units: IN/SEC PK
Data Label: Inner Race Called on 13A
PEAK FREQUENCY PEAK ORDER PEAK FREQUENCY PEAK ORDER

NO. (Hz) VALUE VALUE NO. (Hz) VALUE VALUE
---- --------- ----- ----- ---- --------- ----- -----
1 .94 .0308 .24 13 104.01 .0911 26.42
2 1.66 .0194 .42 14 130.84 .0129 33.23
3 2.04 .0147 .52 15 138.68 .0496 35.22
4 26.83 .0181 6.81 16 141.00 .0102 35.81
5 34.67 .0433 8.81 17 146.53 .0104 37.22
6 38.59 .0099 9.80 18 173.35 .0605 44.03
7 42.51 .0323 10.80 19 200.18 .0175 50.84
8 46.42 .0122 11.79 20 204.10 .0230 51.84
9 69.34 .0203 17.61 21 205.71 .0202 52.25
10 96.17 .0124 24.43 22 208.03 .0551 52.84
11 100.09 .0216 25.42 23 242.70 .0167 61.64
12 101.71 .0116 25.83 24 245.01 .0118 62.23
TOTAL MAG SUBSYNCHRONOUS SYNCHRONOUS NONSYNCHRONOUS
.1669 .0371 / 5% .0748 / 20% .1445 / 75%
F1/Enter=Accept F2=Paging is OFF F7=Title F9=Copy Esc=Quit
Bearing Case #8
As shown in the spectrum above, sidebands spaced at 1xTS are present. The
characteristic of the sidebands spaced at 1xTS around a primary frequency is
associated with an inner race defect. Amplitudes are low but significant
numbers of harmonics and sidebands are present in the spectrum.
Bearing Case #9
530 Baghouse Fan
Belt Driven, Overhung
Bearing Case #9
Displayed is a multiple points plot. There appears to be a problem on both

ends of the fan, inboard and outboard.
Bearing Case #9
Full scale plot of FIH reveals several harmonics of runspeed. This could be an
indication of a Looseness condition on the machine. Structural or Component
Looseness ?
Bearing Case #9
There is a lot of impacting in the Waveform on FIH, 5 to 6 G swing. The
pattern of the impacting looks non-repetitive. What would this lead you to
assume as to the type of problem we have on this one? The Crest Factor is
3.76.
Bearing Case #9
Before you make a decision on this one, let’s take a look at the data collected in
the Vertical direction FIV. Again there are a lot of harmonics of runspeed in
the spectrum. Looseness?
Bearing Case #9
Let’s take a look at the Waveform for FIV. Notice in the vertical direction the
impacting is more of a repetitive type pattern. There is a 10 G swing and also
the crest factor is 5.11. Now what type problem do we have on this fan? If you
said bearing looseness, you are correct.
Bearing Case #9
Displayed in this plot is the horizontal and vertical directions of the outboard
end. There are again a lot of harmonics of runspeed.
Bearing Case #9
When we compare the waveform patterns of the outboard horizontal and
vertical readings we should be able to determine what type of problem we have
on this bearing.
The horizontal direction on this bearing indicates we have a bearing looseness

problem. When the bearings were changed they found both to be very worn
with defects on balls, inner and outer races and very loose.
Note: It pays to look at data taken in all directions when trying to determine a
problem on a machine.
Section Eight
Data Averaging
Objectives
< Use MasterTrend to set up different averaging methods.
< Describe each of the averaging modes available in the Model 2120.
< Determine specific applications for each of the averaging modes.
General Description
Each of the averaging modes lets you perform in-depth analysis in the field as
the need arises. The following lessons discuss each of the averaging modes in
detail. First we will show how to setup MasterTrend to collect route data with
normal as well as advanced averaging techniques. We will also show the use
of the analyze function to collect field data.
Data Averaging
Averaging Modes
( Valuable tools for Analytical Trouble Shooting )
Normal Averaging — Normally used for routine data collection and when
other averaging modes are inappropriate. Normal averaging enhances
repetitive events. --- no tach is needed ---
Peak-Hold Averaging — Use to measure a machine that varies in speed

(frequency) when you do not have a reference pulse (e.g., during coast down).
You may also use this mode to monitor random fluctuations such as steam
whirl in a turbine or oil whirl in a journal bearing. --- no tach is needed ---
Synchronous Time Averaging — Use to minimize all vibration not

synchronously related to the shaft providing the reference pulse. All vibration
at a whole number order of shaft speed (frequency) remains in the spectral data.
Amplitudes at nonsynchronous frequencies are reduced, virtually eliminating
vibration from other components on the same machine.
--- tach is needed ---
Order Tracking — Use to monitor machinery that continuously varies in

speed during the data collection process. This method requires a reference
pulse. --- tach is needed ---
Negative (Linear Minus) — Dynamically subtracts two spectra. This

method is one of the most powerful analysis tools available, because it
subtracts data while the machinery is in operation. No tach is needed.
Data Averaging
MasterTrend Setup - Data Averaging

All the data averaging methods listed on the previous page, with the exception
of Negative Averaging, are accessible in the Analysis Parameter (AP) Sets in
MasterTrend. Let’s go to INTMDVIB.DAT to see the setup.
Select Database Setup Management.
Data Averaging
Then select ADD/EDIT Old Information.
We will look at the AP sets for the various averaging modes.
Data Averaging
The first averaging mode is Peak-Hold.
The choice of averaging methods is made in the Spectral Averaging Mode.

The rest of the choices are shown.
Data Averaging
Acquire Spectrum - Lab

We can also access the various averaging methods from the Analyze, Acquire
Spectrum function. This lab will explain each Analyze function using the
analyzer.
Normal Averaging (also called linear averaging) uses a specific number of

averages — usually four to eight — that carry equal weight. The two types of
normal averaging are linear plus and linear minus. The linear plus averaging
mode is more common. Your routine data collection normally uses this
averaging mode. Linear plus proves most useful for attenuation of
nonrepetitive events.
Press the Analyze function key on the analyzer to display the Analyze
Functions menu as shown below. Highlight 5) ACQUIRE SPECTRUM and
press Enter to define the menu for data collection.
Data Averaging
Define page one of the three-page display as shown below. When you
complete page one, press the Page Up key to advance to the second page of
the measurement definitions.
Normal averaging is the most commonly used for data collection. In this
mode, the analyzer stores the amplitude measured for each cell (bandwidth).
The process then adds the next sample and divides the results by two. The
analyzer gives each succeeding sample the same weight. Define your menu as
shown below and press Enter to proceed with data collection.
Data Averaging
The third page of the Acquire Spectrum Menu is shown below. It contains the
option to choose demodulation or PeakVue analysis and the filter settings for
either. Press Enter to accept page 3.
The following screen displays. When ready to collect data, press Enter once
again to begin data collection.
Data Averaging
What the data being collected might look like is shown below. To display
these plots, press the Utility key on the analyzer and select DUAL DISPLAY.
A plot displaying the full view of the spectrum displays after the analyzer
completes the data collection.
Data Averaging
Press Enter to return to the Analyzer Functions Menu. Highlight 4) STORE

DATA and press Enter. The data will be stored at the measurement point that
was active when the Analyze function key was pressed.
Select the option that best suits the data that you actually want to save. Select
3) STORE BOTH and press Enter.
Data Averaging
To view saved data, highlight 1) DISPLAY SPECTRUM and press the Enter
key.
The spectrum will be displayed. Note the different function keys available.
The waveform can be accessed directly and the frequency units may be
changed easily.
Data Averaging
If more than one spectrum has been stored, you must select the spectrum to
view.
The waveform may be displayed directly by selecting this option from the
Analyzer Functions Menu.
Data Averaging
In this case, multiple waveforms have been saved. Notice that the length of
the waveform displays. Highlight the waveform you want to display and press
Enter.
The waveform displays. The function keys allow for various operations.
Data Averaging
Acquire Spectrum - Definition

Let us examine each of the acquire spectrum options more closely. Press the
Analyze function key at the top of the analyzer. Select the ACQUIRE
SPECTRUM option. PAGE 1 of 3 is the first screen to discuss. The
Frequency selection should be set above the highest frequency of interest.
This is the Fmax.
Frequencies below the Low Cutoff will not be included in the calculation of the
overall measurement nor will they be displayed in the spectrum. This value is
often set at a frequency equal to two times the Fmax divided by the number of
lines.
2 ( Frequency
Low Cutoff '
Number of Lines
Note: 2 Hz is a good selection for Low-End Cutoff for most data collection
Data Averaging
The lines of resolution divided into the frequency result in the Bandwidth
(BW), sometimes referred to as Resolution. Choices in the 2120 analyzer are
100, 200, 400, 800, 1600, 3200, 6400, and 1/3 Octave.
Frequency
Bandwidth '
Number of Lines
Two window choices are available — Hanning and Uniform. The Hanning
window is selected most often because it is best suited to accurate frequency
analysis. The Uniform window is usually reserved for transient analysis.
Data Averaging
The number of averages selected is typically set between 4 and 8 for normal
data collection. The actual values selected depend most on the averaging
mode.
The SST Control set to YES allows for low frequency data collection with an
amplitude correction applied to the low frequency data. Leave set to NO for
now.
Data Averaging
The Units / Integration option allows the analyst to select the spectral units
and the integration mode without having to go back to the Utility function key
/ Change Setup.
Data Averaging
The Initial Setup selection set to YES resets the Acquire Spectrum choices to
the values used in the current Route Point. Leave this set to NO.
Press the Page Up key to move to the second page of the Acquire Spectrum
Menu. In this case, the averaging mode selected is Peak Hold. The options on
this page will be discussed later as they become relevant to the measurement
being made.
Data Averaging
Press the Page Up key to move to page 3 of 3. When Demodulate is set to

YES, the analyzer’s built-in demodulator is activated for the measurement.
When enabled, the filter option is activated, allowing a choice of 500, 1000,
2000, 5000, and 10,000 Hz high pass filters or a 500 to 1000 Hz band pass
filter. These filters remove the signal below the setpoint, amplify the
remaining signal, and then demodulate it, resulting in low-frequency data up to
40 % of the filter set point in Hz.
Toggling the PeakVue option to YES, turns on PeakVue analysis. This allows
spectral analysis of very short duration spikes in the data typically missed in
normal spectra. When the PeakVue option is on, the prefilter option will
automatically enable, allowing different filters to be chosen.
Data Averaging
Peak-Hold Averaging
Peak-hold averaging retains the highest amplitude at every frequency from all
the averages acquired. It is most commonly used for coastdown data for all
frequencies when there is not a once-per-revolution (tach) event marker
available. Peak-hold averaging can also be used when amplitudes are unstable
from sample to sample. Press the Analyze function key to display the Analyze
Functions Menu. Highlight 5) ACQUIRE SPECTRUM and press Enter.
Peak-hold averaging retains the highest value for every spectral line. It uses
these values to develop an envelope that displays the highest amplitude
measured at each frequency during the entire sampling process.
Data Averaging
The number of averages should be set high. Use peak-hold averaging when
the vibration signal varies over a period of time. Peak-hold can also be used
when random fluctuations take place during the data collection period. Press
Enter to initiate data collection. Press Enter again when you are ready to
collect data.
The spectrum shown below illustrates peak-hold averaging. The highest

amplitude that occurred during the averaging process is retained at each
frequency.
Data Averaging
Synchronous Time Averaging

Synchronous time averaging collects time waveforms synchronized to a
tachometer (tach) pulse. The synchronous time waveforms average in the time
domain. The resulting spectral display is the FFT for the averaged waveform.
Use synchronous time averaging when you need data directly related to the
turning speed of a specific shaft. For example, use synchronous averaging
when you have several machines (or shafts on one machine) that all turn at
different speeds. It is important to remember that many faults generate
nonsynchronous energy and they will be reduced with this averaging. This
averaging will enhance turning speed and harmonics of turning speed.
Note: Synchronous averaging reduces but does not eliminate nonsynchronous

vibration in the spectral data. The amount of reduction depends on the
number of averages you take. Use the formula below to calculate the
results:
Reduction Factor ' Number of Averages
For example, 100 averages reduce the nonsynchronous vibration by a

factor of 10, and 10,000 averages by a factor of 100.
It is important to understand that the energy averaged out using Synchronous

Time Averaging can be sub-synchronous, synchronous, or nonsynchronous
energy from another source or machine that you are not collecting data on.
Also, sub-synchronous and nonsynchronous energy will be reduced or
eliminated from the spectrum that you are collecting the data on. Only the
synchronous energy that is occurring on the machine you are collecting data on
using Synchronous Time Averaging with a photo-tach will remain in the
Spectrum.
Data Averaging
Synchronous vs. Harmonic

Vibration signals at integer multiples of a primary frequency (any frequency
where a peak in the signal may occur) are called harmonics of that frequency.
< Harmonics = N x f, where f is the primary frequency, and N = 1, 2,

3, etc.
< Synchronous peaks occur at integer multiples of the operating

frequency of the shaft at the measurement point (1xTS, 2xTS,
3xTS. . .).
Sample No. 1
Synchronous Time Averaging
Sample No. 1: Based on the TTL pulse, the analyzer
stretches or compresses time windows
to match as shown.
Sample No. 2: The process preserves synchronous

Sample No. 2 peaks (harmonics of turning speed).
Mixed fractional multiples (1 1/2,
1 1/3. . .) and noninteger multiples
(bearing defects, electrical frequencies,
etc.) are averaged down.
The reduction factor depends upon the total number of

averages selected.
Sample No. 2 Converted
Note: The more averages you select, the more time the
analyzer has to decrease those vibration
components that are not phase locked to the
rotational speed of the equipment being
measured.
Averaged Waveform
Synchronous peaks are harmonics of shaft frequency. You may also have
harmonics of nonsynchronous peaks as in the case of roller bearing defect
frequencies, but these peaks are not synchronous.
Data Averaging
Press the Analyze function key to display the Analyzer Functions Menu.
Highlight 5) ACQUIRE SPECTRUM and press Enter.
Define the three-page menu as shown on the screens here. On page one,
specify 50 averages. The analyzer reduces the nonsynchronous energy by the
square root of 50, a reduction factor of 7.07.
Data Averaging
Press the Page key to advance to page two of the menu. Toggle Average
Mode to SYNCHRONOUS. The analyzer trigger mode must be toggled to
TACH. A tach signal is required to average this data.
Press Enter to begin the data collection. The analyzer automatically displays
the averaged spectrum when collecting data. After completing the collection
period, the analyzer displays the final average. The spectral display below was
collected with normal averaging. Note the sideband energy. Even though
those sidebands indicate a defect frequency, the synchronous time average data
eliminates them from the plot.
Note the
sidebands
next to the
harmonics of run
speed.
Data Averaging
The spectrum shown below is the same machine seconds later. With the
synchronous time averaged data, the sidebands and all other nonsynchronous
data almost disappear.
Note the
sideband
energy is
no longer
present.
Data Averaging
Order Tracking
Order tracking automatically adjusts the analyzer during data collection to
account for the changes in machine turning speed (RPM) using a reference
tach pulse. Use this function primarily on variable speed machines.
When a machine’s speed changes, the peaks associated with certain faults
often get smeared over many frequencies on the spectral display. It may be
difficult to determine what peak corresponds with 1xTS, 2xTS, etc., making it
impossible to identify bearing fault frequencies, belt frequencies, and other
characteristic fault frequencies. Order tracking, however, automatically
adjusts for changes in the machine RPM to prevent smearing these peaks.
The illustration below shows how order tracking averages spectral data. With
a total of three averages and a speed variance of less than 6 percent between
samples, the order tracking keeps up with both amplitude and speed. Once the
data collection finishes, the analyzer averages the data together in both the X-
axis (frequency) and Y-axis (amplitude). Order tracking also stretches or
compresses the size of the time window so that all windows match.
Data Averaging
Order Tracking Averaging
Sample No. 1 Sample No. 2
Sample No. 2 Converted
FFT From FFT From

Sample No. 1 Sample No. 2
Averaged Spectrum
Data Averaging
Define the three-page menu. On page one, specify 25 averages.
Data Averaging
Toggle the Average Mode to ORDER-TRACK. Define the remainder of

page two as shown below.
Press Enter to begin data collection. The spectrum was collected with Order
Tracking. The average data displays with the frequency axis graduating in
orders.
Data Averaging
Linear Minus Averaging

Linear minus (negative) averaging dynamically subtracts noise from an
averaged spectrum. This method is one of the most powerful analysis tools
available, because you can subtract data while the machinery is in operation.
Note: Noise is any vibratory energy undesirable for analysis. Signals that may
be classified as noise include similar machinery vibration on a common
base plate, belts, fans, and resonance.
The improvement of the signal-to-noise ratio gained by linear minus averaging

is proportional to the square root of the number of averages performed. This
improvement approximates -3dB for every number of averages equal to an
even power of 2. For example:
• Measurement 1: Averages = 2 = 2 (n = 1)
Measurement 2 results in -6dB improvement in the signal-to-noise ratio.

Measurement 3 results in -9dB improvement in the signal-to-noise ratio.
Negative Averaging is a three-step process.
Step 1 Step 2 Step 3

Machine Running Machine Running Machine Running
Impact Machine No Impact No Impact
Normal Averaging Negative Averaging Step 1 - Step 2
Spectrum Displays: Spectrum Displays: Spectrum Displays:
Normal Vibration + Normal Vibration Result of Impact

Result of Impact
Data Averaging
Negative Averaging Case History

The case history provides an example of linear minus averaging. The data
comes from a belt driven fan assembly involved in a dynamic
impact/resonance test. The multiple spectral plot below shows data taken from
the same machine at different periods of time. The top plot shows impact data
collected with the machinery off. The middle plot shows data collected with
the linear minus function. The bottom plot shows data collected with the
normal averaging mode.
Data Averaging
As discussed earlier, the ability of the linear minus function depends on the
time given for noise subtraction. The ultimate objective of this function is to
subtract any vibration data from the plot. If given enough averages, the
unwanted data draws down to the noise floor of the analyzer (10-16).
The support data shown in the next three plots appears in the order the analyst
gathered it. The analyst collected data during the machine’s normal run
condition. He used normal averaging while impacting the structure
approximately every two seconds. The analyst used this data for the dynamic
impact information. After he collected enough dynamic impact data, the
analyst ended the normal averaging mode and stored the data.
Data Averaging
At this point, the analyst changed the averaging mode from normal (linear
plus) to linear minus.
Note: The machine’s operating condition was not interrupted. Also, note the
absence of the operating peaks located in the previous plot. The analyst
subtracted them from the data plot above. This function left only that
data directly caused by the impacts.
Data Averaging
The data below is the impact on the machinery in the off condition. Note that
this plot and the preceding plot are almost identical. The only differences are
the holes left from the linear minus function during operation.
This analysis technique lends itself to a wide range of machinery applications.

This case history illustrates that the Negative Averaging technique proved that
the machine was running near resonance and, when the machine was shut off,
the resonant frequencies were confirmed.
Data Averaging
Define the menus as shown. Negative Linear Averaging collects data first
with normal averaging using the setup parameters given on the first page of the
setup menu.
Data Averaging
Press the Page key to advance to the second page of the menu. Toggle the
Average Mode to NEG-VE AVRG. Press Enter to initiate data collection.
Press Enter again to start the data collection.
Data Averaging
Begin impacting and press Enter. It is important to impact the structure while
the analyzer is autoranging. This is a typical view of Step 1. Try not to get
more than one impact per average. It is generally better to get no impacts in
one average than to get two impacts in one average.
After collecting the data with normal averaging (with impacts), the analyzer
displays a message. Press Enter to start the data acquisition with the linear
minus averaging (without impacts).
Data Averaging
During Step 2, watch as the operational frequencies begin to decrease in

amplitude.
After you are convinced that the operational frequencies have been sufficiently
reduced, press Enter to stop the second step of data collection. The final step
is the spectral displayed of the linear minus averaged spectra (data directly
caused by the impacts). This spectra shows that a distinct resonance appears
just below run speed. This data may be stored and displayed just like any
other.
Data Averaging
Data Collection Setups - Summary

The information below defines the functions of the Acquire Spectrum Menu.
Each setup depends on the other. Therefore, it is important that you
understand all the setup options.
FREQUENCY: — Specifies the maximum frequency, Fmax, for data

collection and display.
LOW CUTOFF: — Defines the lowest frequency displayed by the analyzer.

Frequencies below this are still measured.
LINES: — 100, 200, 400, 800, 1600, 3200, 6400, 1/3 Octave.
WINDOW: — HANNING prevents leakage and is used most often;

UNIFORM applies no shaping function and is often used for transient testing.
AVERAGES: — The number of samples to be collected. Selection depends

on the Averaging Mode chosen.
SST CONTROL: — Activates a Low Frequency amplitude correction for

Slow Speed applications.
UNITS/INTG: — Allows the setup of the spectral units and integration

method without going to the Utility key and using the Change Setup option.
INIT SETUP: — Sets all the Acquire Spectrum choices to a response that
matches the Route Point setup.
Data Averaging
The second page of the setup is not usually changed for most routine
additional analysis, but you need to understand each of these choices.
AVERG MODE: — Select between Normal, Peak Hold, Synchronous Time,

Order-Track, and Negative Linear averaging modes.
TRIG MODE: — Select the type of trigger to be used for data collection.
Select between Off, Normal, Tach, Pre-Trg, Pre-Tach.
TRIG LEVEL: — Units depend on trigger mode selected. Normal and Pre-
Trigger modes require a vibration amplitude. Tach and Pre-Tach require an
RPM value.
% PRE TRIG: — Percent of time waveform that will be collected before the
trigger occurs.
FS RANGE: — Full scale range prevents autoranging, most useful for

transient testing.
A-WEIGHT: — Frequency shaping method that approximates the response

of the human ear. This is often used for airborne noise measurements.
TRIAX CTRL: — Can be used to select individual channels of a Triaxial

accelerometer.
Data Averaging
Page 3 of the acquire spectrum setup contains the DEMODULATE,

PEAKVUE, and PREFILTER options.
DEMODULATE: — Set to YES activates demodulation of the signal using

various high pass filters.
PEAKVUE: — Set to YES activates the PeakVue option, which allows the
detection of very short duration, high-frequency impacts, which can be used to
detect very early rolling element bearing defects and gearbox faults.
PREFILTER: — Allows the selection of various high pass, low pass, and
band pass filters for both demodulation and PeakVue.
Data Averaging
PeakVue Filter Guidelines
50 Hz -- 5 KHz : Felt problems, low to intermediate speed

gearboxes
500 Hz --- 10 KHz : Slow speed machinery (<900 rpm) including

gearboxes with gear mesh frequencies <150 Hz
1.5 KHz --- 15 KHz : Intermediate speed machinery (up to 1800 rpm)
and gearboxes with gear mesh frequencies up to
400 Hz
5 KHz --- 40 KHz : High speed machinery ( > 1800 rpm ) and
gearboxes with gear mesh frequencies up to 2
KHz
Tip : To determine filter selections for gearboxes, select the filter whose
lower corner frequency is greater than or equal to 3.25 times the largest
gear mesh frequency.
Bandpass Filters ( 2120 Analyzer )
20 Hz ---- 150 Hz : Felt problems on paper machines, etc:
50 Hz ---- 300 Hz : Certain structural Resonance excitation,

modulation of gear mesh in low speed
machinery
100 Hz ---- 600 Hz : Gear mesh modulation in intermediate speed

machinery
500 Hz ---- 1 KHz : Gear mesh modulation
Data Averaging
PeakVue Filter Guidelines
High Pass Filters
500 Hz : Low speed machinery having gear mesh <125Hz. Bearing and Gear
problems
1 KHz : Intermediate speed machinery (<2000 rpm with gear mesh <300 Hz)
2 KHz : Medium speed machinery ( <4000 rpm with gear mesh <600 Hz )
5 KHz : Machinery up to 9000 rpm and gear mesh to 1500 Hz. Requires
attention to how the sensor is mounted as well as the sensor’s
frequency response.
10 KHz : High speed machinery with gear mesh up to 3000 Hz. Sensor
"must"be permanently mounted with frequency response of 3 dB in
the 30 KHz or higher range.
20 KHz : High speed machinery with gear mesh up to 6000 Hz. Sensor must
be high frequency and permanently mounted.
Tip : Generally, use
Band Pass Filters when the event of interest is the excitation of a

structural resonance, or the modulation of
known frequencies, such as gear mesh.
High Pass Filters when the objective is to detect stress waves

emitted by metal to metal impacting, fatigue
cracking etc;
Data Averaging
PeakVue Data
Notice, in plot above, the primary cursor is marking the turn speed at 5 rpm.
In this plot displayed above the cursors are marking harmonics of the BPFO.
We are viewing this data in acceleration. Notice the extremely low amplitude.
Data Averaging
Displayed below is the waveform from the spectral data on previous page.
The Analyst must realize the data from PeakVue is reading Stress Waves that
ordinarily would not be picked up by the transducer.
Also note that PeakVue rectifies the signal and places all the impacts on the
positive side of the waveform.
Data Averaging
Displayed below is a plot of another machine diagnosed with a bearing

problem using PeakVue technology.
Below is the corresponding waveform for the spectrum shown above.
Data Averaging
Zoom Analysis
Zoom Analysis can be used to place up to 800 lines of resolution between a
lower frequency and an upper frequency for in-depth frequency analysis over a
higher frequency range. Be aware of the fact that in some cases, better
resolution (a smaller bandwidth) can still be achieved using the Acquire
Spectrum option with 1,600, 3,200, or 6,400 lines of resolution. Select
7) ZOOM ANALYSIS from the Analyzer Functions Menu, and press Enter.
Set the Low Cutoff to a frequency below your frequency range of interest.
The Bandwidth defines the frequency range for data collection. Select the
HANNING or UNIFORM window, the number of Averages, and the desired
Trig Mode.
Data Averaging
After pressing the Enter key, the screen below appears. Press Enter to begin
data collection.
In this case 800 lines of resolution is being used over the frequency range of
500 Hz to 750 Hz. Depending on the preselected Low Cutoff and the
Bandwidth, the analyzer toggles to the nearest internally predefined selections
most closely matching your choices.
If the results of data collection are satisfactory press Enter and store the data.
Only the spectrum is available for storage.
Data Averaging
“True” Zoom Analysis

The zoom feature in the 2120 provides what the industry has come to call “true
zoom” capability. That is, the analyzer will collect a spectrum over a given
frequency at a preset resolution. In the 2120, the preset resolution is 800 lines
and you can move this 800 line “window” anywhere from 0 to 40,000 Hz (the
fmax of the 2120).
This is different from our earlier analyzers in that the maximum resolution of
the 2120 and 2115 is spread out over the fmax for each meter. For example, the
most resolution you could have in a 2115 over 30,000 Hz is 3200 lines of
9.375 Hz/line (30,000 Hz/3200 lines). Although you could have MasterTrend
collect a spectrum from 1800 to 2000 Hz to try to zoom in on this region, the
total resolution is spread from 0 Hz to 2000 Hz. In this case with a 3200 line
spectrum the resolution you would have in this 200 Hz range would be 0.625
Hz/line (2000 Hz/3200 lines) instead of 0.0625 Hz/line (200 Hz/3200 lines).
The zoom feature in the 2120 provides a moveable 800 line window. You can
set the frequency range to anything you want. For example, you could monitor
a gear frequency at 15,250 Hz by setting the zoom range from 15,150 to
15,350. This would give you 800 lines in this range or 0.25 Hz/line (200
Hz/800 lines). In the 2115 you could get 4.8 Hz/line (15350 Hz/ 3200 lines).
Although you cannot collect zoom spectra in a regular route, you can save
zoom spectra in analyze mode and dump it into MasterTrend with regular route
data.
Here is a step by step process on setting up zoom in the 2120:

À Press the analyze key on the 2120
Á Press 7 to go into zoom analysis
Â Set the low cutoff below the frequency of interest
Ã Set the bandwidth to a frequency above what you are analyzing (this
sets the frequency range).
Ä You may choose between hanning and uniform windows. Hanning is
what is normally used.
Å Set the number of averages for the spectrum. Four to six is standard.
Æ You may use a tach trigger or no trigger.
Ç To save your zoom spectrum to a route measurement point, press enter
when prompted and then press 4 for store data. This will allow you to
save the spectrum, the waveform or both. Note that you must have
collected a normal route measurement for this feature to store your
zoom spectrum.
Data Averaging
SST - Slow Speed Technology

The SST feature improves the quality of the very low-frequency vibration data
generally encountered in slow turning machines. We will consider machines
running between 300 and 600 RPM as slow speed. SST will make accurate
measurements down to 0.1 Hz (6 RPM). A few important measurement
considerations must be observed.
Use a low-frequency, low-noise, high-sensitivity accelerometer to collect data.

Integrate the data from acceleration to velocity units using ANALOG
integration.
Apply the SST correction feature to the measurement point. The SST feature
corrects for the deterministic error occurring with the use of the analog
integrator. The end result is the ability to see the low-frequency events at
higher measured amplitudes, allowing for easier detection.
Data Averaging
Accelerometer Selection
To obtain the useful information required to perform analysis on slow speed
equipment, a low-frequency, low-noise accelerometer will provide the results.
The sensor should be minimally responsive to temperature measurement and
should have a sensitivity of at least 500 mV/g.
Most accelerometers have a dynamic range of 100 to 120 dB. This means that
the analyzer will have the limiting dynamic range. 2120 has a range of 90 dB.
If possible when choosing an accelerometer, a ceramic piezoelectric crystal is

preferable to a quartz crystal and a shear mode accelerometer is preferred to a
compression mode accelerometer.
When comparing displacement, velocity and acceleration, it is evident that

displaying the data in units of displacement enhances the low-frequency data
and acceleration depresses the low-frequency data. However, a drawback is
that a displacement probe must be permanently and securely mounted so the
portability factor is lost.
Placing a displacement probe on each measurement point also increases the

equipment cost for your program. Integrating the data to velocity may be the
best compromise.
When integrating data, it is best to use ANALOG integration. As mentioned

in the previous section, there are a number of reasons for this choice.
Ô Analog integration attenuates the vibration signal above the Fmax of

the spectrum and thus improves the dynamic range of the analyzer
in the lower frequency region.
Õ Analog integration reduces the low-frequency flare-up known as

ski slope, which digital integration can actually increase.
Ö Analog integration produces a known effect (deterministic error)

on the vibration data that the Model 2120 Machinery Analyzer can
correct with the SST (Slow Speed Technology) feature.
Data Averaging
Practical Considerations
< Handheld mounts are unacceptable, because you cannot hold the
sensor steady enough.
< From a cost perspective, you recover the money spent for a better
accelerometer through the hours of time saved during data
collection. Using a marginal accelerometer requires extended
averaging to improve the repeatability of the data. This approach
always yields measurements that take more time and incur more
errors.
< Consider the vibration environment of the machine and its

supporting structure. Low-frequency vibrations are not attenuated
by structures and cannot be attenuated by any practical scheme of
mechanical shock absorbers or dampeners. If a machine has an
internal vibration of 5 mils P-P with a structural vibration of 10 mils
P-P in the same low-frequency band, you cannot conduct vibration
analysis.
< In general, special equipment and procedures required for

monitoring extremely low-frequencies do not intermix easily with
regular PDM data collection. Successful programs handle low-
frequency monitoring in separate routes with analysis parameter sets
(APS) tailored for each measurement point. Tailoring APS
minimizes the collection time for each point without corrupting the
integrity of the data.
Data Averaging
2120 Setup
This feature is accessible from the 2120 Acquire Spectrum Menu. Be sure and
follow the integration rules discussed earlier. Press the Analyze key at the top
of the 2120.
Press the Page key to examine the setup for the other three pages.
Data Averaging
Generally speaking the Demodulation and PeakVue Methods should not be

used at the same time as the SST feature, so these can be left with the NO
response.
Data Averaging
MasterTrend Setup
Again, this technique can also be selected from MasterTrend. Let’s examine
the analysis parameter set information needed for this application.
Data Averaging
We have a point setup with standard parameters. Remember, at slow speeds

(less than 600 RPM) the collection times will get long. You may consider
lowering the LOR to 200. This should give an adequate bandwidth.
Under Signal Processing Parameters, select the SST Control feature. For most
applications, the preprocessor should not be selected when the SST feature is
used.
Data Averaging
You may want to collect a special time waveform in G’s acceleration. The
SST feature uses analog integration which will put the waveform in velocity.
We also have a few measurement point setup considerations to look at. The
Units Type Code should be Vel <------ Acc. We will use analog integration.
Data Averaging
The second tab of measurement point setup will have the Sensor Sensitivity
set to 0.5 V/EU or 500 millivolts per G. This is the standard slow speed
accelerometer. Normal 0.1 V/G accelerometers should not be used. We will
set the Lowest Valid Signal Level to .00001. This will take care of the
expected low G level signals. Set the Highest Valid Signal to 10. We should
not see a signal larger than this.
Data Averaging
Low-Frequency Vibration Collection Lab

Ô Add a machine to your data base called SLOWONE.
Õ Add one measurement point. Remember we want to integrate the

spectrum from acceleration to velocity. The sensor should be a 0.5 volts
per g sensor.
Ö Create a parameter set to collect data using the SST method.
× Create a route using analog integration containing this machine.
Ø Download this route and collect data.
Summary
The information in this section provides not only a description of some of the
choices available under the Analyze function key, but also some recommended
selections for data collection. The features found in the Analyze function are
very powerful analysis tools. Being able to use these tools will help solve
many difficult and challenging vibration analysis problems. In later chapters,
we will utilize some of the functions to aid in troubleshooting and analyzing
various faults.
Section Nine
Digital Signal Processing
Objectives
< Relate time waveform length and frequency bandwidth to sampling rate
and sample size.
< Choose the correct analysis window for each vibration analysis
opportunity.
< Recognize limitations of digital signal processing.
< Demonstrate the ability to use the 2120 analyzer and the Analyze
function to acquire data utilizing Demodulation and PeakVue®.
Fast Fourier Transform

The conversion of time domain information to frequency domain information
is the Fast Fourier Transform (FFT).
Often a frequency spectrum is referred to as an FFT. However, the FFT is the

mathematical conversion from the time domain to the frequency domain.
Since the signal that comes into the analyzer is an analog signal as discussed in
the previous section, it must be digitally sampled by the analyzer. Therefore,
the process used by digital analyzers is actually a variation of the FFT, called
the Discrete Fourier Transform (DFT).
For the DFT, the time waveform is recreated in the analyzer by digitally
sampling, and then it is transformed into the frequency domain.
The FFT process works based on the assumption that the signal measured and
digitally sampled is one period of a periodic signal that extends to minus
infinity and to plus infinity. Normally, this is true for most vibrating pieces of
equipment.
It is the digital sampling process that makes the signal processing more
complicated. The information here unlocks the mysteries of digital signal
processing without getting bogged down in too much theory.
Time
In order to understand the FFT digital sampling process, you must understand
the relationship between lines of resolution (LOR), maximum frequency (Fmax),
length of time waveform (Tmax), the digital sample size, aliasing, windowing,
filters, and unit conversion.
Resolution
Once data has been converted to the frequency domain from the time domain,
view all the frequencies of interest in as much detail as possible. Resolution is
the number of parts of the spectrum, usually called lines of resolution (LOR).
The number of LOR selected are divided into the maximum analysis frequency
(Fmax) to arrive at the bandwidth (BW).
BW = Fmax/LOR
The lines are actually the center frequencies of bins of energy. Each bin
contains an infinite number of frequencies, and all the energy in the bin is
summed and represented by a single amplitude at the center frequency
identified at each line of resolution.
First, identify the frequencies of interest so that enough resolution is chosen to

separate closely spaced frequencies. A commonly used LOR is 800 lines.
Also, be aware that more LOR affect the length of the time waveform. For
normal trending, we have to weigh the pros and cons of higher resolution.
Remember that the time to collect one average is equal to 1 divided by the
bandwidth. As the bandwidth decreases, the data collection time increases.
The bandwidth (BW) should be no greater than 5 HZ/Line.
This will give adequate resolution for identifying trend changes and reasonable
data collection time. The time to collect data is dependent on the Fmax and
LOR settings as well as the Number of Averages and Data Overlap Percent.
The Fmax setting is determined by the equipment being monitored as we

covered in Basic Vibration Analysis. One popular way of setting Fmax is to
use an order-based set based on turning speed of the shaft being monitored.
Let’s take a look at the effect of RPM on the BW with a typical 70 x Turn
Speed Rolling Element Bearing Set.
RPM x 70 = Fmax ÷ LOR = BW (HZ)
20 HZ (1200 CPM) 1,400 800 1.75

30 HZ (1800 CPM) 2,100 800 2.65
60 HZ (3600 CPM) 4,200 800 5.3
Since Fmax ÷ LOR = BW and Time = 1 ÷ BW, we can calculate the time to
collect data above with a typical 6 averages:
BW (1 ÷ BW) = Timeone sample x # of Averages = Timetotal
1.75 0.57 sec. 6 3.42

2.65 0.37 sec. 6 2.22
5.3 0.19 sec. 6 1.13
Actually, it is not quite this easy. We can increase our data quality with a
Hanning Window and decrease our collection time with a technique called
Overlap Averaging.
Sampling Time = [( 1 - % of Overlap ) ( # Lines / F-max ) ( # of Averages )]
Case History- Spectral Resolution on I.D. Fan
2500 HP, 1175 RPM Motor
Fluid Coupling
Direct Driven Centerhung Fan
VA2CHSTY.DAT BAL-IMBALANCE, RESOLUTION-ID FAN

BALANCE/ALIGNMENT
* 2500 HP, 1175 rpm Motor

* has a fluid coupling
* fan wheel 16 ft. in dia. and it has sleeve bearings .
* has a history of problems, most of the vibration on the FIH point
* fan moves 1200 degree air in a large kiln in cement plant
The peak at 1X RPM is high with 2X also high.
The fan was subject to buildup and abrasive wear, and imbalance was cited as
the most likely cause with looseness secondary.
All the bolts were checked for correct torque and base inspected with no
abnormalities noted. The maintenance manager made plans for the Vibration
Analyst to return that evening to balance the fan.
The analyst set up the analyzer with the downloadable balance program to take
a reference run. While monitoring the peak-phase data, he noticed the phase
was erratic.
The amplitude data was swinging from a low of 0.02 in/sec to as high as 0.4
in/sec! There are many different causes of peak-phase instabilities, of which
closely spaced frequencies is one. This is known as a beat frequency. The
closely spaced frequencies add to each other as the phase matches, and subtract
as the phases separate.
The analyst decided to capture data with better resolution. The results are
shown. The problem becomes clearer as we zoom in around 1X RPM to show
two distinct frequencies.
The fan vibration was 0.5 mils Pk-Pk. The motor vibration was 1.6 mils Pk-Pk.
.The fan did not need to be balanced. The high motor vibration on the fan
bearings pointed to the fluid drive as the major contributor to high vibration.
An alignment and fluid level change produced the following result.
Recognizing a beat frequency and utilizing the analyzer to troubleshoot and

conduct on-site analysis allowed the maintenance personnel to quickly correct
a problem. Vibration analysts have wasted many hours trying to balance
equipment with beat frequencies with poor results. Bandwidth consideration is
essential to quality route data collection as well as analysis/troubleshooting.
Time Record Length

Calculate the time record length of the time waveform, Tmax, from the
following basic relationship.
Tmax = 1 / BW
or
Tmax = LOR/Fmax
At face value, this is a simple and often used equation.
However, to understand the limitations of some analyzers, it is important to

more fully investigate the relationship between the Fmax, the LOR, and the Tmax.
Tmax = Sample Size / Sample Rate
Sample Size = 2.56 * Lines of Resolution
Sample Rate = 2.56 * Fmax
These terms have already been defined, but be aware that some analyzers have
an upper limit on the sample size.
In the 2120, the limit is 4,096.
* a 400-line spectrum would require 2.56 * 400 = 1,024 samples
* an 800-line spectrum would require 2.56 * 800 = 2,048 samples
* a 1600-line spectrum would require 2.56 * 1600= 4096 samples
* a 3200-line spectrum would require 2.56 * 3200= 8,192 samples.
However, if the analyzer is limited to 4,096 samples, then the 3200-line

spectrum will be created from 4,096 samples since it is the upper limit of
the analyzer.
This is important when discussing the Tmax in the time waveform, because:
C in general, raising the Fmax decreases Tmax
C and raising LOR increases Tmax to the point that the multiple of
2.56 * LOR reaches the sample limit in the analyzer. In this
case, the sample size for anything greater than 1600 lines is
forced to be 4,096.
The table on the next page demonstrates how this limitation affects the Tmax for
a maximum 4,096 sample size analyzer.
Tmax = Sample Size / Sample Rate
Fmax Sample Rate (Sr) = LOR Sample Size (Ss) = Time (sec.) =
Fmax*2.56 LOR * 2.56 Ss ÷ Sr
400 1024 400 1024 1.00
400 1024 800 2,048 2.00
400 1024 1600 4,096 4.00
400 1024 3200 8,192 8.00
4,096 stored by MT 4.0 stored by
MT
The last entry in the table may seem incorrect, but remember 4,096 is the
maximum sample size, and anything that would be greater than 4,096 is forced
to be 4,096 for storage in MasterTrend.
This is the reason your time waveform is not affected when following the
raise-the-LOR-to-lengthen-the-time-waveform rule. You must be aware of the
upper limit of the sample size and the number of LOR to which this number
corresponds.
To increase the amount of time in the time record, it will become

necessary to adjust the Fmax to a lower value.
The following chart shows the effect on the time record of various Fmax
settings.
Fmax Sample Rate LOR Sample Size (Ss) = Time (sec.)=

(Hz) (Sr) = Fmax * 2.56 LOR * 2.56 Ss ÷ Sr
1000
400
200
100
101
10
1
Low Frequency Limit
2
Maximum Stored by MT
3
160 Stored by MT
2120 Analyzer Lab - Setup-Time Waveform

Example: While collecting data on a 2-pole (synch. speed 3600 RPM) electric
motor, the live time spectrum display shows a peak at 2X RPM that changes
significantly in amplitude from average to average. In addition, an audible
humming fluctuation is noted. Realizing this may be a beat frequency, you
decide to view the time waveform for confirmation. With a typical analysis
parameter setup of 70.5 X RPM and 800 lines of resolution, the time record is
calculated as follows:
Fmax * 2.56 = Sample Rate; LOR * 2.56 = Sample Size;
Time = Sample Size ÷ Sample Rate or Time = LOR ÷ Fmax
Fmax = 70.5 x 60 Hz (3600 RPM); Fmax = 4500 Hz
LOR = 800
Time = 800 ÷ 4500 (2120 rounds to 5000 Hz)
Time = 400 ÷ 5000; Time = .16 seconds
Obviously .16 sec is not enough time to see a slow beat frequency, so the
Analyze function is very useful here. Let’s say you decide that 4 to 5 seconds
of data would be appropriate for this analysis. To obtain that data you can
adjust Fmax and LOR to meet your needs. A calculator will be helpful. A
number of calculations will work — here is one:
Press the Analyze button and choose ACQUIRE SPECTRUM.
The FREQUENCY: 200 HZ (Fmax) and LINES: 800 (LOR) will give us the
desired 4 seconds of time record (400 ÷ 100 = 4).
This will also allow us to evaluate the spectrum around 60 and 120 Hz with
high resolution. The Bandwidth will be calculated by: 200 Hz ÷ 800 Lines =
.25 Hz/Line
The resulting spectrum is shown below. Press F1 for the time waveform.
Press the F1 (SPLIT SCREEN) key.
Notice the spectrum data (bottom) taken with an accelerometer has been
integrated to units of velocity. The time waveform remains in units of g’s
acceleration because of the integration method chosen in the acquire
spectrum. The choice of UNITS/INTG: VELOC/DIGITL will collect the
data as shown.
Now collect the data by changing the integration mode to ANALOG and view
the results.
Notice the waveform data is in IN/SEC Velocity. Integration takes place in

the time domain and therefore is integrated and stored in the integrated
(velocity) units.
When collecting a spectrum/waveform with the Analyze function of the 2120

for bearing and gear defect analysis, digital integration will be the preferred
method.
This will store the waveform in acceleration units. Because of the relatively
high frequencies generated, acceleration units makes it easier to determine
impact severity as bearing and gear flaws develop.
Velocity has the effect of smoothing out impacting in the time waveform. In
MasterTrend, we will use analog integration, but collect a special time
waveform in units of g’s acceleration.
Press the Enter key.
The ANALYZER FUNCTIONS screen appears.
Select 4) STORE DATA press Enter.
You have the storage option of spectrum, waveform or both.
Choose 3) STORE BOTH.
This data will be attached to the current point as an Analyze data.
Overlap Averaging
Overlap averaging reuses data stored in the input time buffer of the analyzer.
When using a Hanning window on our vibration data, most of the data has
been multiplied by a factor of one at the center of the window to zero at the
beginning and the end of the time window.
In order to fully use all the data acquired in the analyzer, use overlap
averaging. Fifty percent to 67 percent is recommended. Sixty-seven percent is
the default
The default setting of 67 percent overlap averaging reuses 67 percent of the

time data, saving acquisition time and generally improving amplitude
accuracy. In the 2120 the setup is done with the utility function.
Aliasing
The term aliasing brings to mind the idea of someone masquerading as
someone else. Criminals often do this to elude law enforcement agencies. In
the world of vibration analysis, an aliased frequency is one that appears as
something it is not. Fortunately, it is not trying to hide from us but was forced
into existence by an incorrect digital sampling process.
In a digital analyzer, the time domain data is digitally sampled at the sampling
rate. We know that the sampling rate is controlled by the maximum analysis
frequency ( Fmax ).
In order to get good data, the sampling rate must be set higher than two times
the maximum frequency of interest, Fmax. Two times Fmax is known as the
Nyquist frequency. A very common sampling rate is set at 2.56 times the
Fmax.
An aliased frequency occurs when frequencies higher than the Fmax are present
in the signal. The sampling rate under-samples this high frequency and creates
a low (aliased) frequency equal to the high frequency minus sampling
frequency.
For example:
If the Fmax is 1,000 Hz, then the sampling frequency will be:
1,000 * 2.56 = 2,560.
If a frequency is in the data at 2,768 Hz, then an aliased frequency can be

expected in the data at 2,768 to 2,560 = 208 Hz.
To avoid aliasing, digital analyzers use anti-aliasing filters that remove all
frequencies above 40% of the sampling rate before the time data is converted
to frequency data.
Therefore, with most digital analyzers, aliasing is no longer a concern.
Leakage
Leakage occurs when complete cycles of the vibration signal in the time
waveform are not captured in the time record.
Instead of a discrete frequency in the spectrum, energy spills into adjacent

bands or lines of resolution. A rectangular window can allow time data to be
captured that cannot be transformed into the frequency domain. If a non-
integer multiple of the periods of vibration appears in the time waveform for
all the frequencies present, the frequencies smear over many lines of
resolution.
Digital analyzers use windowing functions to avoid smearing. Some typical

window choices are:
Uniform - which is really no window at all
Hanning - (often called a cosine taper)
Flat Top, Kaiser Bessel, Force, and Exponential.
The purpose of windowing data is to eliminate leakage problems. For most

vibration analysis applications, the window of choice is the Hanning
window. The Hanning window does a great job of forcing the beginning and
the end of the time record to a zero amplitude. This makes the reconstructed
time waveform continuous with no amplitude variations.
A sinewave acquired with a uniform window that has an integer number of

periods in the waveform transforms into the frequency spectrum with all the
energy contained in one spectral line of resolution. However, this condition is
difficult to achieve.
If the time waveform contains a non-integer multiple of periods, then the

frequency data smears over many lines of resolution.
The most typical case involves using a Hanning window on a non-integer

multiple of periods in the time waveform. In this case, the data transformed to
the frequency domain is contained primarily in one line of resolution with
some frequency data smeared into the adjacent bands of the primary frequency.
This, however, tends to be the most acceptable method of data collection.
The Hanning window still spreads a pure discrete tone over three lines of
resolution. To separate very closely spaced frequencies, increase the lines of
resolution a little more.
Integration and Differentiation

The vibration data that is the input signal into the analyzer is a time-varying
voltage proportional to the vibration measured by the transducer. In other
words, an accelerometer produces a voltage that varies over time relative to the
acceleration measured by the transducer. The voltage amplitude in the time
waveform is converted to the desired amplitude units based on the sensitivity
and conversion factor of the transducer.
Most analyzers have the ability to convert from the measurement units of the
transducer to either of the other two units in the time domain or the frequency
domain.
At CSI, integration of the time signal is called analog integration and

integration of the frequency domain is called digital integration.
Integration is a process of converting from acceleration to velocity or

displacement, or converting from velocity to displacement.
Differentiation is the process of converting from displacement to velocity or

acceleration, or converting from velocity to acceleration.
How are these unit types related, mathematically?
D (Displacement) = distance traveled by vibrating object

V (Velocity) = change in Displacement/change in Time
A (Acceleration) = change in Velocity/change in Time
These terms are often represented with the following equations:
D = X
V = X/T
A = X/T/T = V/T
Therefore, if any one of these terms has been measured, integration and
differentiation allows any of the other terms to be calculated, provided the
analyzer or software used is capable of this conversion process.
One drawback to integration is a flare-up of the lower frequency data caused

by the integration process. This effect is often called integration noise or a
ski-slope effect.
This is typically only noticeable when integrating from acceleration to

displacement and tends to affect only the lower one percent of the frequencies.
This may cause the overall vibration level to be higher than usual, and it may
be included in the calculation of the overall vibration level.
Normally, this is not of any real concern unless the analyst wants frequency
analysis below 2 to 4 Hz. If this is the case, the use of SST may be beneficial.
Filters
The vibration industry generally uses three types of filters:
< Low Pass

< Band Pass
< High Pass
Each one filters data out of a signal, which you may find useful when
analyzing signals with large dynamic ranges.
For example, some spectra have both large and small amplitudes relative to
each other. Because of the dynamic range of the analyzer, however, you
cannot analyze the low amplitude vibration in the same plot as the high
amplitude vibration.
A filter helps resolve the problem.
Low-Pass Filter
As shown above, the low-pass filter removes data above the selected
frequency. Use this filter during low-speed (#600 RPM) frequency analysis.
Data collection times can exceed 2 to 10 minutes per average. Therefore, you
can increase data collection times with a low-pass filter by retaining only those
events that take place in the selected frequency range.
Band-Pass Filter
There are two types of bandwidth filters 6 Constant Percentage Bandwidth

and Constant Bandwidth. These filters primarily serve the same function. The
Constant Percentage Bandwidth filter changes width depending on the selected
frequency.
Notice the difference between the two types of filters in the example above.
As shown, the filter to choose is the Constant Bandwidth Filter, because it
provides the best resolution between both high- and low-frequency
components.
High-Pass Filter
The high-pass filter gives you the ability to filter out low-frequency
components for detailed analysis as illustrated above. This proves useful when
low-frequency, high-amplitude data swamps the high-frequency, low-
amplitude data you want to see. This situation often occurs when high-
frequency events appear in the same plot as Run Speed and its relative
harmonics.
Frequency Demodulation
Several publications have described the application of demodulation to
vibration measurements for machinery defect analysis. As an application tool,
demodulation proves helpful in a wide variety of applications.
Demodulation primarily increases the effective dynamic range of the analyzer

for certain types of low-level measurements. This increased range enhances
defect indicators for fault analysis.
The 2120 DEMODULATION feature has the following advantages:
• Provides early detection of bearing and gear faults which might be

masked by high amplitude, relatively lower frequency events.
• Lets you filter input signals for specific analysis requirements.
• Provides detection of faults which normal averaging techniques cannot

detect.
Demodulation Process
Analog Input Signal (Waveform)
500 Hz, 1, 2, and 5KHz Hp

100 Hz HP, 50 & 60 Hz BP
Autoranging
Envelope Demodulator Stage

Envelop
Full Wave Rectification

Modulated Time Waveform
400 Hz
Signal Output to the Analyzer
2120 DEMODULATE Lab
While collecting route data, it is common to have high amplitude frequencies

dominate the spectrum. Often frequencies generated by unbalance,
misalignment, vane pass and other relatively low frequencies “mask” the lower
amplitude, high frequencies generated by rolling element bearing defects.
In gearboxes with meshing gear teeth, the Gear Mesh Frequency (GMF) has a
tendency to mask the lower amplitude sidebands that are generated by various
failure modes.
Even with the >90 dB dynamic range of the 2120, the bearing defects
frequencies will may be lost in the noise floor.
The 2120 has a built in demodulator with a 1000 Hz default high pass filter.
When you choose DEMODULATE, the 2120 will automatically default to a

PREFILTER setting of 1000 Hz.
If you choose the FREQUENCY (Fmax) greater than 1000 Hz the 2120 will
automatically set the PREFILTER to the next highest filter setting above the
FREQUENCY setting.
The 2120 has preset high pass filter (FREQUENCY) settings of:
20, 50, 100, 200, 400, 500, 1000, 2000, and 5000 Hz.
Another tool to help in the analysis of bearings and gears is to use the
Demodulation pre-processing.
With the ANALYZE function, we will introduce the operation of this function
in this chapter, and give specific setup considerations in later chapters as
appropriate.
First we will look at a simple application of demodulation. We will start by

collecting a normal spectrum from a variable speed rotor kit turning at
approximately 3600 RPM (60 Hz).
Normal Spectrum-No Demodulation
Displayed above is a Normal Spectrum - with no Demodulation.
The results are shown above. Notice in this example we have a peak at 820 Hz
surrounded by many lower amplitude peaks. This is most likely an electrical
forcing function related to the drive/motor.
This peak has the appearance of being modulated by another frequency

because of the possible sidebands around it. We will now use demodulation to
get rid of the high amplitude lower frequencies, as well as the carrier
frequency (820 Hz in this example) and show the lower frequency modulator
(with frequency equal to spacing of sidebands).
SETUP FOR DEMODULATION:
On PAGE 1 OF 3 the FREQUENCY chosen will be the Fmax of the spectrum

you want to view. We will use 400 Hz. The LOW CUTOFF and LINES will
be chosen as appropriate. Set: HANNING WINDOW, 4 AVERAGES, and
NO SST CONTROL
Set the UNITS/INTG: ACCEL/ANALOG. This puts the spectrum in g’s

acceleration which is important for proper demodulation operation. The analog
integration mode will extend the dynamic range to the maximum. Press the
PAGE UP key.
Page 2 will not be used.
Press the PAGE UP key to advance to PAGE 3 OF 3.
Toggle DEMODULATE: YES. You will see the PREFILTER: 500 HZ HP

automatically chosen. This filter will default to a value based on your
FREQUENCY setting on the 400 Hz Fmax setting on page 1.
Press enter to collect data.
Normal Spectrum 0-400 Hz
Demodulated Spectrum
Note: Both the Normal Spectrum and Demodulated Spectrum was collected
with a 400 Hz Fmax.
The normal spectrum (top) has a 1xTS peak (59.38 hz) with low levels of
harmonics and not much else.
The demodulated spectrum has filtered off the high amplitude, low frequencies
(1x and harmonics) and now shows us the possible modulator of 120 Hz. The
120 Hz did not show up in the original spectrum because the amplitude was
much lower than 1x, 2x etc.
We will show how this feature is very valuable in bearing defect detection in
the Rolling Element Bearing section.
MasterTrend Setup - Demodulation

Demodulation is a very valuable processing technique for route trending.
Demodulation can help in difficult vibration environments such as slow speeds
or where there is a large interfering signal (1x, 2x, GMF etc.). Here is a basic
MasterTrend setup. The HP filter choices and Fmax settings depend on the
particular vibration environment.
Note: The Signal Processing Parms tab is where Demodulation is selected.
You will also need to collect a special time waveform because the waveform is
not usable with demodulation. Select the third tab, Waveform Parameters.
An order based set with 80 orders and 1024 points will store a time block
which will include about 5 shaft rotations which is ideal for most rolling
element bearing analysis.
A measurement point setup is shown above. The Units Type Code should be
setup for no integration (ACC:<---- ACC).
Filter Stage
The Filter Stage contains eight selections:
C five high-pass filters, two band-pass filters, and a no-filter

bypass option.
The choice of different filters lets you reject large signal components outside
the frequency band you specify. Therefore, large components outside the
specified frequency range do not overload the Model 2120 Machinery
Analyzer, which lets it allocate more of its dynamic range to the band of
interest.
The proper selection of a filter requires some knowledge of the frequencies

you expect for the desired signals. For example, you know the desired carrier
signal falls in the frequency range between 2,500 and 4,000 Hz (e.g., an early
bearing signature). You also have a large, unwanted signal between 400 and
500 Hz (e.g., a gearmesh). For an appropriate filter setting, choose either the
1,000 or 2,000 Hz high-pass filter.
Now change the signal frequency ranges to between 2,000 and 4,000 Hz for
the desired range and between 400 and 900 Hz as the unwanted range. The
filter setting choice now becomes less clear. The 1,000-Hz filter provides only
a small reduction of the 900-Hz signal. The 2,000-Hz filter reduces the
desired 2,000-Hz signal somewhat. In this case, try both filters and select the
one that gives you the best data.
If you undertake gearmesh demodulation, choose a high-pass filter that passes

the gearmesh frequency and attenuates frequencies below gearmesh. If you do
not know in advance the frequencies of the signals, select the bypass mode in
the Model 2115/2120 Machinery Analyzer. Examine all available signals to
determine filter selection.
Note: You will need to play with some of the filter-settings on your own and
examine the data to see what filters will obtain the data you need to
analyze a specific problem.
Band Pass Filters
Use band pass filters primarily to examine electrical line frequencies. These
filters reject all signals that are not within a band around the line frequency.
You can also use band pass filters to detect the running speed of an induction
motor (four poles or more). Simply set the instrument to demodulate the line
current.
Demodulator Stage
The demodulator stage accepts as a portion of its input a signal known as the
carrier.
The output of the demodulator is a signal known as the baseband signal (data
relative to the envelope). The baseband signal contains the information you
want.
Typical carriers in machinery applications include:
C gearmesh frequencies
C structural resonances
C bearing resonances
C electrical line current
C in some cases, the resonance of the sensor itself
The last case (sensor resonance) most often occurs when using the handheld
probe. You must also have a mechanism in the machine that causes the
information you want to be impressed upon or modulate the carrier.
Fortunately, this latter condition is very common in machinery.
Low Pass Filter Stage

The final portion of the demodulation process is a low-pass filter.
The Fmax setting effectively chooses this filter. This filter removes the residual
carrier component that otherwise dominates the baseband output. This filter
works in conjunction with the PREFILTER setting chosen on page 3 of
acquire spectrum.
PeakVue
PeakVue stands for peak value and can be used to detect short duration higher
frequency waves called stress waves.
Stress waves are created when metal is bent. For example, if a rolling element
in a bearing passes over a defect which may still be below the surface of the
inner or outer race, the race will deflect and then spring back.
This motion will create a sine wave. PeakVue measures the highest amplitude
found in this waveform and holds it as the highest value during a waveform
time length equal to 1 over the sampling rate.
PeakVue is a process that also passes the waveform data through a high pass
filter to remove the rotational vibration frequencies from the data. However,
filtering is its only real similarity to demodulation. The high pass filter should
be set above the conventional Fmax of the spectrum.
Generally, the 1000-Hz high pass filter is a good choice. The PeakVue
process alters the length of short-duration, very high-frequency events
occurring in a machine’s vibration data. Be sure to set up a measurement point
for this analysis method that does not integrate the data.
Data taken without PeakVue enabled.
Data taken with PeakVue enabled.
2120 Setup
As with demodulation, PeakVue can be accessed from the 2120 Acquire

Spectrum Menu.
Toggle to the third page of the setup menu.
Collection of at least one data point per machine component is recommended

using the PeakVue technology for early detection of bearing and gear defects
as well as other machine problems.
MasterTrend Setup
This can also be activated from a MasterTrend Analysis Parameter Set

definition. In this case the Fmax is set to 65 orders of turning speed.
If this parameter set is used on a machine turning 1800 RPM.
The Fmax will be 1800 CPM * 65 orders = 117,000 CPM.
If we divide by 60 to get the Fmax in hertz, the Fmax is 1950 Hz.
The PeakVue Filter should be set above the Fmax, so we will choose a
PeakVue filter of 2000 Hz.
Summary
These two techniques are very powerful early detection methods.
In this section, we have examined the differences between Demodulation and

PeakVue as well as methods of enabling these features from both the 2120
Machinery Analyzer and MasterTrend ( use of defined AP-Sets ).
Demodulation and PeakVue Lab
Ô Create a database: INTCLAS.DAT.
Õ Create a station: Signal Processing.
Ö Create a machine: Condensate Pump #1.
× Build five motor and five pump measurement points.
Ø Create Analysis Parameter Set Number 20. Enable the

demodulation of data above 1000 Hz and make sure this set is
assigned to the Motor Vertical points of the Condensate Pump #1.
Ù Create Analysis Parameter Set Number 21. Enable the PeakVue

feature to process data above 1000 Hz and make sure this set is
assigned to the last pump point on the Condensate Pump #1.
Ú Create a route: Pump #1.
Û Load the route and collect your vibration data.
Ü Do not dump the data back into MasterTrend yet.
Summary
The key to collecting worthwhile data is understanding how the analyzer
performs its job.
When setting up for data collection, remember that the:
C spectral Fmax
C the lines of resolution
C the waveform Tmax
C and the waveform size are all related
Don’t forget to choose the correct analysis window for each measurement
application. The conversion from analog data to digital data can alter the way
the data appears, so remember to recognize the limitations of the measurement
process. Using the information in this section should help unlock the
mysteries of digital signal processing.
Section Ten
Waveform Characteristics
Objectives
< Describe five waveform characteristics.
< Identify waveform symmetry using an APD.
< Discuss waveform modulation and how it translates to the frequency

domain.
General Description
Waveform data has been an area of neglect since many analysts have not had
the fundamental training to properly use the waveform for diagnostic purposes.
This section discusses some of the fundamentals of time domain data. The
time domain (waveform) data can be very valuable to the analyst, especially
with complex machines.
A number of different displays use the time domain. Displays such as

synchronous time averaged data is averaged in the time domain. APD
(Amplitude Problability Distribution) this is a function of Wavepak, displays
the symmetry and skewness of the waveform signal.
Orbits use two waveforms from transducers or eddy current probes mounted
90E apart and show the shaft rotational characteristics. These displays and
others will be discussed in detail later in this section.
Each defect type has a characteristic waveform, which subsequently translates

to the frequency domain.
However, there are characteristics and specific events that do not translate to
the frequency domain as discrete peaks. In order to truly understand this
limitation, the analyst must first understand how the time domain data is
gathered and transformed into a spectrum through the Fast Fourier Transform
(FFT) process.
What Is A Waveform?
Time domain data, raw transducer output, signal voltage and many other terms
refer to waveforms. The difference relates to the input transducer type, the
axes being considered (magnitude vs. time), and whether or not the signal has
been manipulated with a digital signal analyzer. Historically, waveform data
has been gathered using analog systems.
Waveform or time domain data is comprised of amplitude with respect to

time. Signals with an amplitude, whether vibration, current, voltage changes,
or other signal types, change with time. The time change may be slow (1 to 10
minutes) or extremely fast (1 to 200 msec). In either case, time is a primary
aspect of understanding waveform.
There are certain things to look for when conducting waveform analysis. In
order to know what to look for, we should discuss the characteristics and what
they mean to the analyst.
BAL - C-20 FLOAT ROLL FAN

C-20 FLOAT-FIH FAN BEARING INBOARD HORIZONTAL
8
Waveform Display
6 25-APR-96 09:36
4 RMS = 1.28
LOAD = 100.0
Acceleration in G-s
2 RPM = 3550.
RPS = 59.17
0
PK(+) = 6.94
PK(-) = 5.84
-2
CRESTF= 5.40
-4
-6
-8
Time: 135.09
0 20 40 60 80 100 120 140
Ampl: .00000
Time in mSecs
As was previously stated in this section, the waveform provides specific

characteristics for defects of a single or multiple nature.
Those machines containing multiple defects are much more complex and
require additional time during the analysis procedure.
Note: The waveform is only as good as its definition. If the resolution of your
waveform lacks definition, the data can be worthless, or poor at best.
Once the characteristics have been properly identified, the analyst can rule out
certain fault types.
For example:
If a waveform is periodic (sinusoidal)
C looseness
C cracks
C resonance
C antifriction bearings
Could probably be ruled out. You may not know what the problem is, but
you know what it is not.
Listed below are waveform characteristics an analyst should look

for when analyzing the waveform:
< Amplitude < Periodic < Complexity

< Asymmetry < Spikes/Impacts < Discontinuities
< Electrical v Mechanical < Non-Periodic < Low Frequency Events
< Distortions < Modulation < Truncation/Restrictions to
Motion
C-20 - C-20 FLOAT ROLL FAN

8
Waveform Display
6 25-APR-96 09:36
4 RMS = 1.28
FAULT LOAD = 100.0
Acceleration in G-s
2 ALERT RPM = 3550.
RPS = 59.17
0
PK(+) = 6.94
PK(-) = 5.84
-2
CRESTF= 5.40
ALERT
-4 FAULT
-6
-8
Time: 76.72
0 20 40 60 80 100 120 140
Ampl: -.109
Time in mSecs
Amplitude
When diagnosing machinery faults using the time waveform, similar to
spectral data, we are concerned with the amplitude of the waveform.
When we are discussing bearing and gear waveforms, we use the peak to peak
amplitude of the waveform. This is often referred to as g swing.
The g swing is the sum of the absolute value of the maximum positive and
negative amplitude in that period.
MasterTrend calculates this value and gives us the ability to trend and alarm
based on this and other waveform values.
BAL - ZONE 6 EXHAUST

C-30 Z6X -FOH FAN BEARING OUTBOARD HORIZONTAL
0.6
Waveform Display
13-JUN-95 14:52
0.4
RMS = .1390
LOAD = 100.0
Acceleration in G-s
0.2
RPM = 1000.
RPS = 16.67
-0.0
PK(+) = .3672
PK(-) = .4322
-0.2 CRESTF= 3.11
-0.4
-0.6
0 100 200 300 400 500

Time in mSecs
VA2CHSTY.DAT, BAL-IMBALANCE, C-30 Z6X, ZONE 6 EXHAUST
Periodic
Sometimes referred to as a deterministic simple signal, this is an ideal signal

which repeats itself exactly after a fixed period.
This is not possible in the real world. However, there are some machinery
faults which have this characteristic.
A single plane balance problem will have a very periodic waveform due to the
mass rotational center and the rotor shaft of other component center line
differences.

C-20 FLOAT-FOH FAN BEARING OUTBOARD HORIZONTAL
6
Waveform Display
25-APR-96 09:37
4
RMS = 1.59
LOAD = 100.0
Acceleration in G-s
2
RPM = 3508.
RPS = 58.47
0
PK(+) = 4.88
PK(-) = 5.25
-2 CRESTF= 3.30
-4
-6
0 20 40 60 80 100 120 140

Time in mSecs
Label: LOOSE, OUT OF BALANCE
Complexity
To determine the complexity of the waveform, establish whether the signal is:
C periodic in nature
C estimate the harmonic content
C determine if the signal is synchronous
C non-synchronous
C identify whether the waveform correlates directly to the spectral
data.
Confirmation will be covered in detail later in this section.
Impacts/Spikes
Impacts or Spikes may or may not be repetitive in nature.
The non repetitive spikes generate white noise. In other words, they excite all
of the frequencies present.
Repetitive impacts or spikes, such as those produced by rolling element

bearing defects or broken gear teeth, may excite discrete frequencies and
therefore show up well in the spectrum. The excitation is proportional to the
energy induced by the impact.
This characteristic is best detected by defining a waveform amplitude type in

acceleration. Acceleration data is proportional to force, and the sudden
impact induces a force which could be of grave importance.
The crest factor, which is equal to the maximum peak (positive or negative)
divided by the RMS of the waveform, is a good indicator of the impacting.
This value can be setup as an analysis parameter and trended in MasterTrend.
Discontinuities
This characteristic is usually associated with faulty equipment due to the
discontinuous nature of the data.
Data with this characteristic has breaks in the data where there appears to be a
loss of input signal or a significant increase/decrease in amplitude.
This is not a uniform change such as resonance, load changes, or even sudden
component failures.
Discontinuous data is typically unpredictable, and very distinct.
MISC - #1 H2O BOOSTER

4661 -MIV MOTOR INBOARD VERTICAL
2.5
Waveform Display
2.0 16-NOV-95 10:18
1.5 RMS = .5155

LOAD = 100.0
Acceleration in G-s
1.0
RPM = 1789.
RPS = 29.82
0.5
PK(+) = 2.24
0
PK(-) = 1.43
-0.5 CRESTF= 4.35
-1.0
-1.5
-2.0
0 60 120 180 240 300

Time in mSecs
Label: LOOSE BASE
VA2CHSTY.DAT, MISC, #1 H2O BOOSTER
Asymmetry
Asymmetry refers to the relationship between the positive and negative energy.
A waveform is asymmetric when there is more energy in the positive plane

than the negative or vice versa.
Notice above, the waveform display shows the peak positive/negative energy.
We can use this information to make some assumptions about our data.
Asymmetry refers to the direction of movement relative to the transducer

mounting with a positive signal representing energy into ( towards ) the
accelerometer and a negative signal representing away.
A tool which is designed to check this type of characteristic is the APD,

Amplitude Probability Distribution.
Also a tool that can be utilized is to select the Analyze Data feature in Diagnostics
Plotting when in Waveform Analysis. Select option from Graphics Function Bar.
Amplitude Probability Distribution

An APD or Amplitude Probability Distribution is similar to a Hystorgram.
The signal is broken down into amplitude percentages, and then the amplitude is
plotted.
The X-Axis is the amplitude and the Y-Axis is the percentage of the signal that falls
into that amplitude range.
This can be extremely helpful in finding the root cause of a problem.
The APD is typically used for acoustical analysis.
It can also be used for machine vibration analysis to find the balance of the signal
(asymmetries), the direction, and possibly the location of a specific defect especially
those that may not stand out in the waveform or the spectrum.
Sinewaves
Sinewaves are very symmetrical, which means there is a balance of energy in the
positive and negative planes.
If most of the vibration signal is evenly distributed and sinusoidal, there is a strong
possibility it is due to a synchronous component such as imbalance, misalignment,
gears, blades, etc.
The waveform and APD below show the shape of a sinewave and the probability
related to this type of signal.
Notice that the APD at the bottom of the above display shows a set of peaks at the
maximum and minimum amplitude locations. Each of the peaks represent the total
probability of the signal in that location.
This could also be called a Hysteresis display. The probability of the signal being in
the ± 10 volt location is much more probable that the signal being at the zero location
of the display.
This information lets the analyst check for the direction of motion and the skewness
of the data. Looseness and other machinery faults may be identified easier if the APD
is used with the waveform and spectrum displays.
Triangle Wave
With a triangle wave, we see the relationship of the waveform and a different type of
APD display.
The waveform takes an extended time to pass through just above the zero crossing of
the waveform about the 1.5 - 2.0 volt location. Therefore, the APD shows a high
probability that the signal is in the range of energy previously mentioned.
Also note that the data is skewed to the negative plane. Again, this provides the
analyst with the direction of motion.
The previous illustration displays the direct relationship between the waveform and
the APD. Bear in mind that the APD provides another tool to determine location,
direction, and asymmetry.
Squarewave
The squarewave below provides some insight into the use of the APD for checking
asymmetries. Remember that symmetry refers to the balance of energy. Therefore,
with a slightly more complex signal, this becomes more important especially when
performing Root Cause Failure Analysis(RCFA).
In the illustration above, the signal is asymmetric, and there is more energy in the
positive plane than the negative.
If this is a mounted transducer, such as an accelerometer, the analyst should look for
more movement in the direction of the transducer.
The energy in the positive plane shows movement toward the transducer, and the
negative plane is obviously the opposite.
Truncation/Restrictions to Motion
Truncation means to abruptly shorten, or to appear to terminate.
In waveform analysis, this characteristic indicates restrictive motion. It is usually

caused by a light rub. With enough waveform resolution and understanding of the
phase characteristics of specific amplitude types, the analyst can locate and identify
the severity of the rub. This is another application for multi-channel analysis,
specifically orbits.
Modulation
All the waveform characteristics up to this point have dealt with signals of a constant
amplitude.
A varying signal will cause the waveform to become modulated. The type of
modulation occurring determines its classification. Commonly referred to as Beat
frequencies, these may be broken into three specific categories.
Amplitude
If the amplitude of a single frequency signal changes during each cycle, it is no
longer periodic. If this signal is a single frequency and the amplitude varies
periodically at a lower, but unrelated frequency, then the spectrum has an easily
recognized pattern. The spectrum will have a peak at the signal’s frequency with one
peak on each side spaced at the frequency of the amplitude change. These peaks are
referred to as sidebands.
Amplitude modulation is common when analyzing inner race bearing defects. This
occurs when the defective bearing component passes in and out of the bearing load
zone. The middle of the load zone is typically where the highest amplitudes in the
waveform show up.
The spectrum and waveform show slot pass frequency from an AC induction motor.
The primary signal at 34xTS is marked with a vertical line. The sideband cursors
mark the amplitude change at 120 Hz.
Beating
A beat pattern in the waveform is related to amplitude modulation. The difference is
modulation caused by a beat is comprised of two unrelated single frequency signals,
closely spaced in frequency. The signal’s waveform appears like a modulated signal
at the mean frequency of the two components, amplitude modulated at a frequency
equal to their mean difference frequency.
Beating is often found in two pole induction AC motors. The close proximity of two
times line frequency and the second harmonic of turning speed cause this beat. The
two times line frequency component can be high in amplitude if the motor has a
rotor or stator defect.
AMGL - CENTAC 3 STAGE COMPRESSOR

#1 CENTAC -1BA MOTOR OUTBD AXIAL TO 200 Hz
0.04
PK Vel in In/Sec
REFERENCE SPECTRUM
0.03
06-DEC-94 10:15
0.02 OVRALL= .0581 V-DG

PK = .0380
0.01 LOAD = 100.0
RPM = 3575.
0
RPS = 59.58
60 80 100 120 140 160 180
Frequency in Hz
Acceleration in G-s
0.10 WAVEFORM DISPLAY

0.05 06-DEC-94 10:15
-0.00 RMS = .0678
-0.05 PK(+) = .1300
-0.10
PK(-) = .1790
-0.15
CRESTF= 2.64
-0.20
Freq: 119.00
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Ordr: 1.997
Time in Seconds Spec: .02520
Dfrq: 1.000
An example of beating is shown above. The 2x RPM and 2x line frequency are
separated by less than .5 Hz. The waveform shows the amplitude modulation
associated with beating.
Frequency
Rarely seen in a routine environment, this is a change in frequency without a change
in the signal amplitude. Frequency modulation typically occurs in gearmeshing
vibration, due to the small speed fluctuations caused by tooth spacing errors and
faults as they develop. Almost all effects which cause frequency modulation in
gearboxes and other machines will also produce amplitude modulation. It may be
difficult to separate out the two effects, but a very wide spread of sidebands in the
spectrum is usually an indication that significant frequency modulation is present.
The example above is from a winder gearbox. The vertical line in the spectrum
marks gearmesh frequency at 24xTS. The sideband cursors mark the output shaft
speed with labels identifying sidebands spaced at input shaft speed. The waveform
has been expanded to show the frequency modulation occurring. A good illustration
is shown between 170 and 180 msecs. As stated above, amplitude modulation can
also be seen.
As seen from the examples showing modulation, the signal is usually made up of a
combination of amplitude and frequency modulation. One thing is for sure, if the
spectrum contains sidebands, the waveform will be modulated.
Now that modulation has been discussed, the remaining characteristics of the
waveform can be presented. These characteristics build on the ability to identify
modulation.
Low Frequency Events

When performing detailed analysis, you need to be able to collect and analyze data in
excess of one minute for low frequency problems. This is extremely important when
the machine in question has an operational speed below 200 RPM. The challenge in
identifying low frequency defects is having sufficient time in the waveform. A low
frequency event may only appear once in the collected time domain. As discussed
earlier, this event will not be transformed into the spectrum.
The waveform data above displays an event that occurs approximately every two
seconds or 0.5 Hz (30 CPM). Therefore, unless you were specifically looking for a
low frequency event with a very low frequency setup using special transducers, you
could have missed the event completely.
Electrical vs. Mechanical

Determining if the source of energy is mechanical or electrical is sometimes
difficult. Appropriately set up waveforms can be a great help. Setting up for a long
enough time to capture the operational conditions and the machine shutoff point can
identify the source. The advantage of using the time domain as opposed to the
frequency domain is there is no need to worry about the screen update time or
sampling rate.
Waveform and Spectrum Relationship

Now that we have identified the specific characteristics of the waveform, let’s
compare the waveform and its associated spectrum.
For our purposes, each spectrum has an associated waveform. The spectrum is
made of this waveform. As discussed earlier in this section, some of the
characteristics in the waveform do not translate to the FFT due to the way the
calculations are made. The assumption is that there is a repetitive cycle of
events made up of sines and cosines. However, this is not actually the case.
If an event happens only once, then this event has no frequency; therefore, the
spectral representation is a continuous spectrum. The data below shows the
waveform and spectrum of an instrumented force hammer. Though used for
multichannel impact testing, the characteristic differences are shown.
In the waveform above, there is no repetition in the event; therefore, there is no

frequency. An impact with little or no time associated with it results in the
spectrum above. The analyst input approximately 80 pounds of force, yet there
is only 0.1 pounds of force at all frequencies ranging from 0 to 200 Hz with
little deviation in the actual amplitude. This may cause the spectrum to look as
if the problem has fixed itself. However, the data in the waveform is extremely
severe. The waveform amplitude and spectral characteristics don't match
because the defect may have become so nonsynchronous and nonrepetitive that
there is no definable frequency content.
Modulated Waveforms
Finally, when modulation is involved, there is a direct relationship between the
waveform and the spectrum depending on the differential time ( t).
Knowledge of the modulation t helps determine the resolution required for

detailed spectral analysis. Also, from our previous discussion on modulation,
we know there is a carrier frequency that the modulation must follow.
Gears, bearings, and electrical defects each have carrier frequencies. For gears
the carrier is the frequency where the gears mesh. However, a carrier
frequency for an electrical defect could be the line frequency (FL ) or 2 * FL.
Believe it or not, the data above is real. Generated from a slow-speed

application, this is a gear defect with sidebanding of the output gear speed.
There is a direct relationship between the waveform and spectrum, but in some
cases it is not easy to identify.
Waveform Analysis As Confirmation

One of the most functional uses of waveform analysis is to confirm the
diagnosis of a particular fault type.
Every fault condition has a corresponding waveform characteristic.
Unbalance, for example, has a sinusoidal pattern with one major event per
revolution.
Misalignment, which is primarily offset, typically has harmonic activity with

the waveform having the same number of events per cycle as the spectral data
has peaks. A misalignment condition generating a second and possibly a third
order peak shows two or three sinewaves per revolution.
Looseness will have a complex waveform with many peaks within one
revolution. This will confirm the spectral characteristics of multiple harmonics
of turning speed.
The following examples show how to use waveform data to confirm existing
spectral analysis.
We are not suggesting that the primary diagnosis be conducted in the time
domain. But, it should be used to reassure the analyst of his/her judgement
call.
Vertical Turbine Pump Unbalance Example

The vertical pump motor bearing spectrums show unbalance as the major
forcing function. The time waveform confirms the diagnosis. The motor was
shaking so bad, the leads were chafing in the terminal box and grounding to
the frame.
MOH
MOV
MIV
MIH
Multispectra Plot - Data Comparison
The multiple point spectrum plot below shows radial and axial measurements
taken from the top of the vertical motor.
FWEL - FRESH WATER BOOSTER PUMP 1
131-546-03 - PTS=MOH MOV MOA
Max Amp
.65
Plot
Scale
131-546-03-MOA
PK Velocity in In/Sec
0.7 09-FEB-96 09:22
131-546-03-MOV
0
09-FEB-96 09:22
131-546-03-MOH
09-FEB-96 09:22
0 3 6 9 12 15 18 21 24 27
Frequency in Order
VA2CHSTY.DAT, BAL-IMBALANCE, 131-546-03, FRESH WATER BOOSTER PUMP 1
The spectrums show a dominant peak at 1X turning speed. The highest level is
at the MOV position. Note the absence of harmonic activity and low level of
non-synchronous energy. The radial directions are much higher than the axial.
Single Spectrum - Amplitude Relations
The single spectrum plot below taken from the top of the motor, MOV, shows
the predominant signal at 1X RPM (1776).
The harmonic cursor shows the lack of any harmonic activity. The sharpness
of the peak indicates that it has been created from a waveform dominated by a
single frequency.

131-546-03-MOV MOTOR OUTBOARD VERTICAL
1.0
Route Spectrum
09-FEB-96 09:22
0.8
OVRALL= .6466 V-DG
PK = .6464
LOAD = 100.0
0.6 RPM = 1776.
RPS = 29.60
0.4
0.2
0
Ordr: 1.000
0 5 10 15 20 25 30 35 40 45 50
Freq: 1776.2
Frequency in Order Spec: .646
Time Waveform - Sinusoidal
The time waveform below shows the last time block stored by the analyzer.
Approximately 270 milliseconds of time (8 shaft revolutions) shows the clear

one per revolution signal generated by the unbalance condition. The
repeatability of the signal and the lack of impacting confirm the diagnosis.

131-546-03-MOV MOTOR OUTBOARD VERTICAL
0.8
Waveform Display
0.6 09-FEB-96 09:22
0.4 RMS = .2679

LOAD = 100.0
Acceleration in G-s
0.2 RPM = 1776.

RPS = 29.60
-0.0
PK(+) = .5932
PK(-) = .6215
-0.2
CRESTF= 2.32
-0.4
-0.6
-0.8
0 60 120 180 240 300

Time in mSecs
Fan Bearing Looseness Example

The fan bearing looseness data provides the initial spectral data for diagnostics
and the waveform data to confirm the looseness fault diagnosis.
The fan ran in an out of balance condition for two years. The bearings now
have excess clearance, allowing the shaft to move around.
Multispectra - Data Comparison
The order-based plot below shows radial and axial measurements on the
outboard and inboard turbine bearings.
The major peaks in the radial directions are all found at one order of running
speed (i.e., 1 x running speed). Also, the highest amplitudes of the plot appear
to be in the horizontal planes at the inboard and outboard locations.
Also note the small amounts of harmonic activity and axial data amplitude.

C-20 FLOAT - PTS=FIH FIV FIA FOH FOV FOA
Max Amp
.46
C-20 FLOAT-FOA
Plot 14-JUN-95 08:10

Scale
C-20 FLOAT-FOV
1.0 14-JUN-95 08:10
C-20 FLOAT-FOH
14-JUN-95 08:09
C-20 FLOAT-FIA
0
14-JUN-95 08:09
C-20 FLOAT-FIV
14-JUN-95 08:09
C-20 FLOAT-FIH
14-JUN-95 08:08
Ordr: 1.000
0 2 4 6 8 10 12 14 16
Freq: 3499.0
Frequency in Order Sp 1: .395
Single Spectrum - Amplitude Relations

0.6
Route Spectrum
14-JUN-95 08:08
0.5
OVRALL= .5095 V-DG
PK = .5065
0.4
LOAD = 100.0
RPM = 3498.
RPS = 58.30
0.3
0.2
0.1
0
Ordr: 1.000
0 2 4 6 8 10 12 14 16
Freq: 3499.0
Label: HARMONICS-BALANCE/LOOSENESS
The spectral plot above shows vibration in the horizontal direction on the fan
outboard bearing.
The cursor markers note the locations of harmonics of running speed.
Virtually all the vibration energy in this spectrum is caused by turning speed
and harmonics. The sides, or skirts, of this peak are also very narrow.
The number of harmonics tells us that the spectrum is derived from a complex,
repetitive time waveform.
Time Waveform - Sinusoidal Character

The waveform below shows instantaneous acceleration over 240 milliseconds
in the horizontal direction on the fan inboard bearing. Vertical lines show the
time required for the fan shaft to make one complete revolution. The
waveform is in units of acceleration since the probe used to collect data was an
accelerometer.

4
Waveform Display
3 14-JUN-95 08:08
2 RMS = 1.06
LOAD = 100.0
Acceleration in G-s
1 RPM = 3498.
RPS = 58.30
0
PK(+) = 3.08
PK(-) = 3.01
-1
CRESTF= 2.88
-2
-3
-4
0 30 60 90 120 150 180 210 240

Time in mSecs
Label: HARMONICS-BALANCE/LOOSENESS
A clear and repeatable waveform occurs once per shaft revolution, 1 x RPM.
There is also multiple peaks within one revolution The waveform shows the
acceleration created on the bearing housing by the looseness. The repeatability
of the waveform in time with respect to the shaft turning speed and amplitude
means that the vibration force is tied to the shaft running speed. This is true of
waveforms collected in acceleration, since they can be influenced by high-
frequency energy bursts, such as impacts.
Motor to Pump Misalignment Example
The pump is installed below ground and supplies water for household use from
a water treatment facility. The pump has had high vibration since installation
and numerous seal/packing and bearing failures. The maintenance personnel
stated that the alignment was “difficult” because the base was drilled
incorrectly at the manufacturers facility.
Multispectra - Amplitude Comparison
#1 - TIMBERLINE BOOSTER (PROSPECT

TIMBSTRPRO - PTS=MOH MOV MIH MIV MIA
Max Amp
.43 TIMBSTRPRO-MIA
21-JUN-95 16:12
Plot
Scale
TIMBSTRPRO-MIV
0.5 21-JUN-95 16:11
TIMBSTRPRO-MIH
0
21-JUN-95 16:11
TIMBSTRPRO-MOV
21-JUN-95 16:11
TIMBSTRPRO-MOH
21-JUN-95 16:11
0 5 10 15 20 25 30 35 40 45 50 55
Frequency in Order
VA2CHSTY.DAT, MIS-MISALIGNMENT, TIMBSTRPRO, TIMBERLINE BOOSTER
The multispectral plot above shows inboard and outboard motor points. At
first glance, the problem might appear to be unbalance. If we take a closer look
we see that 2X running speed peaks are present in all directions, which may
indicate an alignment or looseness problem. This is a good time to examine the
time waveform.
Single Spectrum - 2X T.S.

TIMBSTRPRO-MIV MOTOR INBOARD VERTICAL
0.24
Reference Spectrum
0.21 AAAAA 21-JUN-95 16:11
0.18 OVRALL= .1780 V-DG

PK = .1771
0.15 LOAD = 100.0

RPM = 1768.
RPS = 29.47
0.12
A=MOTOR HARMONIC
0.09
: 1.00
0.06
0.03
0
Ordr: 1.004
0 3 6 9 12 15 18 21 24 27
Freq: 1774.9
The single spectrum display of the motor inboard vertical position is shown
above.
Harmonics of running speed are denoted by the fault frequency markers

(dashed lines).
The first through sixth orders of running speed are visible with the 2X T.S.
predominant. The peaks are sharp and very little non-synchronous energy is
present, therefore the waveform is repeatable.
The diagnosis can be confirmed by viewing the time waveform.
Time Waveform - Twice per Revolution

TIMBSTRPRO-MIV MOTOR INBOARD VERTICAL
0.8
Waveform Display
0.6 21-JUN-95 16:11
1
2
0.4 RMS = .1784
LOAD = 100.0
Acceleration in G-s
0.2 RPM = 1768.

RPS = 29.47
-0.0
PK(+) = .5682
PK(-) = .5457
-0.2
CRESTF= 3.19
-0.4
-0.6
-0.8
0 60 120 180 240 300

Time in mSecs
The time waveform for the motor inboard horizontal position is shown above.
The vertical lines denote the time required for one shaft revolution.
The waveform is repetitive for each revolution with two distinct peaks for each
period.
Pump Bearing Looseness Example
The diagram above shows a centerhung pump with bearing housing dimensions
worn oversize .
The motor turns at about 1,775 RPM.
The worn housings makes the pump very loose . Typical of many looseness
problems, this has grown worse over time. A small dimension problem has
gradually made itself worse.
Multi spectral - Broadband

CWTR - COOLING WATER PUMP 1
341-545-01 - PTS=PIV PIH POV POH POA
Max Amp
.14 341-545-01-POA
18-APR-96 08:47
Plot
Scale
341-545-01-POH
0.14 18-APR-96 08:47
341-545-01-POV
0
18-APR-96 08:46
341-545-01-PIH
18-APR-96 08:46
341-545-01-PIV
18-APR-96 08:46
0 400 800 1200 1600

Frequency in Hz
VA2CHSTY.DAT, BRG-BEARINGS, 341-545-01COOLING WATER PUMP 1
A Multi spectral plot of all five pump points is shown above.
Inboard and outboard radial measurements and an outboard axial measurement

were made.
Many harmonics of running speed are visible on all measurement positions.

Baseline or floor energy is also very visible.
Single Spectral - 9-15x T.S. and Broadband

341-545-01-POA PUMP OUTBOARD AXIAL
0.14
Route Spectrum
18-APR-96 08:47
0.12
OVRALL= .3663 V-DG

PK Velocity in In/Sec 0.10 PK = .3675
LOAD = 100.0
0.08 RPM = 1775.
RPS = 29.58
0.06
0.04
0.02
0
Freq: 29.58
0 400 800 1200 1600
Ordr: 1.000
Frequency in Hz Spec: .03901
A full-screen view of the motor inboard horizontal position is shown above.
A cursor is positioned at 1x running speed and on the harmonics of running

speed.
The peaks are broad and have wide skirts.
Notice, no individual peak exceeds .1 in/sec, but the overall energy is .3663
in/sec.
This is common with looseness. Broad humps of energy show up in the 9X to

15x running speed range.
This indicates that the time waveform cannot be cleanly transformed into a
spectrum. Therefore, the waveform must have random, non-periodic energy
present.
Time Waveform - Random Energy

341-545-01-POA PUMP OUTBOARD AXIAL
10
Waveform Display
8 18-APR-96 08:47
6 RMS = 2.12
LOAD = 100.0
Acceleration in G-s
4
RPM = 1775.
RPS = 29.58
2
PK(+) = 7.68
0
PK(-) = 6.42
-2 CRESTF= 3.63
-4
-6
-8
0 60 120 180 240 300

Time in mSecs
The time waveform used to calculate the motor inboard horizontal position is
shown above.
Vertical lines show the time required for the shaft to complete one full
revolution. This waveform is non-repeatable and non-periodic.
There is no similarity in its pattern from revolution to revolution. Therefore,

the vibration is not tied to the turning speed of the shaft. It makes sense that
the waveform trace is not repeatable if the shaft does have a looseness
problem and that the pattern would not be tied to the turning speed of the
shaft.
Non-periodic, random patterns do not convert well in the FFT process. It is

very difficult to assign specific frequencies and amplitudes to patterns in
waveforms like this one.
This difficulty leads to the broadband energy humps in the spectrum. Broader
humps indicate more random energy. Higher humps indicate more impacting
in the waveform.
Rolling Element Bearing Example
The motor and gearbox above turn the cable reel for a large ski-lift.
The maintenance personnel reported vibration from the back end of the motor
after only 200 hours operating time on a newly installed drive.
The analyst investigated and found visible flakes of a bronze colored material
near the back end of the motor. The motor manufacturer was contacted to
determine if the 6330 bearings had a bronze retainer and the reply given was
no.
Multi spectra - Non-Synchronous Energy

NSTR - BACKSIDE QUAD
BACKSIDEQD - PTS=MOV MOA MOH
Max Amp
.19
Plot
Scale
BACKSIDEQD-MOH
0.20 05-JAN-96 08:56
BACKSIDEQD-MOA
0
05-JAN-96 08:57
BACKSIDEQD-MOV
05-JAN-96 08:46
0 400 800 1200 1600 2000

Frequency in Hz
VA2CHSTY.DAT, BRG-BEARINGS, BACKSIDEQD-BACKSIDE QUAD
This is a Multi spectral plot of the motor point outboard points.
All the levels appear very low in amplitude, but notice the location of the
dominant peaks.
There appears to be groups of many peaks closely spaced in the mid to higher
frequency range. These “mounds of energy” can indicate bearing defects.
We need to look at the spectrum a little closer.
Single Spectrum - Bad Bearing

BACKSIDEQD-MOH MOTOR OUTBOARD HORIZONTAL
0.30
Analyze Spectrum
C CC CC CC CC C 05-JAN-96 08:56
0.24
PK = .3611
LOAD = 100.0
RPM = 1298.
0.18 RPS = 21.64
>SKF 6330
0.12 C=BPFO : 77.64
0.06
0
Freq: 77.50
0 400 800 1200 1600 2000
Ordr: 3.582
Frequency in Hz Spec: .03297
Label: OUTER RACE FREQUENCIES W/CAGE SB
A full screen plot of the motor outboard horizontal position is shown above.
The fault frequencies for the 6330 bearing ball pass frequency outer race are
marked. Notice the number of peaks surrounding the higher frequency defect
harmonics.
The large number of harmonics and sidebands will be created from a complex
waveform.
Time Waveform - Bad Bearing

BACKSIDEQD-MOH MOTOR OUTBOARD HORIZONTAL
8
Waveform Display
05-JAN-96 08:56
6
RMS = 1.75
4 LOAD = 100.0
Acceleration in G-s
RPM = 1298.
2 RPS = 21.64
PK(+) = 5.81
0
PK(-) = 5.13
CRESTF= 3.32
-2
-4
-6
0 40 80 120 160 200

Time in mSecs
Label: OUTER RACE FREQUENCIES W/CAGE SB
The number and height of the spikes in the time waveform confirm the
presence of severe impacting.
Many exceed ± 10 g’s in amplitude. Remember that this is a very large DC

motor with an armature that weighs 4,000 lbs., so large amounts of energy are
generated by these impact levels. This impact energy is absorbed by the
bearings and is damaging them rapidly.
The waveform shape is random and complex. This shape cannot be

transformed into a clean spectrum, so the spectrum on the previous page with
broad humps of energy is created.
The motor was disassembled in place because of the difficulty of removing a

large motor in 12 feet of snow at 11,000 foot elevation. The bearing cage
turned out to be bronze! It was deteriorating and did not have much life left.
The outer race had major spalls from impacting balls. The bearing was
replaced.
Summary
Waveform data may be used for much more than what is typically seen in
industry. The ability to check for specific characteristics such as periodicity
and modulation, helps the analysis process.
Energy balance (asymmetry) may be checked for direction of signal and for the
predominant traits of the signal.
Overall waveform is much more understandable and useful than most would
lead us to believe. However, this section enhances your analysis abilities using
the time waveform.
Section Eleven
Peak / Phase Analysis
Objectives
< Define peak/phase data and the convention used to measure it.
< Describe reasons for monitoring peak/phase data.
< Describe and interpret display formats for peak/phase data.
General Description
The monitor peak/phase data defines the location (in degrees) of a machine’s
vibration peak with respect to a fixed reference mark on the rotor. Several
reasons to collect peak/phase data appear below:
< Peak/Phase data provides information regarding resonant frequencies of a

machine when collected during startup or coastdown.
< Peak/Phase data proves useful for field balancing work.
< Peak/Phase data helps diagnose various machinery faults such as coupling
misalignment and bent shafts.
1/99 © Copyright 1999, Computational Systems, Incorporated. All Rights Reserved. 11-1
Phase Data
Phase can be defined in two ways. First, phase is that part of a cycle (0E to
360E) through which a particular part of a machine travels relative to a fixed
reference point. Second, phase refers to the location in degrees that marks a
machine’s high vibration peak and its frequency relative to a fixed reference
mark on a rotating component of the machine. This second type of phase is also
known as synchronous phase.
Photo T
Tach
CL
Tim
Phase = •2
e
tT (radians)
x
Phase = t T •360
Single Channel Phase Measurement (@

1XTS)
Phase varies with the monitored frequency. For machinery diagnostic work,
synchronous phase refers to 1xTS or any harmonic of turning speed up to 128
orders. CSI measures phase counter to the rotation of the shaft.
11-2 © Copyright 1999, Computational Systems, Incorporated. All Rights Reserved. 1/99
Specific Fault Types

Unbalance
Static unbalance shows a zero degree phase shift across the rotor radial-to-radial
or horizontal-to-horizontal, a 90E (±20E) phase shift from vertical to horizontal
at the same bearing location.
Dynamic unbalance shows a phase shift across the rotor radial-to-radial or

horizontal-to-horizontal related to the heavy spots on each end of the rotor. If
the heavy spots are 180E out of phase on each end, then the phase measurements
will also be 180E out of phase.
Misalignment
Angular misalignment typically shows a 180E (± 30E) phase shift across the
coupling in the axial direction.
Parallel misalignment tends to show a 180E(± 30E) phase shift across the
coupling in a radial direction.
Bearing misalignment (cocked bearing) shows a 180E(± 30E) phase shift from
one side of the bearing to the other or from the top of the bearing to the bottom.
Looseness and Soft Foot

Phase reading with looseness will be erratic from point to point around the
machine train. A soft or loose mounting foot usually shows a phase shift from
the foot itself to the foundation. This shift is often greater than 90E. Also, a soft
foot shows a phase difference from the other machine feet.
Resonance
Typically, for a lightly damped system, the phase shifts 180E as the turning
speed moves through the resonant frequency with approximately a 90E phase
shift occurring at the resonant frequency.
Phase Measurements
Phase is measured by using the once-per-revolution pulse as a reference point.
Depending on the machine fault, a machine’s structure moves either as one unit,
in phase, or separately, out of phase.
Usually the amplitude and the phase of the 1xTS frequency is of the most
interest. The relationship between the high vibration amplitude and the
tachometer pulse determines the phase. Phase is the difference in time,
measured in degrees, between the leading edge of the tach pulse and the peak
amplitude in the waveform. As the vibration sensor changes from a vertical
position to a horizontal position, a 90E shift, the time between the tach pulse and
the peak amplitude will also change 90E.
Transducer shifted 90E yields a 90E

shift in phase.
Monitoring the amplitude and phase response during a startup or coast down is a
method of using peak and phase measurements to determine resonances. The
analyzer should be placed in monitor peak and phase mode at the turning speed
harmonic of interest. The machine should then be shut off. Take care to ensure
that the peak and phase data are sampled fast enough to capture meaningful data.
Monitor (Synchronous) Peak/Phase Lab

This lab covers hands-on operation of the Model 2120 Machinery Analyzer
using the Monitor Peak/Phase option.
Press the Analyze key to display the Analyzer Functions menu. Highlight 6)
MONITOR MODE and press Enter.
Highlight 5) MONITOR PEAK/PHASE and press Enter.
For ORDER (XRPM), specify an integer multiple of the machine turning speed
or RPM, 1 is normally used.
BANDWIDTH (XRPM) sets the bandwidth of the tracking filter. The analyzer
attenuates all frequency components outside the specified bandwidth. Multiply
peak frequency times parameter bandwidth for filter bandwidth. Accepted
values range from 0.05 to 0.5, but 0.10 works well for most applications. A
more narrow filter (a smaller bandwidth) helps isolate the indicated peak, but it
requires more time per measurement.
When you set AVERAGER ENABLED to YES, the analyzer calculates a new
vector average for each peak/phase average. This helps when readings fluctuate
or when nearby components modulate the frequency. Use the CLR key at any
time to clear the averaged vector and restart the averaging process. The
TACHMTR TIMEOUT will display the NO TACH message when set to YES.
Setting this to NO will disable the message. The FS RANGE when set to 0, will
AutoRange the signal before each measurement. Setting a value will allow you
to collect data more quickly. The units are in the same units as the waveform.
Press Enter to collect data.
When you finish data collection, press Enter to show the Display Functions
menu. Highlight 5) SAVE DATA and press Enter.
Coastdown / StartUp Tests
To view the collected data, follow the same procedure used to monitor the 1xTS
peak and phase data. Press the Analyze key. Highlight 6) MONITOR MODE
and press Enter to continue.
Highlight 5) MONITOR PEAK/PHASE and press Enter.
The setup above is identical to the previous setup.
With the machine running, the Monitor Peak/Phase menu displays the amplitude
and phase data on the machine before the machine is shut off.
You can monitor the amplitude and phase as the machine coasts down to a stop
as the power to the machine is shut off.
The machine is near resonance at this point. Notice how the amplitude has
reached over 2 In/Sec and the phase has shifted 115 degrees.
As the machine coasts to a stop, the amplitude is negligible and the phase has
shifted more. Press Enter to exit the data scan.
To avoid inadvertently losing data, it is best to store the data right after data has
been collected.
After data is stored, view the Peak vs. RPM data by selecting 1) PEAK VS
RPM and pressing Enter.
The resonance appears in the data at a turning speed of 937.9 RPM.
Press Enter to return to the Display Functions Menu.
Select 2) PHASE VS RPM to view the phase change with respect to speed.
The phase has gone through about a 180E phase shift with about a 90E phase
shift at resonance.
This plot helps verify whether the peak you observed was caused by resonance.
Press Enter to return to the Display Functions menu.
The phase reading from a monitored machine should change by approximately

180E as it runs up to speed, passes through the resonance range(s), and moves
into the free space region.
The phase reading always becomes unstable at the resonant frequency. Phase at
resonance should, however, differ by roughly 90E from measurements in the free
space region. The same holds true for measurements taken when the machine
turns too slow to experience vibration amplification from resonance.
Highlight 3) NYQUIST PLOT and press Enter to view a polar plot of the data.
Bode plots (Peak or Phase vs. RPM) yield important information about
resonance. The presence of runout or a bow in the shaft, however, can
significantly alter the appearance of the plots. Nyquist plots, on the other hand,
remain unaffected by runout and bowed shafts. Use nyquist plots to confirm any
conclusion based on Bode plots.
The illustration below shows the important features of the Nyquist plot. Note
that any loops in the plot are evidence of the other resonant frequencies.
If the machine happens to be running and you wish to collect more data, press 4)
COLLECT MORE DATA. If you collect more data, you may want to clear the
data before continuing.
Interpreting the Coastdown Results

Resonant frequencies on a particular machine do not change unless degradation
occurs (e.g., cracked welds or eroded grout). For critical machines, know where
resonances occur before repair or rebuild work begins. This information
provides a baseline for comparison after the work is finished.
Resonant frequencies should remain the same before and after a rebuild unless
you make structural or component modifications. If the before and after
frequencies do not coincide, most likely the machine was not assembled
properly. Loose, missing, or improperly fitted components can cause resonant
frequencies to shift.
Machines that experience frequent bearing failures or structural cracking often

operate at or near a resonant frequency. Use coastdown analysis to verify this
condition.
Viewing Stored Peak and Phase Data
Press the Analyze key and highlight 6) MONITOR MODE and press Enter.
Select item 5) MONITOR PEAK/PHASE and press Enter.
Press F6 to display the stored data list.
Highlight the data set of interest and press Enter to display the data.
Select the desired plot and display function. In this case, select 1) PEAK VS.
RPM.
The peak versus RPM data displays here.
Summary
It is important to be able to monitor the amplitude and phase data at different
multiples of turning speed. This is another tool for doing “investigative”
vibration analysis. Phase is a useful tool for analyzing many of the common
faults in machinery. It also helps the analyst to visualize the operating
movement to diagnose the root cause of the vibration.
Monitoring the amplitude and phase data during a coastdown or a startup test is
a powerful analysis tool for confirming the presence of a resonance. Remember
to look for the amplitude increase relative to the frequency as well as the phase
change relative to the frequency.
Section Twelve
Gears
Objectives
< Describe the basic purpose of gears and their components with needed
gear theory.
< Identify transmission error and other gear defects and the effect on
vibration data.
< Select the best sensor, sensor location, and measurement point setup for
meaningful data collection.
Gear Design
Gears are commonly used in industry to provide the speed and power
transmission requirements needed in industrial applications.
Gears can provide these speed changes and torque transmission without slip.
Gear designs have specific characteristics that can affect the measured
vibration. Gears are often considered too complex to diagnose their defects
properly, but with an understanding of a few gear design basics and
terminology, troubleshooting gearboxes can be accomplished more easily.
Gears
Gear Types
Different gear types are available for various speed and power considerations.
Each of these different gear types shows the same basic vibration patterns when
gear defects appear.
Spur Gears
Spur gears are most commonly thought of when discussing gears. The teeth
are cut parallel to the shaft. These gears are good at power transmission and
speed changes, but are noisier than other gear types.
Gears
Helical Gears
Helical gears have teeth cut at an angle to the shaft.
These gears are much quieter than spur gears; but, due to the angular nature of
the gear meshing, axial thrust and, therefore, axial vibration is higher on these
gears than spur gears.
To avoid the higher axial thrust, double helical gears are used. These gears,
often called herringbone gears, are divided in the middle with each side having
an equal magnitude angle but in an opposite direction.
If a gap exists between the two halves of the gear, then it is a double helical
gear.
If no gap exists and each tooth is continuous, then the gear is called a
herringbone gear.
Gears
Bevel Gear
Right angle gearboxes transmit power to an output shaft perpendicular to the

drive shaft. These gears may use a bevel gear design to transmit the power
better. The bevel gear may have a straight bevel or a spiral bevel.
Worm Gear
( For Worm Drives ) GMF = (#T)out * (TS)out
A worm gear is also used to transmit the rotational motion between

perpendicular shafts. The worm has one or more teeth wrapped around its
shaft. It drives a worm wheel which has the appearance of a helical gear.
Gears
Basic Gear Theory

Tooth Shape
All the previously mentioned gear types use the same basic tooth design, called
a tooth profile. The best tooth profile is one that allows for the radial velocity
of the gears to be constant.
For modern gearing, the tooth profile that works best is called the involute.
The involute design minimizes the effect of fabrication errors on the radial
velocity of the gears keeping the vibration and noise levels down.
Conjugacy
The purpose of a gearbox is to provide power and or speed changes with a
minimum of excessive noise and vibration. To accomplish this, the power
from the drive gear must be transmitted though a line perpendicular to the
common tangent and intersecting the center-to-center line. The common
tangent is a line tangent to both of the meshing gears. This point of
intersection is called the pitch point. The pitch point of each gear tooth must be
on the center-to-center line between the gears. The circle that connects the
pitch points is called the pitch circle.
This is the principle of conjugacy. The use of the involute tooth profile allows
this condition to be met more easily.
Gears
Prime Number Theory

The number of teeth on each gear can be factored down to a series of prime
numbers.
Prime numbers are 1, 2, 3, 5, 7, 11, 13, 17, 19, etc. For example, the number
10 can be broken down to its prime factors 1 x 2 x 5, the number 26 can be
factored into 1 x 2 x 13. Prime numbers are important when trying to
understand some of the gear defects and their frequency components.
When the largest prime factor common to both gears, called the largest
common factor (LCF), is 1, one tooth on a drive gear will mesh with every
tooth on the driven gear before it remeshes with the first tooth on the driven
gear.
If the LCF is greater than 1, some teeth will mesh more often; and this leads to
an reduced gear life. Also, machining defects and wear patterns develop that
show up as defect frequencies based on the largest common prime factors
between the meshing gears.
Vibration Analysis
Vibration analysis of gears can provide a wealth of information about the
mechanical health of the gears. This section discusses the source of the
frequencies that may be present in a gear box.
The source of most of the defect frequencies is transmission error between two
meshing gears.
Transmission error is caused by machining errors, tooth deflections, looseness,

eccentricity, or anything that causes the power to be transferred through any
point other than the pitch point.
Gears
Gear Mesh Frequency

Gear mesh frequency (GMF) is the most commonly discussed gear frequency.
However, GMF by itself is not a defect frequency.
GMF is always present in the spectrum regardless of gear condition. Its

amplitude may vary, however, depending on the gear condition.
Single Reduction
GMF is simply defined as the number of teeth on a gear multiplied by its
turning speed.
GMF = (#Teeth) * (Turning Speed)
Example:
GMF = 44t * 29.59 Hz = 1301.96 Hz
1301.96 * 60 = 78,117.6 cpm
Gears
If the turning speed in the above equation is in units of RPM (or CPM), the
GMF will be in units of CPM. If turning speed is in orders, then the GMF will
be in orders. This relationship can be used to advantage when trying to
determine the output speed of a driven gear when the input speed and the
number of teeth on each gear is known. This is possible because any two
meshing gears must have the same gear mesh frequency. Therefore, the above
equation can be rewritten slightly.
(#T)in * (TS)in = GMF = (#T)out * (TS)out
When faced with the need to calculate an output speed for a single reduction
gear drive.
Calculate the gear mesh frequency for the known gear and divide by the
number of teeth on the output gear.
This results in the determination of the output speed. (This is also the same as
multiplying the input speed by the gear reduction ratio.)
For example, if the input speed is 1750 CPM and the input gear has 25 teeth
and the output gear has 17 teeth, then these values can be put into our
relationship and we can find the turning speed of the output gear.
(#T)in * (TS)in = GMF
(25) * (1750) = 43,750 CPM
GMF / (#T)out = (TS)out
43,750 / (17) = 2573.5 CPM
Gears
Multiple Reduction
A multiple reduction gearbox is not any more difficult to evaluate if two facts
are remembered.
Ô Gear mesh frequency is the product of the number of teeth on the

gear and its turning speed.
Õ Any two meshing gears must have the same gear mesh frequency.
The following example demonstrates these two principles.
The input drive gear’s turning speed is 59 Hz and it has 256 teeth. It meshes
with an intermediate gear that has 157 teeth and an unknown turning speed,
(TS)in.
The intermediate gear meshes with the output gear that has 94 teeth and an
unknown turning speed, (TS)out.
First determine the gear mesh frequency for the input gear.
(#T)in * (TS)in = GMF
(256) * (59) = 15,104 Hz
Gears
Since the gear mesh frequency is the same for two meshing gears, the
intermediate GMF is also 15,104 Hz.
The turning speed for the intermediate gear, (TS)int, is calculated below.
GMF / (#T)int = (TS)int
15104 / (157) = 96.2 Hz
Again for the output gear, the gear mesh frequency is the same between two
meshing gears. The output gear speed is determined below.
GMF / (#T)out = (TS)out
15104 / (94) = 160.7 Hz.
Even complicated gear drives can be computed if they are examined one
meshing pair at a time.
Gears
Worm Gears
Worm gears are sometimes confusing because there is a question as to how
many teeth are on the input worm drive.
In the case of a worm gear, it is not the number of teeth that is of concern
(often a worm drive only has one tooth), but the number of flights on the worm
gear.
The flights refer to the number of teeth that mesh with the driven gear during
one revolution of the worm drive.
This can be readily identified if the output gear speed, the number of teeth on
the output gear and the input shaft speed are known.
In this example an output drive gear with 24 teeth turning at 10 Hz is driven by

a worm gear turning at 29.5 Hz.
The number of flights (#F) on the input gear can be determined as follows.
(#T)out * (TS)out = GMF
(24) * (10) = 240 Hz
GMF / (TS)in = (#F)in
240 / 29.5 = 8.13
This worm gear has 8.13 flights meshing with 24 teeth on the output gear.
Gears
Planetary Gears
Probably the most confusing gear mesh frequency to calculate is for a planetary
gear set. There are many different types of planetary gear designs. One of these
is shown below. In this planetary gear, there are some new components that
need to be identified.
The input shaft is attached to the planet carrier which drives the planet gears.
The planet gears mesh with the ring gear and the sun gear which drives the
output shaft. For this gear set, the GMF is equal to the number of teeth on the
planet gear (#T)planet, multiplied by the speed of the planet gear. GMF is also
equal to the number of teeth on the sun gear (#T)sun multiplied by the output
shaft speed.
Now, the speed of the planet (TS)planet is determined by multiplying input shaft
speed (TS)in by the number of teeth on the ring gear (#T)ring and dividing by the
number of teeth on the planet gear (#T)planet.
(TS)planet = (TS)in * (#Tring / #Tplanet)
Then,
GMF = (#T)planet * (TS)planet
Once the gear mesh frequency for the planet gear system is found, any of the
turning speeds can be determined by:
Dividing the GMF by the number of teeth on that gear.
Gears
Fractional Gear Mesh

Gear mesh frequency is always present in the vibration signal of a gearbox.
Depending on the wear patterns and the Largest Common Factor (LCF),
fractional gear mesh harmonics may appear. If the LCF is 1, only one gear
mesh will appear. If the LCF is 2, the ½ gear mesh will appear in the spectrum
as the gears become worn. Likewise if the LCF is 3, the 1/3 and 2/3 gear mesh
frequency will appear in the spectrum.
Eccentricity also causes fractional harmonics of GMF. If the largest common

factor is one, but every other tooth is raised, then again ½ GMF will be present.
If a gear has five spokes then it is possible to have five high spots around the
gear and 1/5, 2/5, 3/5, 4/5 GMF will appear in addition to the GMF.
Multiples of Gear Mesh

Gear misalignment shows up as harmonics of the gear mesh frequency.
Typically the second and third harmonics are most significant when trending
this defect.
If the second gear mesh harmonic is higher in amplitude than the GMF itself,
then it is very possible too much backlash exists in the gear set and the gear
teeth may be impacting twice during the meshing process: the normal impact
during the initial contact and an additional impact during the end of the mesh.
Effect of Load on GMF

The effect of load on the gear set has two contrasting effects on the GMF
amplitude depending upon the defect. The general effect of increased load is to
increase the amplitude of the gear mesh frequency.
The opposite effect can be expected if the gear has too much backlash present.
Too much backlash occurs as the gears become worn and the clearances
between the meshing gears increase.
Gears
Sidebands
In gear analysis, sidebands can be valuable in diagnosing gear defects.
Sidebands show up as frequencies on either side of the GMF.
The sideband frequency spacing is equal to the turning speed of either the input
shaft speed or the output shaft speed.
The spacing of the sidebands is equal to the turning speed of the gear with the
defect. Sidebands appear most commonly because of wear, looseness, and
eccentricity.
While the presence of sidebands is important, the amplitude relative to the

GMF amplitude is more significant than the amplitude of the GMF.
If the amplitude of the sidebands approaches the amplitude of the GMF, the
defect could be severe.
Gear Resonance
One frequency not easy to calculate is the gear resonant frequency. Resonant
frequency occurs naturally in all structures, but does not appear in the spectral
data unless some other frequency excites the resonance. In gearboxes,
eccentricity problems and excessive looseness cause the teeth to mesh with
excessive force causing high levels of impacting in the machine that cause the
gear resonant frequencies to be excited.
Hunting Tooth Frequency

If a tooth has a manufacturing defect, then it will have a defect frequency
associated with it. This Hunting Tooth Frequency (HTF) is subsynchronous as
the tooth repetition frequency is less than turning speed. The HTF is simply
equal to the product of the GMF and the Largest Common Factor (LCM)
between the meshing gears divided by the product of the number of teeth on
each gear.
HTF = (GMF * LCF) / (#Tin * #Tout)
This frequency, if present, will be very low and may even be present as a
sideband frequency, detectible only by using envelope demodulation.
Sometimes HTF is referred to as the tooth repetition frequency.
Gears
Broken Tooth
The effect of a broken tooth is difficult to detect when using only the spectral
data.
If one tooth is broken then a pulse will be generated once per revolution of the
gear with the broken tooth.
This is simply a 1xTS frequency. The way to detect a broken tooth is to

examine the time waveform and look for an impact occurring at a time spacing
equal to 1xTS. The time waveform is not sinusoidal but has the impact and
ring down once per revolution.
Audible Noise
Unfortunately many gearboxes are considered problems because they are noisy.
However, noise levels are not always a good indication of gear condition.
Very often the GMF or any of the other gear defect frequencies simply excite
the natural resonances of the gearbox cover.
This causes the radiated airborne noise levels to increase significantly.
Sometimes lubricating oil gets trapped between the meshing teeth and is forced
out at extremely high velocities which can cause the audible noise levels to be
high.
Gears
Measurement Considerations
Sensor Selection
Now the spectral data will not do any good if the frequencies of interest are not
measured. Several things must be identified before the measurement points are
developed.
Which frequencies are important in the analysis of gears?
Low frequencies, such as the hunting tooth frequency all the way to 2x or
3xGMF.
Often it is recommended to set the Fmax at (2xGMF) + 5xTS in order to see the
gear misalignment defect in addition to any sidebands around the 2xGMF.
However, if the maximum frequency selected, Fmax, is higher than the usable
frequency range of the transducer, then a high frequency accelerometer will
need to be used in addition to the sensor normally used.
This would mean two measurements taken at the same position. Otherwise, a
lower Fmax could be selected at the cost of unmeasured data. Be sure and use a
sensor that accurately measures all the frequencies of interest.
Sensor Attachment
Once the proper sensor has been chosen, make sure the proper attachment
method is used. Be aware of the frequency response and the mounting
resonances present due to the different mounting methods.
A high-frequency accelerometer attached with a small, but strong magnet may

be acceptable. Some cases of very high frequencies, above 10,000 Hz may
require stud mounting to get good vibration data.
For spur gears the radial directions provide the most important information
because of the direction that the forces are being transmitted though the gears.
Helical gears experience a significant amount of axial thrusting and therefore
the axial direction contains the best information for the analysis of these gears.
The gearbox covers are not good locations for data collection because of
resonances in the gear covers. The bearing locations or the heads of bolts are
the most acceptable measurement locations.
Gears
Measurement Point Turning Speed

As the measurement points are being defined for data collection, it is important
to realize that the turning speeds will be changing as the speeds are reduced or
increased through a gearbox.
If the Fmax is set to 2xGMF, this could be acceptable for each measurement
point along the gear train.
However, pay attention to the lines of resolution that have been selected and
adjust them to keep the bandwidth at an acceptable range.
Summary
This section has covered a wide range of topics including the theory of gears,
vibration analysis of gears, and measurement point definitions for data
collection.
An understanding of the topics covered in this section will lead to a more

confident ability to perform vibration analysis of even the most complex gear
trains.
Gears
Case Histories
Product Winder Gear Case #1
Ô 1250 HP DC motor driving gearbox. 811 rpm
Õ Input pinion has 24 teeth and meshes with a 72 tooth gear.
Ö Each output shaft has a 24 tooth pinion.
Gears
Gear Case #1
The above spectrum was taken on the gearbox at the outboard horizontal
position of the input shaft. Shaft speed is 811 rpm.
What is the calculated Gear Mesh Frequency ?
GMF = # T x turning speed
24t x 811cpm = 19,464 cpm (19, 464 divided by 60 = 324.4 Hz)
Gear Mesh Frequency (GMF) is marked by the primary cursor at 24 orders.
1 order = 811cpm ( 24 orders x 811cpm = 19,464 cpm )
Notice the harmonic cursors are showing the presence of multiples of GMF.
The amplitudes of these multiples are low, however, their presence does
indicate potential problems.
Gears
Gear Case #1
This is still looking at the input shaft data.
This spectrum shows the set mark at GMF, with sideband cursors marking
peaks at .333 x TS or 1/3 order.
It appears there are many multiples of this sideband. What does this sideband
spacing indicate? Recall the case history information stated the reduction ratio
in the gearbox is 3:1.
We need to ask ourselves what would create a frequency at 1/3 order of the
turning speed of the shaft we are analyzing the data on.
Again the turning speed of this shaft is 811 rpm.
811 rpm divided by 3 = 270 rpm
Again, what would create a sideband frequency at 1/3 order (270 rpm) ?
Let’s look at some more data.
Gears
Gear Case #1
This is the intermediate shaft data.
A set mark has been placed on the GMF. The sideband cursor shows sidebands
spaced at 1xTS. The spacing of these sidebands determine which shaft has the
defective gear.
Check the reference rpm on this measurement point on the spectrum. The
reference is listed as 272rpm.
The DORD is listed as .995 orders. Or simply ( 1 Order ).
These sidebands are spaced at the turn speed of this shaft, the data displayed
above is of this shaft.
This means that the gear on “this shaft” has bad teeth on it.
Case solved ?
Not exactly, let’s look at some more data
Gears
Gear Case #1
Data displayed below is from the last Output shaft.
This shaft speed is 270 rpm.
The primary cursor is marking the GMF, the sideband cursors are marking the
DORD at 1 order.
This means this shaft also has a gear that has bad teeth on it.
When the gear-box was taken down for repairs and inspected for damage,
the gear on the intermediate shaft was bad and the gears on the output
shaft also were bad.
Now is the Case solved ?
Yes, but we should find the root cause of this failure.
Was it lubrication, poor installation, faulty gears, excessive loads, or did it

finally die of old age?
Gears
Rotary Screw Compressor

Gear Case #2
Ô 500 HP, 1800 RPM Motor.
Õ Compressor is driven by intermittent gearing.
Ö The motor gear has 66 teeth and the compressor gear has 61 teeth.
Gears
Gear Case #2
The above spectrum shows data collected from the compressor inboard
horizontal. A cursor is set on the high speed shaft’s GMF (61xTS).
Notice the peaks above and below GMF.
Gears
Gear Case #2
The compressor inboard vertical measurement point data is seen above. The
cursor has been set on GMF(61xTS). The vertical data also shows the presence
of peaks around GMF.
The next page shows the same data with these peaks marked.
Gears
Gear Case #2
A mark has been set on GMF and the cursor marks the highest peak.
The spacing on this peak is equal to 1xTS.
This spacing has determined the defect to be on the compressor gear. Also,
notice the amplitude of this sideband compared to the GMF amplitude.
Gears
Surge Cake Mixer

Gear Case #3
Õ Double-Reduction gearbox direct driven from the motor.
Ö Input pinion has 15 teeth and meshes with a 91 tooth gear.
Gears
Gear Case #3
The multiple spectrum plot is displayed above from the gearbox outboard
vertical point for the input shaft.
The cursor marks GMF (15xTS) of the input shaft.
The data from April shows the peak between 20 - 40 orders has changed.
The next page shows data from February.
Gears
Gear Case #3
A harmonic cursor is set at GMF (15xTS) and shows five multiples of GMF.
Notice the 2x GMF peak is higher than the primary GMF peak.
As with other types of equipment, 2x GMF is an indication of misalignment.
Gears
Gear Case #3
The 2x GMF peak has changed from the data collected in February.
It appears the 2x GMF peak has developed sidebands.
The next page shows an expanded view of this group of peaks.
Gears
Gear Case #3
A mark has been set at 2x GMF with the sideband cursor showing a spacing of
1xTS.
This spacing indicates a problem with the input shaft pinion.
This unit was sent in for repairs and the input pinion and gear were found to be
misaligned.
The misalignment had caused an uneven wear pattern across the face of the
gear teeth.
Section Thirteen
Electrical Defects
Objectives
< Explain the electrical defects associated with AC and DC motors.
< Describe the differences between Induction, Synchronous and DC motors.
< Identify the various electrical defects associated with the above types of
motors.
< Explain analysis techniques using single-channel data acquisition.
General Description
Industry utilizes a large number of both AC and DC motors for a wide variety
of applications.
These motors can be broken down into two main components:
The rotor and the stator.
How these two components electrically interact will determine the

characteristics associated with specific electric motor defects.
Although most vibration problems associated with electrical systems are motor
related, do not overlook other sources.
C Loose laminations in the power transformer

C SCR pulses in the speed controller
C Unbalanced supply voltage
C High currents from welders
C Solenoids
Electrical Defects
AC Induction Motor Analysis

Rotor Related Defects
Broken or Open Rotor Bar
Broken Rotor Bar

Rotor Bars
Sidebands Equal
Rotor Bar Pass ± 2 L.F.
Sidebands Equal
Slip Freq • #Poles
Broken or Open Rotor Bars

The predominant vibration appears at 1xTS with sidebands spaced at a
frequency equal to the number of poles on the motor multiplied by its slip
frequency.
Rotor slot frequency may also be present with sidebands spaced at 2x line
frequency.
Electrical Defects
For the Vibration Analyst to be proficient at calling electrical problems on their

equipment it is essential the calculations below are understood.
Please consider the following:
Slip Frequency = Magnetic Field Freq. - Rotor Freq.
The magnetic field frequency is the rated speed listed on the motor tag.
The rotor frequency is the actual speed of the motor.
Example:
Rated motor speed = 1800 rpm
Actual motor speed = 1785 rpm
Slip Frequency = 1800 - 1785 = 15 rpm
-----------------------------------------------------------------------------------------------
Rotor Slot Frequency = # of Rotor Slots * Rotor Freq.
The # of Slots = how many slots the rotor has on it
Rotor Frequency = the actual speed of the rotor
Example:
44 slots on rotor x 1785 = 78,540 cpm = Rotor Slot Frequency
-----------------------------------------------------------------------------------------------
Stator Slot Frequency = # of Stator Slots * Rotor Freq.
The # of slots on the Stator x actual speed on the rotor
Example:
46 slots on the Stator x 1785 = 82,110 cpm = Stator Slot Frequency
-----------------------------------------------------------------------------------------------
# of Motor Poles = [Line Freq(Hz) * 120] / Magnetic Field Freq(CPM)
Example:
60 Hz x 120 = 7200 divided by 1800 = 4 poles
Electrical Defects
The waveform resembles that for imbalance, because the rotor actually is out of
balance. In a live, real time, display of the waveform, however, you can see the
electrically induced vibration disappear immediately when you turn off the
power to the motor. The signal disappearance helps you differentiate this fault
from mechanical imbalance, where the vibration would coast down with shaft
speed.
Loose Iron or Slot

The primary vibration appears in the spectrum at 2xLF and slot frequency with
sidebands spaced at 2xLF. The waveform shows a dominant cycle of 2xLF.
Variable Air Gap

Air Gap
Stator
Dynamic Eccentricity
The principal vibration appears in the spectrum at 1xTS with sidebands spaced
at slip frequency multiplied by the number of motor poles and/or 2xLF.
Slot frequency may also be present with sidebands spaced at 2xLF. The
waveform resembles that for imbalance, because the rotor is actually out of
balance. In a live, real time display of the waveform, however, you can see the
motor power. The signal disappearance (when power is shut off) helps you
distinguish this fault from mechanical imbalance.
Electrical Defects
Stator Related Defects
Static Eccentricity
Static eccentricity is also known as stator eccentricity.
An eccentric stator will cause an unequal air gap between the rotor and the
stator. This results in the spectrum having a primary peak at 2xLF that is very
directional. Slot frequency with sidebands spaced at 2xLF may also be present.
A poorly built stator, improper lamination, stamping, loose laminations, warped
motor base or a soft foot are all possible causes of this defect. The waveform
demonstrates a dominant cycle at a frequency related to the 2xLF. As with
other electrical faults, you can confirm this defect by monitoring the signal in a
live, real time waveform and securing the motor. The energy should disappear
immediately.
If a soft foot is suspected, monitor 2xLF in a live spectrum and loosen one
motor foot bolt at a time observing any change in the amplitude at 2xLF. If the
amplitude at 2xLF is reduced during this procedure, then a motor soft foot is
probable. This test does not determine which of the motor feet has the
potential problem.
Electrical Defects
Loose Stator Laminations or Loose Stator Core

The primary vibration appears in the spectrum at 2xLF and harmonics of 2xLF.
The waveform demonstrates cycles at frequencies related to 2xLF. As with

other electrical faults, you can confirm this problem by monitoring the signal in
a live or real time waveform and turning off the power. The energy should
disappear immediately.
Open or Shorted Windings

The primary vibration appears in the spectrum at 2xLF and increases in
amplitude as motor temperature climbs.
A dominant cycle at a frequency associated with 2xLF is shown in the

waveform.
This problem can be confirmed by monitoring the live waveform and securing
the power to the motor. The energy should disappear immediately.
Insulation Breakdown and Unbalanced Phase

Both appear in the spectrum at 2xLF. The waveform demonstrates a dominant
cycle at a frequency related to the 2xLF.
This fault can be verified by monitoring the real time waveform and turning off
the power to the motor. The energy should disappear immediately.
Loose Connector
The primary frequency appears in the spectrum at 2xLF with sidebands spaced
at one-third line frequency.
The waveform shows a dominant cycle at a frequency identified with 2xLF.

This problem can be particularly difficult to identify if the loose connector is
intermittently making contact.
Electrical Defects
AC Synchronous Motor Analysis

A synchronous motor will have a running speed of exactly the magnetic field
frequency of the stator.
This is accomplished by designing the rotor as a fixed pole magnet.
This rotating magnet can be a permanent magnet but is usually an

electromagnet energized from an external direct current source through slip
rings.
The rotating magnetic field on the stator and the fixed magnetic field on the
rotor lock in and cause the rotor to turn at synchronous speed.
Unequal Pole Resistance

The primary vibration appears in the spectrum at 2xTS. The waveform
demonstrates a dominant cycle at frequency related to 2xTS.
This fault can be verified by monitoring the real time waveform and securing
the rotor field. The energy should disappear immediately.
The principal vibration appears in the spectrum at 1xTS.
A 2xLF peak or stator slot frequency with sidebands spaced at 2xLF may also
be present.
The waveform resembles that for imbalance, because the rotor is actually out of
balance.
In a live, real time display of the waveform, however, you can see the
electrically induced vibration disappear immediately when power is secured to
the motor. The signal disappearance helps you distinguish this fault from
mechanical imbalance.
Electrical Defects


a live, or real time waveform and turning off the power. The energy should

The primary vibration shows up in the spectrum at 2xLF and increases in

waveform. This problem can be confirmed by monitoring the live waveform
and shut off the power to the motor. The energy should disappear immediately.

Both appear in the spectrum at 2xLF.
The waveform demonstrates a dominant cycle at a frequency related to the

2xLF. This fault can be verified by monitoring the real time waveform and
turning off the power to the motor. The energy should disappear immediately.
Electrical Defects
Static Eccentricity
An eccentric stator will cause an unequal air gap between the rotor and the
stator. This results in the spectrum having a primary peak at 2xLF that is very
directional.
Slot frequency with sidebands spaced at 2xLF may also be present.
A poorly built stator, improper lamination, stamping, loose laminations, warped

motor base, or a soft foot are all possible causes of this defect.

2xLF. As with other electrical faults, you can confirm this defect by
monitoring the signal in a live, real time waveform and securing the motor.
The energy should disappear immediately.
Loose Connector
The primary vibration appears in the spectrum at 2xLF with sidebands spaced

Variable Frequency Drives For AC Induction Motors

The need to vary machine operating speeds is required in certain industrial
applications.
Variable Frequency Drives (VFDs) or Inverters are commonly used to control

the speed of an AC induction motor.
The operating speed of the motor is solely dependent on the output from the
VFD. The motor speed will vary directly as the output from the VFD increases
or decreases.
Electrical Defects
Broken or Open Rotor Bars

The predominant vibration appears at 1xTS with sidebands spaced at a
frequency equal to the number of poles on the motor multiplied by its slip
frequency.
Rotor slot frequency may also be present with sidebands spaced at 2x line
frequency. Please consider the following:
Slip Frequency = Magnetic Field Freq. - Rotor Freq.
Rotor Slot Frequency = # of Rotor Slots * Rotor Freq.
Stator Slot Frequency = # of Stator Slots * Rotor Freq.
# of Motor Poles = (Line Freq. * 120) / Magnetic Field Freq.
The waveform resembles that for imbalance, because the rotor actually is out of
balance.
In Monitor mode you can see the electrically induced vibration disappear
immediately when you turn off the motor power. The signal disappearance
helps you differentiate this fault from mechanical imbalance, where the
vibration would coast down with shaft speed.
Electrical Defects
Loose Iron or Slot

The primary vibration appears in the spectrum at 2xLF and slot frequency with
sidebands spaced at 2xLF. The waveform shows a dominant cycle of 2xLF.
The principal vibration appears in the spectrum at 1xTS with sidebands spaced
at slip frequency multiplied by the number of motor poles and/or and 2xLF.
Slot frequency may also be present with sidebands spaced at 2xLF.
The waveform resembles that for imbalance, because the rotor is actually out of
balance. In a live, real time display of the waveform, however, you can see the
motor power. The signal disappearance helps you distinguish this fault from
mechanical imbalance.


a live or real time waveform and turning off the power. The energy should

The primary vibration shows up in the spectrum at 2xLF and increases in

waveform. This problem can be confirmed by monitoring the live waveform
and securing the power to the motor. The energy should disappear
immediately.
Electrical Defects

Both appear in the spectrum at 2xLF.

2xLF. This fault can be verified by monitoring the real time waveform and
turning off the power to the motor. The energy should disappear immediately.
Static Eccentricity
An eccentric stator has an unequal air gap between the rotor and the stator.
This results a very directional spectrum with a primary peak at 2xLF. Slot
frequency with sidebands spaced at 2xLF may also be present.
A poorly built stator, improper lamination, loose laminations, stamping, warped

motor base or a soft foot are all possible causes of this defect.

2xLF. As with other electrical faults, this defect can be confirmed by
monitoring the signal in a live, real time waveform and securing the motor.
The energy should disappear immediately.
Loose Connector
The primary frequency appears in the spectrum at 2xLF with sidebands spaced

Variable Frequency Drive Defects

Open or Shorted SCR
The primary vibration appears in the spectrum at multiples of line frequency.
A dominant cycle at a frequency associated with line frequency is shown in the
waveform. Further electrical testing is recommended if this defect is
suspected. This could include jumpering out or replacing one set of SCRs at a
time.
Electrical Defects
DC Motor Analysis
A DC motor consists of a stationary component called the field.
The field windings create the magnetic North and South poles on the stator.
The rotating component of the DC motor is called the armature. The armature
winding current is supplied through brushes and creates the magnetic field on
the rotor.
A commutator is attached to the ends of the armature coil and is used to reverse
the current flow in the coil.
All DC motors operate on the principle of Motor Action which states: a

current carrying conductor placed in a magnetic field tends to move at right
angles to the field either up or down.
DC motors are classified by the way the main field is connected to the
armature. This will determine how the motor reacts to different types of faults.
Electrically testing the motor will identify most of the electrical faults
associated with the field and armature.

Commutator or Brush Problems
The primary vibration tends to show up in the spectrum at the (number of
commutator segments) x (number of brush rows) x (shaft turning speed ). A
dominant frequency associated with the commutator or brush defect is shown
in the waveform.
DC Drive Defects
Open or Shorted SCR
The primary vibration appears in the spectrum at multiples of line frequency.
A dominant cycle at a frequency associated with line frequency is shown in the
waveform. Further electrical testing is recommended if this defect is
suspected. This could include jumpering out or replacing one set of SCRs at a
time.
Electrical Defects
Case Histories
Boiler Feed Pump

Electrical Defect Case #1
Ô 300 HP, 3600 RPM AC Induction Motor.
Õ The motor supply breaker was intermittently tripping.
Ö Speed fluctuations of the motor were observed.
× 47 amp surges were measured with clamp on ammeter.
Electrical Defects
Boiler Feed Pump

The spectrum above comes from a boiler feed pump that was suspected of
having an electrical fault. The vibration data was collected after the stator was
tested and determined to be in satisfactory condition.
The cursor is marking the running speed peak of the motor. Notice the
skirt width of the running speed peak.
Also note that we are only looking out to 1.4 orders.
Electrical Defects
Boiler Feed Pump

When the spectrum from the previous page is expanded, the sideband around
motor turning speed is now more obvious.
This may have been identified as unbalance if a high resolution spectrum was
not taken. This demonstrates the importance of collecting high resolution data
when an electrical defect is suspected.
What could be the possible cause of the motor running speed sideband?
Electrical Defects
Boiler Feed Pump

A mark has been set on the motor 1xTS peak. The difference frequency of the
sideband is 0.66 Hz, which equals the number of motor poles (2) times the slip
frequency (0.34 Hz).
Slip Frequency = Magnetic Field Speed - Run Speed

= 60 Hz - 59.66 Hz
= 0.34 Hz
Sideband = Number of Motor Poles x Slip Frequency

= 2 X 0.34 Hz
= 0.68 Hz
Electrical Defects
Boiler Feed Pump

The sideband spacing of 0.66 Hz around 1xTS of the motor approximates this
frequency.
The presence of this sideband is a good indication of a rotor bar defect.
A current measurement analysis would confirm if a rotor bar problem exists.

In this case, the repair shop found five open rotor bars on this motor.
Electrical Defects
Case #2 Inverter Driven Induction Motor
MIV MOV
MIA
MOA
MIH MOH
PH1 PH2 PH3
Ô This 50-HP, four-pole motor is powered by an inverter, so the

line frequency is not necessarily at 60 Hz.
Õ The motor drives an exhaust fan in a painting booth.
Ö In addition to vibration readings, current measurements were

also collected on Phase 1 of the motor.
× This is a new motor. Other motors supplied at the same time did
not have these same vibration characteristics.
Electrical Defects
The spectrum below comes from the inverter driven motor.
Examine the spectrum to find the 1xTS peak at 26.1 Hz or 1566 RPM.
Note the high frequency energy around 1700 Hz. This energy first seems to
indicate a bearing fault.
Further examination shows a peak at 1732 Hz, which is 66 orders of TS. This
value is probably the rotor bar pass frequency, which means the rotor has 66
rotor bars.
The line frequency is 54.1 Hz, so 2xLF equals 108.2 Hz.
This spectrum shows very clear sidebands of 108.0 Hz around the peak at
1732 Hz.
Therefore, these peaks are sidebands of 2xLF around the rotor bar pass
frequency. These peaks indicate an electrical problem.
Electrical Defects
The line frequency appears at 54.1 Hz.
The current amplitude at turning speed approximates 41 dB amperes.
From the previous page, the motor turning speed is 26.1 Hz.
Because this is a four-pole motor, the magnetic field speed or synchronous

speed would be half the line frequency or 27.05 Hz.
Slip frequency ' Magnetic field speed & run speed
Slip frequency ' 27.05 Hz & 26.10 Hz
Slip frequency ' 0.95 Hz ( 57 RPM )
Electrical Defects
The sidebands in the current spectrum should occur at a spacing equal to the
number of motor poles times the slip frequency.
Sideband ' Number of motor poles × slip frequency
Sideband ' 4 × 0.95 Hz
Sideband ' 3.8 Hz
The sideband spacing of 3.9 Hz nearly equals this frequency.
The reason for examining the current spectrum in dB amperes is to see these
sidebands.
If the sidebands are less than 50 dB down from the line frequency peak, then
suspect a rotor bar problem. In this case, the sidebands at 50.2 Hz and 65.9 Hz
are less than 50 dB down.
Electrical Defects
Two-Pole Induction Motor
MIV MOV
MIA
MOA
MIH MOH
PH1 PH2 PH3
Ô This 50 HP, two pole induction driven motor operates on 60 Hz line

frequency power.
Õ The actual running speed of this motor is 59.54 Hz, or 3572.4 RPM.
Ö This example demonstrates the need for high resolution spectra when
troubleshooting electrical faults.
Electrical Defects
You need high resolution spectra to diagnose electrical faults.
The 400-line FFT below comes from the Route mode of the Model 2110/2115
Machinery Analyzer.
The 800 and 1600 line FFTs come from Acquire Spectrum in the Analyze
mode of the Model 2110/2115.
Electrical Defects
The bottom spectrum in the multiple-spectra plot above shows only broad
peaks at 1xTS and 2xTS.
The top two spectra indicate that many closely spaced peaks combine to form
those broad peaks in the bottom spectrum.
Electrical Defects
The spectrum above shows a 1600-line FFT.
Note that the 2xTS harmonic cursor is not on the highest peak.
Another peak, located at a frequency just above the 2xTS frequency, has a
much higher amplitude.
The plot on the next page expands the frequency scale of the spectrum above.
The expansion locates 120 Hz at the center of the horizontal (frequency) scale.
Electrical Defects
When you expand the plot on the previous page, the peak at 120 Hz shows up
more clearly.
This major peak is located at exactly 120 Hz or 2xLF. Note that the 2xTS
peak of the motor actually occurs at 119.1 Hz.
The sidebands are spaced at about 0.9 Hz, which equals the number of poles
on the motor (2) times the slip frequency (0.46 Hz).
slip frequency =magnetic field speed - run speed

= 60 Hz - 59.54 Hz
= 0.46 Hz (27.6 RPM)
Electrical Defects
The sidebands around the 1xTS peak should be spaced at a frequency equal to
the number of poles on the motor times the slip frequency.
sideband = number of motor poles x slip

frequency
= 2 x 0.46 Hz
= 0.92 Hz
The sideband spacing of 0.9 Hz around 1xTS of the motor approximates this
frequency.
Finding these sidebands might cause you to suspect a defective rotor bar.
Electrical Defects
The same portion of the spectrum on the previous page is now plotted in units
of dB velocity.
When the sidebands caused by the possible rotor defect are less than 20 dB
down from the actual run speed peak, then a significant problem may exist.
Further work with a current probe is needed to examine the electrical

characteristics of the motor.
Use the data taken with vibration transducers as a first step in determining
whether the machine has an electrical or mechanical problem.
In this case, the motor appears to have an electrical problem.
Note: To change amplitude to dB’s, from diagnostic plotting, go to

“Global Options”, Engr. Units, change Units Mode to desired dB
units.
Electrical Defects
Process Chiller Pump

Ô 15 HP, 1800 RPM AC Induction Motor.
Õ High motor vibration on the motor horizontal points.
Ö Motor vibration at 4.1xTS.
Electrical Defects

The multiple plot above comes from a case history involving a direct-driven
chiller pump.
The unit running speed peak at 28.75 Hz (1725 RPM) is not even seen in the
plots.
Note, the dominant nonsynchronous peak at 4.17 orders in the horizontal

direction.
The spectrum for the motor inboard horizontal appears on the next page.
Electrical Defects
The amplitude of the 4.17 order peak in the spectrum is 0.7 IPS. The
waveform shows some modulation and a definite sinusoidal pattern related to
the vibration. What could be the source of this directional, nonsynchronous
vibration? The next page shows a high resolution spectrum that confirms the
suspected problem.
Electrical Defects

The located spectral peak is at 4.17 orders or 119.9 Hz.
The vibration at 119.9 Hz (120 Hz) or 2xLF is the problem that needs to be
corrected.
Upon investigation, it was found that the directional 2xLF vibration was
caused by a soft foot condition on the motor.
Electrical Defects
Fan Pump Motor

Ô 450 HP, 1200 RPM Inverter Driven AC Induction Motor.
Õ Motor was running hot and it was replaced.
Ö New motor had the same vibration as the one it replaced.
× Current readings were: 93, 95, and 75 amps.
Electrical Defects
Fan Pump
The trend plot for the motor’s 3 to 4xTS band is shown above. This case
history is from an inverter driven fan pump.
Notice the large increase in amplitude from the data collected in December.
The next page shows the spectrum that was collected in December.
Electrical Defects
Fan Pump
The trend data clearly shows that something changed in December.
Spectral data from December shows the highest vibration peak at 59.99 Hz.
Setting a harmonic cursor on 59.99 Hz reveals the presence of harmonic peaks.
An electrical defect should be considered since the motor is powered from a

variable frequency drive.
Electrical Defects
Fan Pump
Troubleshooting was performed to determine the source of the harmonics of

line frequency.
This spectrum was taken with one of the boards containing a pair SCRs
jumpered out.
Notice that the amplitude of the 59.99 Hz peak went from 0.48 IPS to 0.003
IPS. It was decided that a defective SCR in the variable frequency drive was
the source of the 1xLF vibration.
Electrical Defects
Synchronous Band Motor

Ô 500 HP, 900 RPM AC synchronous motor.
Õ The motor was sent in for repair due to high vibration.
Ö High vibration was measured at 2xTS.
× The 2xTS vibration was primarily in the horizontal direction.
Electrical Defects

The multiple plot above shows data from the synchronous motor before repairs
were performed.
All the points on the motor show a dominant peak at 2xTS. Troubleshooting
was begun to determine the source of the 2xTS vibration.
Electrical Defects

The Motor Outboard Horizontal (MOH) spectrum and waveform are shown
above. Spectral data shows the dominant 2xTS vibration with no harmonics.
As expected, the waveform is repetitive and lacks complexity. Notice the low
amplitude level in the waveform.
Electrical Defects

After extensive troubleshooting, the defect was isolated to the rotor.
Individual rotor pole resistances were taken and some variances were
discovered.
The spectrum on this page shows the MOH point after the rotor was rewound.
Notice the large reduction in the vibration at 2xTS.
Section Fourteen
Belts
Objectives
< List four causes of belt vibration.
< Describe the vibration characteristics of each belt defect.
< Measure the belt vibration in the most effective locations and directions.
Introduction
Belt defects are considered non-critical faults by many maintenance groups due
to the relative ease of replacement requiring minimal downtime. However, belt
drive system defects can have a significant effect on the reliability of the
equipment. Belt defects themselves are a major contributor to the overall
vibration environment of the individual components. Excessive belt tension
provides additional load to the component bearings and under tensioning leads
to operational performance problems. One main factor affecting component
and belt reliability and longevity is proper sheave alignment.
Belts
Defect Characteristics
Belt Faults
Belt faults, such as cracks, broken or missing pieces, hard or soft spots, lumps
on belt faces, or a crooked belt which has set to a deformed shape, can generate
vibration at the turning speed of the belt (1xBelt) and harmonics.
The predominant harmonic is typically the 2xBelt or the harmonic equal to the
number of sheaves in the system, i.e., 3xBelt for a 3 sheave system.
The predominant plane is typically radial and in-line with the belts. Due to the
length of the belt in relation to the sheaves, the 1xBelt frequency is
subsynchronous and often the 2xBelt may be subsynchronous as well.
Severity is judged by the number and amplitude of the harmonics seen in

the spectra.
The belt frequency can be calculated from the following equation:
Belt Freq = (3.142 * Shaft TS * Pulley Pitch Diameter)/(Belt Length)
Note: The Pitch Diameter and the Belt Length must be in the same units.
The belt length may be known or it can be measured or calculated. If it must

be calculated, the following equation can be used:
Belt Length = 2C * 1.57 * (D + d) + (D - d)2/4C
Where,
C = center-to-center distance between sheaves

D = pitch diameter of the large sheave
d = pitch diameter of the small sheave
Note: All the length dimensions must be in the same units.
Belts
A timing belt will have a specific frequency related to the belt frequency and
the number of teeth on the belt and the pulley.
Timing Belt Freq = (Belt Freq) * (# Belt Teeth)
or
Timing Belt Freq = (Pulley TS) * ( # Pulley Teeth)
Remember, the belt frequency and harmonics appear when the belts are worn,
loose or mismatched.
The defect amplitudes may increase and decrease relative to the driver or
driven pulley’s turning speed.
Belts
Misaligned Sheaves/Sheave Runout

Belt misalignment is one of the primary reasons for reduced belt life.
If the misalignment is greater than 1/16 of an inch for each 12 inches of

belt—center to center distance—very rapid wear will develop.
Very often, sheaves are out of line because the motor and the machine shafts
are not parallel.
Sheave misalignment results in high axial vibration at shaft turning speed. If

belt defects are also present, belt defect vibration will also become more
evident in the axial plane.
Misaligned sheaves show high vibration at 1xTS vibration predominantly in the

axial direction. Whether the driver or driven 1xTS peak is dominant is
dependent on the mass and stiffness of each component.
Not always, but often with sheave misalignment the highest axial vibration on
the motor will be at the fan turning speed and the highest axial vibration on the
driven unit will be at motor turning speed. This is referred to as, Cross Over
effect. Belts that are adjusted too tight can also cause this situation.
Also 3xTS and 4xTS multiples of the belt frequency may show up in line with
the belts and in the axial direction.
A wobbling sheave whips the belt from side to side subjecting it to excessive
lateral strain. This can lead to higher axial 1xTS vibration.
Belts
Eccentricity/Imbalance
If the sheave is true but unbalanced, vibration occurs at shaft turning speed
radially. This relates to an unbalance problem which will show up at 1xTS
frequency.
If the sheave is balanced but has excessive runout, symptoms of inadequate

tensioning become evident but at low amplitudes.
Runout also tends to generate excessive belt-flapping and adds to the radial
turning speed amplitude of the defective shaft.
High vibration will show up at 1xTS as the sheave literally has a high spot that
affects the belt tension once per revolution of the sheave.
Belts
Belt Resonance
Every system has resonant frequencies and belt systems are no different. There
are two types of belt resonance problems.
Ô Belt frequencies exciting other system resonances.
Belt defects can generate many different frequencies that can excite system
resonance, such as turning speed and harmonics of both the driver and the
driven shaft. Also, belt frequency and multiples of belt frequency. This may
be the case when one of the multiples of a belt defect has a high amplitude
compared to the other belt frequency harmonics.
Õ Belt Strand Resonance (BSR) being excited by other system

frequencies.
The belts also have a system resonance of their own. This frequency is often
referred to as the Belt Strand Resonance. The BSR is related to the portion of
the belt suspended between the two sheaves. The frequency of the BSR is
dependent on the mass, stiffness, damping properties of the belt as well as the
belt tension. This is similar to the properties of a violin or a guitar string.
The BSR can typically be determined by an experimental measurement. With

the machine off, the belt can be plucked and the resonant frequency can be
measured. The analyzer setup will be similar to that for an impact test. The
accelerometer can either be taped to the belt or placed at the sheave bearing.
The BSR should be adjusted if it will be excited by one of the operational

frequencies of the machine train. Adjustments could be made by altering the
belt length, the tension, or by placing an idler pulley in the system to effectively
change the belt length.
Belts
Data Collection Considerations

Vibration Data
Due to the proximity of the 2xBelt frequency to the shaft turning speed
frequency (1xTS), the ability to separate the two peaks for analysis is extremely
important.
Use a high resolution measurement point on each shaft in the in-line position
for effective analysis of belt-driven systems.
For most belt-drive systems, using a normal data collection resolution (800
lines) is usually sufficient with an Fmax of 3xTS of the high speed shaft.
However, each system should be evaluated individually to ensure adequate

separation of belt defect harmonics from other peaks. Caution should be taken
with your accelerometer mounting technique to allow the 1xBelt to be viewed
without interference from integration noise.
This high-resolution data may be collected in addition to normal route-

collected data in the analyze mode based on field observations. The most
effective method for trending and fault progression analysis is to make these
measurements part of your normal route data collection.
The best measurement directions to detect belt defects and to see the highest
amplitude vibration is in line with the belt tension (horizontal) and
perpendicular to the belt tension (vertical).
Belts
Physical Observations
An extremely effective technique for analyzing belt system faults is to use a
strobe light in conjunction with your data collection.
Belt defects are often visible when the belts are frozen.
Sheave misalignment is also typically visible when the sheaves are frozen.
Applying a reference line across the belts during installation will provide a
comparison point for determining the amount of slippage in the system.
Summary
This section has covered the basic belt defect causes and the frequencies that
can be generated from these frequencies.
Defects related to problems with the belt itself can generate 1xBelt and
harmonics.
2xBelt can be dominant because of the two sheaves that a belt defect must pass
over.
Other problems, such as belt misalignment and sheave runout can lead to a
higher 1xTS in the axial direction.
A sheave can be well balanced but still generate a 1xTS in the radial direction
due to the sheave being eccentric (out of round). This causes uneven belt
tension effects once per revolution.
The belt strand resonance can be a significant problem that is often overlooked.
Generally, the BSR must be measured in order to know if it can be excited by
some other system operating frequency. A belt drive system may rapidly fail if
this frequency is excited.
Belts
Case Histories
Belt Case 1
This case history typifies data for misaligned sheaves. The data shows a severe
condition affecting the motor measurement points.
Allowing this condition to progress would lead to premature bearing failure.
Identification and correction of this condition increases both the reliability and
efficiency of the motor.
Belts
Belt Case 1
The above multiple-spectra plot shows data from the Motor Inboard
measurement points.
Notice the highest amplitude appears at 1xTS and the predominant plane is the
axial reading.
Very little vibration is seen at harmonics of turning speed indicating that this
problem is in its early stages and has not caused any looseness to develop.
With a plot span of 1.0 in/sec, it is difficult to determine whether bearing

defects have begun to appear as a result of the misalignment condition.
Belts
Belt Case 1
The spectrum for the Motor Inboard Axial (M2A) is shown above.
Most of the overall vibration is due to the energy located at 1xTS and further
analysis would be required to determine whether other problems have been
driven by this condition.
Belts
Belt Case 1
As suspected, the waveform is showing a fairly nonimpacting, sinusoidal

pattern with one event per cycle. The amplitude is fairly high and helps verify
the source of the 1xTS vibration as misaligned sheaves.
Belts
Belt Case 1
Expanding the frequency axis allows us to determine the condition of the belts.
The fault frequencies are displayed for the primary belt frequency with
harmonics.
Notice the primary belt defect frequency (1xBelt) does not appear in the
spectrum, possibly due to the amplitude scale.
The 2xBelt frequency may be adding to the 1xTS peak amplitude. Notice that
the skirt width of the 1xTS peak includes the frequency for the 2xBelt.
The occurrence of the wide-skirted peak should cause the analyst to consider
whether the peak is made up of two or more frequencies added together.
Obtaining high resolution spectral data would help to separate the 1xTS from
the 2xBelt frequency and allow us to determine the condition of the belts.
Belts
Belt Case 2
This case history examines the need for resolution for defect analysis and the
identification of the need for additional data during normal route data
collection.
Belts
Belt Case 2
The multiple-point spectral plots show route-collected data for the

measurement points on the motor.
Notice most of the energy appears to be around 1xTS with the radial points
showing the higher amplitudes.
Measurement point M1H, Motor Outboard Horizontal, appears to have the

most energy and is displayed on the next page.
Belts
Belt Case 2
The full-screen view of M1H autoranges to 1.6 in/sec and does not provide much
detail of the lower amplitude frequencies.
Viewing this data in the field may lead you to consider the high amplitude peak as
turning speed.
The cursor on the main peak identifies the peak frequency is actually about
0.8xTS.
The person collecting the data recognized this may be due to a belt defect and
acquired additional data as shown on the next page.
Belts
Belt Case 2
Notice the wide-skirted peak in the spectrum from the route spectral setup at
08:43.
The 2xBelt frequency and 1xTS cannot be separated in this spectrum.
The analyst suspected a belt problem and acquired additional high-resolution

spectra with a lower Fmax and higher lines of resolution (LR).
The spectrum from 08:46 had an Fmax of 500 Hz and 3200 LR.
The spectrum from 08:47 had an Fmax of 200 Hz and 3200 LR.
The following pages show the analysis of the high-resolution spectrum from
08:47.
Belts
Belt Case 2
The spectrum above displays the belt defect fault frequencies for the belt defect.
Notice the 2xBelt frequency is the dominating peak in the spectrum.
The overall value for the spectrum is 1.38 in/sec and the amplitude of the 2xBelt
frequency is 1.2 in/sec.
Most of the overall energy is a result of the 2xBelt frequency, not 1xTS.
Obtaining this data in the normal route mode would simplify the data collection
process and allow this fault to be accurately identified during the data collection
process.
Having trend data of this machine would then allow the analyst to determine the
rate of progression more accurately and provide valuable information for
determining the recommended maintenance time frame.
Belts
Belt Case 2
The cursor marks the location of the 1xTS peak.
Without the resolution available in this spectrum, the 1xTS and 2xBelt frequency
may actually appear as the same peak.
The belt problem may have been incorrectly diagnosed as unbalance when the
1xTS amplitude was actually acceptable.
Belts
Belt Case 2
As expected, the associated time waveform shown above shows a strong

sinusoidal pattern.
At first look, it appears to have a regular one event per cycle.
Notice that the events do not occur at regular intervals because the predominant
vibration is due to the 2xBelt.
There also appears to be some modulation resulting from the minor amount of
vibration at 1xTS.
The modulation occurs at a frequency equal to the difference between 1xTS and
2xBelt.
Belts
Belt Case 2
The spectrum shown above was collected after the belts were replaced.
Notice the amplitude of the 2xBelt frequency is still fairly high but has dropped to
about 0.09 in/sec and the overall level has decreased to 0.1749 in/sec.
Although the immediate problem was corrected, 0.09 in/sec at 2xBelt should not
be considered acceptable.
This case history not only allows us to identify that a defect exists, but also
contains a probable cause for the belt defect.
Belts
Belt Case 3
The multiple-spectra plot shown below indicates that all measurement points have
skirted peaks at what appears to be turning speed.
The Motor Inboard Axial (M2A) measurement point appears to have the highest
amplitudes and we will examine it further.
Belts
Belt Case 3
The multiple-trend plot from point M2A shows that the overall amplitude has
increased significantly over the last reading.
The trend of the SUB and 1xTS band has seen the same amplitude increase. The
2xTS band has increased only slightly.
Belts
Belt Case 3
The single spectrum displayed above shows a mound of energy around 1xTS.
The frequency scale needs to be expanded to see what may be occurring there.
Note that the low amplitudes of the other frequencies in the spectrum may be due
to the amplitude scale (0-0.7 in/sec).
Belts
Belt Case 3
The spectrum displayed above is an expanded view of the plot around 1xTS.
The fault frequencies are displayed for the belt defect.
Notice the 2xBelt and 4xBelt frequencies are clearly seen in the spectrum.
However, the 1xTS frequency is the dominating frequency.
As with the previous case, the 1xTS skirt width may provide an indication that
there are two closely spaced peaks appearing as a single peak.
This spectrum does not have enough resolution to separate the two peaks.
The dominance of the 1xTS peak in the axial plane proves this is a misaligned
sheave problem and the harmonics of the belt defect frequency axially indicate
that the belts are suffering from the alignment condition.
The sheave misalignment has damaged the belts.
Belts
Belt Case 3
The waveform shows one event per cycle as expected. However, the time span is
limited and we are unable to identify any modulation due to the belt fault.
Belts
Belt Case 4
Belts
Belt Case 4
In this multiple points plot we can see the amplitude of the vibration in the Axial
direction.
Belts
Belt Case 4
Note the high amplitude at 1 x Turning Speed, 1.97 in/sec in the axial direction.
Data collected 19th Dec 1997.
Belts
Belt Case 4
Compile report was written as follows:
Belts
Belt Case 4
It was discovered that both sheaves where worn and needed to be replaced. The
sheave was replaced on the Motor, but not on the Gearbox. New belts where
installed and straight edge alignment was performed. As displayed below the
vibration was reduced 10 times.
Belts
Belt Case 4
This data shows the improvement achieved on the system. The plot at the top is
after the repairs were made. The plot on the bottom is before repairs.
Section Fifteen
Resonance
Objectives
< Define what resonance is and how resonance affects machine vibration.
< Understand the three parameters that affect the amplitude and frequency
of the resonance.
< Recognize the single-channel analysis tools available for diagnosing

resonance as well as the limitations.
What is Resonance?
Resonance occurs naturally in all machines, structures, and even animals.
Often resonance is called natural frequency or critical frequency.
Resonance is a frequency at which vibration in a structure is amplified.
Several factors work together allowing resonance to occur, such as low

stiffness and/or low damping at the resonant frequency. Resonance is not
necessarily a problem unless machine defects create vibration or nearby
machinery transmits vibration at the same frequency as the resonant frequency.
Resonance does not create vibration; it only amplifies it.
Resonance
Resonance is not itself a defect, but it is a property of the whole mechanical

system. Therefore, resonance will not be a problem unless some other
vibration excites the resonance. This is just like a bell or a drum. The bell
possesses the potential to resonate at a particular frequency (or frequencies) if it
is excited by being struck. If it is not excited, then it does not vibrate at its
resonant frequency.
Resonance is a property of the machine whether running or not. Be aware that

the dynamic shaft stiffness when the machine is running may be different
enough from the at-rest static stiffness to cause the resonance to vary slightly.
The rule of thumb has always been that a resonant frequency measured with
machinery shut off should be at least 20 percent away from any forcing
frequency.
Individual parts of a machine have resonant frequencies such as shafts, rotors,

casings, and foundations.
When these machines are assembled these resonant frequencies shift because of
the mass, stiffness, and damping effects that occur when the machine is put
together.
Also, dynamic stiffening effects, as mentioned earlier, may shift the static
resonant frequencies when the machines are running at their operational speed.
Most machines are designed so that the rotor does not have the same resonant
frequencies as the shaft.
Once the two are assembled, the machine is not designed to run at the resonant
frequency. However, as machines wear and clearances change, resonances
often shift close to operating frequencies.
An unexpected defect frequency, such as a harmonic of looseness or some

other machine defect, may excite a resonant frequency.
Machinery vibration at acceptable levels can be amplified to unacceptable

levels if the vibration excites a resonant frequency.
For example:
A two-speed machine runs at 900 RPM and 1200 RPM. The machine has a
resonance at 1200 CPM which amplifies the 1xTS vibration. At 900 CPM, the
1xTS vibration is 0.1 IPS and 1200 CPM and the resonance causes the 1xTS
vibration to increase to 0.5 IPS.
Resonance
Resonance may be observed as a machine starts up and runs through a resonant

frequency.
The amplitude of the 1xTS will increase to a maximum amplitude at the

resonant frequency and decrease as it passes through the resonant frequency.
The phase will go through a 180E change as the shaft passes through resonance
with a 90E phase shift occurring at the resonant frequency.
The 180E phase shift often only occurs on simple single-plane types of rotors.
More complex shaft/rotor systems exhibit a phase shift, although not 180E.
Whenever an analyst is diagnosing a high amplitude vibration problem, the

analyst needs to consider the possibility of acceptable vibration exciting a
resonance and causing unacceptable levels of vibration.
A
M
P
L
I
T
U
D
E
FREQUENCY
fr
P
H
A
S FREQUENCY
E
Resonance
Mass, Stiffness, and Damping

Mass, stiffness, and damping are the three parameters that affect the frequency
and amplification of a resonance.
Mass is the property that creates the vibration forces.
Stiffness is the property that counteracts the inertia due to the mass forces.
Damping is the property that converts mechanical energy to thermal energy,

actually absorbing the vibration.
Resonance
Increasing the mass of a structure will decrease the resonant frequency.
Increasing the stiffness of a structure will increase the resonant frequency.
Resonance
Increasing the amount of damping in a structure will decrease the amplitude of

the resonance.
At resonance, damping is the only property that controls the amplitude of the
vibration.
The added damping also reduced the resonant frequency slightly.
If adding mass will decrease a resonant frequency, then taking mass away will
certainly increase the resonant frequency. Likewise, if adding stiffness
increases a resonance then decreasing the stiffness will decrease the resonant
frequency.
An analogy could be made to a guitar string.
The tighter the string ( the more stiffness) the higher the tone (resonant
frequency) that the string will produce when it is plucked. If a thicker string is
used (more mass), the lower the tone will be.
Resonance
Measuring Resonance
One of the commonly used methods for measuring a structure’s resonant
frequencies is to impact the structure with a hammer (not a steel hammer) and
measure the response.
This works because the impact inputs a small amount of force into the machine
over a specific frequency range.
When performing this technique, it is important to try impacting different

locations on the structure since all a structure’s resonant frequencies will
always be measurable by impacting at one location and measuring at the same
location. Both drive-point and transfer-point measurements should be tried
when attempting to identify machine resonances.
If the hammer has a soft tip, more energy will be input at lower frequencies.
Harder tipped hammers place less energy at each frequency, but allow energy
to be input at higher frequencies.
Only the response to the impact is being measured with a single-channel

analyzer, so the machinery must be shut off to do this properly, unless the data
collection is being performed with negative linear averaging.
Resonance
Phase is one of the parameters that can clearly

identify resonant frequencies.
However, phase cannot be measured during an impact test with a single-
channel analyzer; and, in order to confirm the presence of a shaft resonance,
the phase must be collected using other methods.
When using a single-channel analyzer, measuring amplitude and phase of 1xTS

during a coastdown (or start-up) will identify a resonant frequency as an
amplitude increase at the resonant frequency and a phase change of 180E from
below resonance to above resonance.
This allows the analyst the two pieces of information needed to identify a shaft
resonant frequency. If the phase does not change, a structural resonance and
not a shaft resonance is probably being measured.
In either case, it is recommended that a “bump test” be performed in addition

to the startup/coastdown data.
Resonance is most definitively measured using a multichannel analyzer to

measure impact and response data at the same time, thus allowing for phase
and coherence data to be collected at the same time.
Coherence is a dual-channel function that relates how much of the input signal
caused the output signal. This means that resonance frequencies can be
identified more rapidly.
Resonance
Machinery Considerations
Some consideration needs to be given to different types of machinery and some
conditions that may make resonant testing difficult. Due to structural
differences between the horizontal and the vertical directions, the resonant
frequencies may vary slightly from the horizontal to the vertical direction. So
resonances can be directional.
As discussed earlier, resonant frequencies may vary slightly as machinery is

running or stopped. Vertical equipment is generally very troublesome because
the equipment, being vertical mounted, has a resonance resulting from the
cantilever action of the motor.
Some machines are too massive to excite with a hammer and require alternative
means to excite the resonant frequencies. Sometimes on very large structures
using a dynamic shaker setup to vibrate through a frequency range is used
because shakers are able to place more energy in at each frequency.
One last consideration — before performing resonance testing it is very helpful

to measure the background vibration levels. This can prevent misdiagnosing a
large amplitude frequency in the background data as a resonant frequency that
may need to be corrected.
Summary
This section has discussed the effect of resonant frequencies on machinery
vibration. All structures have resonant frequencies, but the resonances will not
affect a machine unless there are frequencies present that excite the resonances.
If machine vibration excites a machine’s natural frequency, three options exist.
Ô Move the forcing frequency away from the resonant frequency.
Õ Move the resonant frequency away from the forcing frequency.
Ö Add damping to reduce the amplification factor of the resonance.
Options 2 and 3 generally involve some structural design changes that should
not be made unless a modal analysis and/or a finite element analysis study has
been performed on the structure.
Resonance
Case Histories - Resonance
Process Vacuum Blower

Resonance Case #1
MIH
MIV
BIH
BIV
Õ Unit a direct-driven overhung fan.
Ö Directional vibration was present.
Resonance
Resonance Case #1
The above multiple-point plot shows the spectra from the motor and fan
inboard points. Notice the large difference in the horizontal and vertical 1xTS
amplitudes.
The points most affected are the points on the fan.
Resonance
Resonance Case #1
The full-screen spectrum from the blower inboard horizontal is displayed.
The cursor shows the 1xTS amplitude to be over 0.7 IPS. That is
approximately ten times the vertical amplitude.
The next page shows the results of the impact test performed on the blower
housing.
Resonance
Resonance Case #1
The impact spectrum clearly shows the highest peak is at motor turning speed.
This resonance of the housing amplifies the turning speed vibration.
The waveform for this impact is displayed on the next page.
Resonance
Resonance Case #1
The display above shows the waveform for the housing impact. The flat line
before the data represents the pre-trigger setup in the analyzer.
Normally set to 35% pre-trigger.
Resonance
Process Pressure Pump

Resonance Case #2
Ô 20-HP, 3600 RPM motor.
Õ Unit is a direct-driven overhung pump.
Ö High directional vibration.
× The pump impeller has three blades.
Resonance
Resonance Case #2
The multiple-point plot above shows the motor and pump data.
The highest vibration is on the pump horizontal points.
The cursor shows the peak to be located at 3xTS.
A multiple plot showing just the horizontal pump points is on the next page.
Resonance
Resonance Case #2
The pump horizontal data is displayed in the above plot.
The cursor marks the 3xTS peak on the pump outboard horizontal.
The large differences between the horizontal and vertical readings suggest a
possible resonance problem.
Remember the Rule, when there a ratio of 3 to 1 or greater when comparing the
Horizontal reading to the Vertical reading, suspect resonance.
Resonance
Resonance Case #2
The spectral and waveform data are presented above. The analyst used an
accelerometer to collect the impact response. Notice the spectrum shows the
resonance peak more clearly in units of acceleration.
Resonance
Resonance Case #2
The data above shows the after repair vibration on the horizontal pump points.
The three-bladed impeller was replaced with an impeller that had five blades.
This moved the forcing frequency caused by the pump’s blade pass to above the
pump casings resonant frequency.
Resonance
Reboiler Feed Pump

Resonance Case #3
Õ Unit is a direct-driven overhung pump.
Ö High directional vibration is present.
Resonance
Resonance Case #3
The multiple-point plot from the pump shows the vibration is primarily at 1xTS
and in the horizontal direction.
The next page displays the Pump Inboard Horizontal (PIH) measurement point.
Resonance
Resonance Case #3
The cursor marks the amplitude of the 1xTS peak at .31 IPS.
This spectrum has all the characteristics of a balance problem.
If this was unbalance, the amplitude of the 1xTS peaks in the vertical direction
should be higher.
The directional vibration is caused by resonance, but the excitation force is

different.
This resonance was caused by improper mounting of the pump. In this case,
one of the pump foot bolts was stressing the casing because its hole was drilled
at an angle in the base.
Resonance
Aerator
Resonance Case #4
AY AX
BY
BX
Õ Gearbox has two intermediate shafts.
Ö High directional vibration is present.
Resonance
Resonance Case #4
A multiple-point spectrum shows the presence of high vibration at 1xTS in the

radial Y direction.
The vibration at shaft turning speed is not present in the axial measurement.
Resonance
Resonance Case #4
A full-screen spectrum of the upper motor bearing’s radial Y position shows the
1xTS peak is almost 1.5 IPS.
This directional flexibility is not uncommon in vertical equipment due to their

increased sensitivity to resonance.
Resonance
Resonance Case #4
The waveform shows a repeatable, sinusoidal pattern that lacks complexity.

The amplitude is relatively low and is free of impacts.
Resonance
Effluent Pump
Resonance Case #5
AX
AY
BY
BX
Ô 60 HP, 3600 RPM, motor.
Õ Vertically mounted well pump.
Ö Directional vibration is present.
Resonance
Resonance Case #5
The multiple-point plot for the effluent pump is displayed above. The cursor
marks the running speed peak at 1 order.
The amplitude of this peak is over 2 IPS.
Resonance
Resonance Case #5
The full-screen spectrum for the upper motor bearing point AX is displayed.
The overall vibration level for the plot is 2.05 IPS and the 1xTS peak
comprises almost all this value.
Resonance
Resonance Case #5
The waveform shows the sinusoidal pattern with the one distinct peak per shaft
revolution.
This repetitive shape, however, lacks any real complexity.
Resonance
500 HP Vertical Pump

Resonance Case #6
Ô 500-HP, 1200 RPM motor.
Õ Motor was just refurbished.
Ö Data was taken for certification.
× High directional vibration was present.
Resonance
Resonance Case #6
The certification data for the motor is displayed on the multiple point plot.
The cursor marks the 1xTS peak on the measurement point BH.
Again, the high vibration is only present one direction.
Resonance
Resonance Case #6
The data for the upper motor bearing is shown with a cursor marking the 1xTS
peak.
The analyst inquired about the support structure the unit was mounted on. He
was told the pump was mounted on an I-beam support that could move in the
direction of the vibration.
Additional support in the horizontal direction was recommended to reduce the

vibration.
Resonance
Resonance Case #6
This multiple-point plot for the motor was collected after temporary supports
were installed.
The cursor marks the 1xTS peak for the lower motor bearing’s measurement
point. Remember the initial vibration level was 0.5 IPS.
The supports reduced the vibration by a factor of three.
Resonance
Resonance Case #6
The multiple-spectrum plot for measurement point AH is displayed showing

the three readings that were taken.
December’s data shows the unit after the motor was repaired.
The amplitude of the 1xTS peak was .9 IPS.
The data for March shows the same measurement point after the supports were
permanently installed.
The amplitude of the 1xTS was reduced by a factor of seven to a value of 0.13
IPS.
Section Sixteen
Pump Vibration Basics
Objectives
< Describe the primary vibration generation principles that cause
vibration in centrifugal pumps.
< List vibration characteristics that point to pump defects.
Fundamentals of Centrifugal Pumps

Centrifugal pumps are commonly used in industry for fluid movement. Pumps
can exhibit typical mechanical vibration generated by defective bearings,
mechanical unbalance, looseness, etc. This section is dedicated to vibration
related to the hydraulic forces occurring within the pump and the piping
system.
Vibration is caused by the interaction between the moving pump impeller and
the stationary parts of the pump such as the volute and the diffuser vanes.
Also, vibration is caused by the interaction between the impeller blades and the
fluid being pumped.
Pump Operation
In a centrifugal pump, a rotating element called an impeller is enclosed in a
case. The fluid to be pumped enters into the case through the suction piping
into the impeller and is forced out the discharge side of the pump by means of
centrifugal force pushing the fluid out radially through the impeller. Several
components within the pump can generate vibration. Some of the vibration is
broadband and some is discrete.
Impeller Basics
Single-stage pumps use one impeller to create the necessary head and flow.
Pumps with more than one impeller are called multistage pumps. Multiple
impellers are used when the system requires more head than one stage can
provide.
Different types of impellers are used, depending on the type of flow required.
Ô Radial Flow Impellers - the fluid enters the impeller radially.
Õ Axial Flow Impellers - the fluid enters the impeller axially

(through the eye of the impeller).
Ö Mixed Flow Impellers - a combination of Radial and Axial Flow.
× Single Suction Impellers - the fluid enters the impeller from only
one side.
Ø Double Suction Impellers - the fluid enters the impeller from both
sides.
Other characteristics affect the impeller design such as the enclosure of the
impeller and the orientation of the shaft (vertical or horizontal).
Pump Casing
The pump casing forms the pressure boundary around the fluid. The pumped
fluid needs to have some of its velocity energy converted to pressure energy.
Typically, this is done through the use of a volute or diffuser vanes.
The casing can also be described by the way it must be separated for impeller
access— radial or axial split casing.
Volute Design
A volute is an ever widening spiral casing surrounding the impeller that reaches
its greatest area at the pump discharge. The increase in area converts some of
the velocity energy to pressure energy. Some pumps have twin volutes that
reduce excessive flow turbulence caused when the pump operates away from
its best efficiency point (BEP).
Diffuser Vanes
These are stationary vanes at the discharge of the impeller that convert some of
the velocity energy into pressure energy. No volute casing is needed when the
diffuser vanes are used.
Diffuser vanes work well only if the pump is operating near the design point so
that the angle that the fluid is leaving the impeller vanes is the same angle that
it is entering the diffuser vanes. If the pump is not operating within 5 to 10
percent of its BEP, then the flow path can become turbulent.
Net Positive Suction Head (NPSH)

NPSH is a measure of the difference between the total suction head and the
fluid vapor pressure. A fluid will flash to vapor if the pressure that the fluid is
under drops below the vapor pressure.
If the suction pressure of the pump does not exceed the vapor pressure of the
fluid, then fluid will change to a vapor and the pump will cavitate. Pump
cavitation causes excessive vibration and noise and will eventually destroy the
pump impeller and cause the pump to fail.
There is a difference between the available NPSH and the required NPSH:
Ô Available NPSH is the difference between the absolute suction

head and the vapor pressure of the fluid at its operating
temperature. This is a characteristic of the pumping system.
Õ Required NPSH is the smallest margin necessary between the

pump suction head and the vapor pressure. This is a characteristic
of the pump design.
Pump Vibration Analysis

Vane Pass Frequency
When most people think of pump vibration, they think of vane pass as one of
the pump’s primary defect frequencies. Vane pass frequency is not always a
defect frequency, but actually is an operational frequency. It is normal to see
this frequency in pump vibration data.
The amplitude of this frequency may increase if the pump is operating away
from the BEP. Also, higher amplitude vibration may be present if there is not
an equal gap between the rotating impeller and the pump’s stationary parts,
such as the diffuser vanes.
There are two different vane pass frequencies depending on the pump design.
Ô Impeller Vane Pass Freq. = (# Impeller Vanes) * (Shaft Turning

Speed)
Õ Diffuser Vane Pass Freq. = (# Diffuser Vanes) * (Shaft Turning

Speed)
The diffuser vane pass is only possible if the pump has diffuser vanes.
Cavitation
Cavitation occurs when the pressure of the fluid drops below the vapor
pressure for the temperature of the fluid. When this pressure drop occurs,
whether it is a system pressure drop or a localized pressure drop, voids or
cavities (bubbles) will form in the liquid. These bubbles implode or collapse
when the fluid moves through impeller to the high pressure side of the pump,
causing the impeller to erode. This can be very damaging and eventually the
impeller will fail.
Vibration due to cavitation is typically seen as harmonics of turning speed up to

the impeller vane pass frequency. Higher frequency broadband noise can also
be measured as well as heard by the analyst. A steady crackling sound (often
compared to popcorn or gravel moving through the impeller) is present.
The broadband vibration is usually present above the impeller blade pass
frequency. Since cavitation is caused by reduced NPSH, the solution is to raise
the NPSH above the fluid vapor pressure.
Recirculation
This is a pump problem with symptoms similar to cavitation. However, the
problem is very different. Recirculation occurs when the pump has such a high
suction pressure that it actually pulls the fluid from the output back through the
pump. The fluid can flow backward through seals and even the pump impeller
clearance.
The major contributor to recirculation is reduced flow in addition to high

suction pressure. The solution to this is usually to increase the fluid flow
through the pump or to install a bypass.
Recirculation is a hydraulic source of 1xTS vibration as well as

subsynchronous broadband noise and it can produce a random crackling noise
similar to what is heard with cavitation. The difference to remember is that,
with recirculation, the crackling noise will be intermittent and near the pump
inlet while with cavitation, the crackling noise will be noticeable on both the
inlet and outlet and will be steady.
Hydraulic Unbalance
Uneven flow distribution entering the impeller can introduce a hydraulic
unbalance that will show up at 1xTS in an axial direction. This can be caused
by the poor design of suction piping such as an elbow close to the pump inlet.
This causes the fluid to enter the pump impeller at different flow velocities. A
straight run of pipe equal to between 20 and 40 diameters before the pump inlet
is recommended to let the flow equalize before it enters the pump.
Hydraulic unbalance may also occur if the impeller is mechanically balanced

but not symmetrical in the fluid flow areas. This is similar to an eccentric rotor
defect.
Flow Turbulence
Flow turbulence may be more difficult to detect. This turbulence is
hydraulically induced by low suction pressure due to system problems that
produce inadequate NPSH. This generally shows up in a low-frequency region
between 0.1 and 0.4 orders of turning speed. Excessive turbulence has been
known to produce higher frequency broadband vibration as well.
Turbulence often occurs in pumps with diffuser vanes or a twin volute. If the
pump is not operating near its BEP, then the fluid leaving the impeller will not
enter the diffuser vanes at the correct angle which in turn causes the turbulence.
This tends to lead to higher axial vibration at the subsynchronous frequencies.
Resonance
Resonance is a problem that can occur in pumps as well as piping. Vertical
pumps produce classic resonance problems due to the long and slender nature
of the structure as well as the low natural frequency of most vertical pump
foundations.
Piping can also produce a resonance problem. This can be both a structural
pipe resonance and acoustic fluid column resonance. Both of these resonances
can be excited by various pump operational and pump defect frequencies.
Structural resonance problems should be corrected with modifications made to

the pipe or the pump foundation unless the forcing frequency can be altered to
prevent resonant frequency excitation.
Summary
This section has reviewed some basic pump vibration concepts. The basic
components of a pump—the impeller, volute and diffuser vanes—were covered
as well as the concept of net positive suction head and its importance to pump
vibration.
Vibration frequencies to watch for include impeller vane pass frequency and
diffuser vane pass frequency. These frequencies are generally operational and
do not indicate pump defects. However, if the amplitude increases
dramatically, this can be an indication of a developing problem.
Cavitation and recirculation can be similar, but cavitation generates broadband

noise around vane pass frequency and recirculation creates low-frequency
vibration.
Resonance can be a problem in pumps just as it can in other structures.

Vertical pumps are often prone to resonance problems and structural
modifications can be made to correct the resonance problem.
Appendix A
Glossary of Terms
Accelerance — see inertance.
Acceleration — the rate of velocity change of a mechanical system. Usually measured

in units of g (or sometimes G) in English units (1 g = 386.4 in/s2 = 32.2 ft/s2); the
international standard unit is m/s2, 1 g = 9.806 m/s2.
Acoustics — the study of sound and its interaction with the human hearing mechanism.
See also sound intensity and sound pressure.
Admittance — see receptance.
Amplitude — magnitude of a measured signal.
Analog — describes a signal represented by a proportional electrical voltage, current,

charge, etc. By association, any device which operates directly on an analog signal is an
analog device, for example analog amplifiers and analog filters are used to condition
analog signals at the front end of a digital signal analyzer.
Analog Integration — converting an analog signal representing one parameter into an

analog signal representing a different parameter by using analog electronics. For
example, an analog signal representing acceleration can be analog integrated once to get
an analog signal which represents velocity, or integrated twice to get an analog signal that
represents displacement.
Analog-to-Digital Converter (ADC or A/D converter) — a device for converting an

incoming analog signal to a series of discrete digital values by sampling. Described by
the number of bits it uses, (e.g., 8 bits or 12 bits), it is a key element in a digital analyzer.
The number of bits sets a theoretical upper limit on the analyzer's dynamic range,
equivalent to approximately 6 dB per bit.
Anchor — a term used to describe a reference mark used for measuring delta time or for
marking sideband intervals in the CSI Model 2400 cursor functions.
Apparent mass — the frequency response function representing force per unit
acceleration. Note: only in special cases is this function the inverse of the inertance.
Autocorrelation — a time domain function that compares a signal with delayed copies
of itself, using all possible time delays, and shows at which time delays the signal repeats
itself (periodicities). Although it has seen some applications in characterizing signals
buried in noise, usually more information is found in the frequency spectrum of the
autocorrelation which is the autospectrum (or power spectrum). Autocorrelation is a
special case of cross correlation.
1/99 ©Copyright 1999, Computational Systems Incorporated. All rights reserved. A-1
Glossary of Terms
Autoranging — the process of automatically adjusting the input gain of an analyzer to

match the amplitude of a signal. Optimizes the use of the dynamic range of the analyzer
and improves signal-to-noise ratio.
Autospectrum — another term for the averaged power spectrum of a signal. It has
magnitude only, phase information being effectively discarded during processing.
Averaging — a process of summing a number of time waveforms, sample by sample,

or a number of spectra, frequency by frequency, to obtain a better estimate of the mean
properties of a signal in the presence of noise or other interfering signals. May be carried
out in a number of ways, with or without weighting, including linear (power), negative
linear, exponential, peak hold, and synchronous, linear vector types.
A-Weighting — a frequency spectrum shaping applied to frequency spectra in acoustics.

The effect is designed to approximate the way that the human ear perceives the loudness
of sound. Sound levels are reduced at low frequencies and at very high frequencies,
where the ear is less sensitive. There are national and international standards for A-
weighting.
Bandwidth — (a) the overall frequency range of an analyzer, e.g. 0 to 1000 Hz; (b) the
frequency range over which a filter passes a signal without attenuation; (c) the effective
frequency range represented by one line in an FFT spectrum.
Baseband — the frequency range from the low-frequency cutoff to the maximum
analysis frequency for an FFT analysis.
Baud Rate — unit of speed for data transmission over a serial communications link.
Bodé Plot — in general, a dual plot of both the phase and magnitude information in a
signal as a function of frequency. Often used in a more specialized way to describe the
plots of synchronous amplitude and phase as functions of synchronous frequency for a
machinery spinup or coastdown.
Coherence — a function of frequency which describes the degree of linear relationship

between two signals. Used to assess cross-channel measurement quality, locate noise
sources, and to check out transmission paths.
Compliance — or Dynamic Compliance: see receptance.
Correlation — see autocorrelation and cross correlation.
A-2 ©Copyright 1999, Computational Systems Incorporated. All rights reserved. 1/99
Glossary of Terms
CPM — cycles per minute. Favored by many in machine vibration analysis since the
vibration caused by unbalance shows up at a frequency in CPM equal to the RPM of the
shaft. Sixty cycles per minute (CPM) is equivalent to one cycle per second which equals
one hertz.
Cross correlation — a time domain function that compares two signals using all
possible time delays and shows at which time delays the two signals are strongly related.
Although it has applications in transmission path analysis and system identification,
usually more information is found in the frequency spectrum of the cross correlation,
which is the cross spectrum.
Cross spectrum — the basic cross-channel measurement. Used most in calculating

other functions such as the Transfer Function, Frequency Response Function and
Coherence. Its magnitude measures how strongly two signals are related, frequency by
frequency; its phase measures the phase shift between two signals, frequency by
frequency. Whether the cross spectrum is from A to B or from B to A depends on
whether the phase of channel B is referenced to channel A or the phase of channel A is
referenced to channel B. Both forms have the same magnitude, but their phases are equal
and opposite. Sometimes called the cross-power spectrum.
Cursor — a manually controlled marker that can be moved across a plot to indicate the
amplitude at each time or frequency. See also harmonic and sideband marker.
Decibels (dB) — a logarithmic system of nondimensional units that measures the size
of a quantity relative to a reference level. Any quantity can be measured in this way, as
can any two quantities with the same dimensions be compared using decibel measure.
Given a reference power (amplitude squared) level Wref, any other power quantity W,
having the same dimensions, may be expressed in decibels using the formula:
dB = 10 log10 (W / Wref)
If a quantity X is in RMS amplitude units, and Xref is a suitable reference level, the
formula may be rewritten using W = X2, to give:
dB = 20 log10 (X / Xref)
Digital — describes a signal whose level is represented by a discrete series of numbers,

in a format suitable for processing by a digital computer. The representation may be in
the form of a waveform or a spectrum. By association, any device that operates on a
digital signal is a digital device, for example digital filters are used to filter digital signals
after conversion from analog to digital form in a digital signal analyzer.
Glossary of Terms
Digital Integration — converting a digital signal representing one parameter into an

analog signal representing a different parameter by using digital processing. A (digital)
frequency spectrum representing acceleration can be digitally integrated once to get a
(digital) frequency spectrum which represents velocity, or integrated twice to get an
analog signal which represents displacement. A single digital integration of an FFT
spectrum (in RMS, 0-pk or pk-pk amplitude format) is carried out by dividing the
magnitude of each spectrum line by the frequency (in radians per second) of the line.
Displacement — the distance that an object moves, especially when vibrating. There
are two types of vibrational displacement in common use:
(a) relative displacement: e.g., the movement of a shaft relative to a bearing

surface, measured by proximity or eddy current probes.
(b) absolute displacement: as measured from a doubly integrated acceleration

signal, picked up by an accelerometer on the casing of a machine. Units are
mils (thousandths of an inch) and microns (millionths of a meter), which
may be shown in RMS, 0-Pk or Pk-Pk formats.
Downloadable firmware (software) — refers to firmware (software) for controlling an

analyzer that may be transferred to the analyzer from a computer over a computer
interface. The Model 2400 has downloadable basic operating firmware, as well as
downloadable applications software, including the standard FFT program supplied with
every unit.
Dynamic Flexibility — see receptance.
Dynamic Range — the ratio between the largest and the smallest signals that an analyzer
can detect when measured at one and the same time. Not to be confused with the input
range of an analyzer, which depends on the available gain settings in the analyzer, nor
the equivalent range of the A/D converter, which sets a theoretical upper limit. Actual
dynamic range depends upon the quality of the analog electronics in the input signal
conditioning stages, the number of bits and error level in the A/D converter, the jitter in
the sampling clock, effects of quantization, and the accuracy of the computation.
Dynamic Stiffness — the frequency response function representing force per unit
displacement. Sometimes known as effective stiffness. Note: only in special cases is this
function the inverse of the receptance or compliance.
Glossary of Terms
Exponential — (a) a type of exponentially decaying window applied to transient

waveform data to improve its decay rate and minimize leakage in the spectrum; (b) a type
of frequency weighting used to give a better measure of the averaged value of a time-
varying signal.
External sampling — using an external sampling clock signal to override the normal
internal sampling clock of an analyzer. Usually uses a synchronous signal giving a fixed
number of pulses per revolution of a shaft. In this way all FFT frequency components
synchronous with the shaft rotation occur exactly at the center of frequency lines and may
be measured without leakage using a uniform window. Nonsynchronous frequency
components will appear to shift in frequency and are likely to be subject to leakage.
Absolute frequency information in the spectrum is lost: each line represents frequencies
which are some fixed fraction of the rotation frequency of the machine.
FFT — Fast Fourier Transform. An efficient method of computing a frequency

spectrum from a sampled signal waveform, especially suitable for digital computers.
Filter — an analog or digital device which removes or attenuates unwanted frequencies

in a signal.
Firmware — the software which controls or instructs the basic operating functions of
CSI Machinery analyzers. So-called because this type of software is often burned in to
the microchips.
Flat Top — a type of time window designed to minimize amplitude errors in the
frequency range, at the expense of frequency discrimination. Mostly used for analyzer
calibration at a given frequency.
Force/exponential — a window combination used to improve the quality of the analysis

for an impact test (or other test with impulsive excitation). The exponential window is
applied to the response channel data, to minimize leakage, and the force window to the
exciting impulse channel data, to improve the signal-to-noise ratio. A force window is
a short rectangular (uniform) window which brackets the impulse and forces all other
data points to zero.
Frequency — number of times an event repeats itself per unit of time. Units are hertz
(Hz = cycles per second) or cycles per minute (CPM).
Frequency response function — a spectrum representing the input/output relationship

for a system, e.g., the vibratory response of a structure to an exciting force. The
frequency response function is computed as a special case of the transfer function.
Glossary of Terms
Fundamental frequency — the first frequency in a series of harmonic frequencies. For

example, the orders of shaft turning speed occur at harmonics (integer multiples) of shaft
turning speed.
G (or g) — a unit of acceleration, commonly used with the English system of units; 1 g
represents the acceleration due to gravity at sea level and is approximately equal to 386.4
in/s2, or 32.2 ft/s2 (9.806 m/s2).
Group velocity — the rate of change of phase with frequency for the cross spectrum of
two signals. If the two signals represent two measurement locations, the group velocity
can be used to estimate the time of flight of a signal passing between the two points.
Hanning window — a shaping function applied to a time record before the FFT is
calculated in order to smooth out end effects and reduce leakage in the spectrum.
Usually the default window type to use for analyzing continuous signals because of the
compromise between frequency discrimination and leakage suppression. See also
Windowing.
Harmonic — an integer multiple of a fundamental frequency.
Harmonic Marker — a marker used to indicate the multiple harmonics in a frequency

spectrum.
Hertz — a unit of frequency equal to cycles per second (CPS), usually abbreviated as
Hz. Favored by many in general signal analysis because it is the international (SI)
standard frequency unit. One hertz is equivalent to one cycle per second which equals
60 cycles per minute (CPM).
Impact Test — a type of test used to investigate the properties of a structure, in which
the structure is caused to vibrate by an impulsive load from an instrumented hammer and
the vibratory response is picked up by a vibration transducer.
Impedance — (a) mechanical impedance is the frequency response function

representing force per unit velocity. Note: only in special cases is this function the
inverse of the mobility; (b) acoustic impedance is a frequency response function relating
the sound pressure produced by a sound source (such as a vibrating surface) per unit
volume velocity; (c) specific acoustic impedance is a frequency response function
representing the sound pressure per unit area of vibrating surface; (d) electrical
impedance is the frequency response of an alternating current electrical system,
representing the ratio between voltage and current.
Inertance — the frequency response function representing acceleration per unit force.
Also known as accelerance.
Glossary of Terms
Instantaneous spectrum — the basic spectrum computed by the FFT process from a
single data record. A complex-valued function, it contains information about both
magnitude and phase with respect to the start of the time record.
Intensity — a quantity that measures the rate of power flow through a surface per unit
area, in a given direction. Usually refers to acoustic or sound intensity, but may also be
defined for vibratory power flow. May be measured with FFT analyzers by using two
transducers and the cross spectrum.
Linear — describes a system with special properties that has an output spectrum directly
proportional to its input spectrum and that does not generate new frequencies. The
constant of proportionality is fixed, but may differ from frequency to frequency. Many
systems can be analyzed as if they were linear, at least over restricted parameter ranges.
Linear averaging — a type of averaging in which the mean square magnitude of the
instantaneous spectra of a number of time records is computed for each frequency. Also
known as power averaging, the averaged spectrum is often expressed in the RMS format.
Linear vector averaging — a type of frequency averaging in which amplitude and phase
information in each instantaneous spectrum are used to define a vector and averaged in
a vector sense with spectra from other time records. Only meaningful if analysis is
triggered, in which case the spectrum is identical to the FFT of the time averaged
waveform.
Linear vector spectrum — a spectrum that has both magnitude and phase information,
or, equivalently, a complex spectrum. Examples include an instantaneous spectrum, the
cross spectrum, and a linear vector averaged spectrum.
Live-Time — a term growing in popularity, generally means showing the waveform and
frequency spectrum at the same time as data is being acquired. A live-time display gives
a visual impression of how a signal varies with time. Compare with real-time.
Mil(s) — a unit of displacement equal to one thousandth of an inch.
Mobility — the frequency response function representing velocity per unit force.
Modal Analysis — the process of modeling the modes of structural vibration, including
resonance frequencies and damping, by mechanical testing, frequency analysis and
computer processing.
Glossary of Terms
Negative linear averaging — a type of averaging that starts from an initial linear
averaged spectrum and proceeds to compute a new averaged spectrum by subtracting the
contribution of each new data record, instead of adding as in linear averaging. Very
useful for subtracting background effects from averaged data.
Nonlinear — describes a system whose output is not proportional to its input.
Nyquist Plot — in general, a polar plot of the real and imaginary parts or magnitude and
phase, of a complex spectrum, such as an instantaneous spectrum, a linear vector
averaged spectrum or a cross spectrum. The term is also used in a special sense in the
analysis of machine spinup or coastdown data to mean a plot of the synchronous
magnitude and phase of one of the orders of shaft running speed in a polar format as the
synchronous first order frequency changes. At any speed, the magnitude and phase in the
Nyquist plot are exactly equal to the magnitude and phase in the corresponding Bodé
plot.
Octave band — describes a type of bandpass filter that has a bandwidth equal to 70.7%
of its center frequency. Conventionally used for analyzing sound levels, there are
standardized center frequencies and filter characteristics for such filters. Adjacent octave
band filters have center frequencies spaced approximately a factor two (one octave) apart,
hence the name. A frequency analysis in terms of octave bands is called an octave (also
whole octave or 1/1 octave) analysis. An FFT analyzer can be used to synthesize octave
band filters from an FFT spectrum and thereby carry out an octave band analysis,
although this is strictly only valid for steady signals.
One-third octave — like octave band, but for a bandwidth of approximately 23.1%.
There are three one-third (1/3) octave bands for each octave band. A one-third octave
spectrum can be synthesized from one or more high resolution FFT spectra.
Operating Deflection Shape (ODS) — the way that a machine or structure is

deforming, or moving, at one or more frequencies under the action of normal operating
loads. Use of multichannel analysis greatly simplifies and improves the range and
quality of ODS analysis; dedicated ODS or modal analysis software can be used to
generate visualizations of the operating deflection shapes.
Order — a multiple of a shaft turning frequency. The first order is the shaft frequency
itself, in CPM numerically equal to the machine RPM.
Order Analysis — a frequency analysis in which the frequencies are expressed as orders
of shaft frequency, either by normalizing by the shaft frequency, or by using external
synchronous sampling.
Order Tracking — a measurement of a signal from a machine whose speed is changing

with time, showing the level of one or more orders as a function of machine speed or
time.
Glossary of Terms
Overlapped processing — a way of acquiring and processing data when using a

Hanning window, in which each successive time record uses part of the previous record.
This gives an increase in the rate at which the display can be updated, giving an apparent
increase in live-time rate for lower frequency ranges. There is also an improvement in
the smoothness of the data and the statistical error is reduced for a given data acquisition
time. However, the method is inappropriate when triggering data acquisition and it
causes an actual reduction in the real-time rate. An overlap of at least 50% (1/2) is
required to gain benefits from smoothing the data, but the smoothest data is achieved for
overlaps of 2/3, 3/4, 4/5, 5/6, etc.
Peak — (a) the overall maximum level of a signal in a given period of time. For
sinusoidal (single frequency) signals, the peak level is 1.414 (%2) times the RMS level;
for non-sinusoidal (multiple frequency) signals, this is no longer true, and the peak and
RMS are not simply unrelated. Abbreviated as pk, 0-p or 0-pk. (b) An isolated
maximum in a frequency spectrum, either due to a single-frequency component or the
resonance of a system.
Peak-to-Peak — (pk-pk, p-p) the difference between the maximum and the minimum
levels (positive or negative) in a signal over a given period of time. For a sinusoidal
(single frequency) signal, the peak-to-peak level is always 2 times the peak level and
2.828 (2%2) times the RMS level. For non-sinusoidal (multiple frequency) signals this
is no longer true and there is no simple relationship between peak-peak, peak and RMS
levels.
Peak hold — a type of averaging in FFT analysis in which the maximum level is
retained for each frequency line over all data records processed.
Period — time required for one complete cycle of a periodic signal.
Periodic signal — an ideal signal that repeats itself exactly after a fixed finite interval
of time and exists for all time. Although not possible in the real world signal, many
signals behave like periodic signals for a certain length of time, for practical purposes.
Fundamental concept behind FFT analysis.
Phase — (a) the relationship between the angular location of the high spot and heavy
spot for an unbalanced rotor; (b) the angular relationship between the peak in the
vibration at a synchronous frequency and a phase reference (tachometer) pulse, for a
rotating machine; (c) the delay between two signals at a given frequency, expressed as
a fraction of a cycle, usually in degrees.
Power averaging — see linear averaging.
Glossary of Terms
Power spectrum — a spectrum of a signal formed as the mean square average level of
a number of instantaneous spectra. Often displayed in RMS format, by taking the square
root of the mean square level at each frequency.
Power Spectral Density (PSD) — a representation of the power in a signal that

compensates for the bandwidth of the analysis. In FFT, PSD is computed from a power
spectrum (in units of power = (amplitude)2), by dividing by the bandwidth of each line
in hertz. The units of PSD are then (amplitude)2/Hz. Sometimes, an RMS format is
used, in which case the units are (amplitude)/%Hz.
Pre-/Post Trigger — triggered data acquisition using a delay so that the time record
starts before (pre-trigger) or after (post trigger) the trigger event. The Model 2400 can
use trigger delays from 8 time record lengths before the trigger to 100 time record lengths
after the trigger.
Real-time rate — refers to the frequency at which the time an analyzer takes to compute
an FFT is equal to the time required to acquire the data. Commonly used as a measure
of the speed of an analyzer and equally commonly confused with the rate at which the
display is updated. Overlapped processing reduces the actual real-time rate.
Receptance — the frequency response function representing displacement per unit force.
Also known as admittance, (dynamic) compliance, and dynamic flexibility.
Resolution — the frequency range represented by one line of an FFT spectrum. Found
by dividing the maximum analysis frequency by the number of lines. The resolution in
Hz is equal to the inverse of the data record length in seconds.
RMS — root mean square; as applied to a dynamic signal such as vibration or sound
refers to an averaged level of a function obtained by averaging the square of the signal
level over a period of time (or number of data records), then taking the square root of the
result.
RS232 — a serial, asynchronous communications standard; a type designation for cables

used to connect communications ports on a computer with other digital devices such as
digital analyzers, printers and modems.
Sideband — a frequency component that represents the effect of modulation on a signal.

If a modulated signal has more than one component, each component will show
sidebands. A sideband is spaced off from the frequency of the modulated signal by an
amount equal to the modulating frequency. If the modulating signal has multiple
components or if there is frequency modulation, the sideband pattern may be very
complicated including sum and difference frequencies between the sideband component
frequencies (intermodulation effects).
Glossary of Terms
Sideband Marker — a marker used to indicate the sidebands around a center frequency
marked by setting a mark, then highlighting an adjacent frequency component.
Signal-to-noise ratio — the ratio of the power of the signal to the power of the
background noise effects in a measurement, usually expressed in decibels. In a signal
analyzer, the signal-to-noise ratio is typically improved by increasing resolution or the
number of averages, among other factors.
Software — computer programs for calculating functions or controlling digital devices

with a digital computer.
Sound — vibratory movement of the air, or some other conducting fluid, characterized
by a compressive wave mechanism with a constant speed of propagation in a
homogeneous unrestricted medium.
Sound Pressure — the pressure exerted by the movement of fluid particles in a sound
wave.
Spectrum — the frequency domain representation of a signal. In practical

measurements, the spectrum usually displayed as a plot of magnitude against frequency
over a limited frequency range. See also cross spectrum, power spectrum and linear
vector spectrum.
Synchronous averaging — a type of averaging in which successive time records are

averaged together without computing a frequency spectrum. If the analysis is triggered
synchronously from a rotating shaft or some other periodic event, the averaged waveform
will emphasize the synchronous components of the signal and suppress the asynchronous
components like noise and background effects. The spectrum of the synchronously
averaged signal is a linear vector averaged spectrum of the data, having both magnitude
and phase information.
Tachometer — device that generates a pulse signal corresponding to the revolution of

a shaft, used to measure turning speed. A single pulse per revolution may be used to
trigger data acquisition synchronously with shaft rotation.
Time record length — the time required in FFT analysis to acquire the number of
samples required to obtain a given number of lines at the sample rate required to achieve
the maximum analysis frequency selected.
Transfer function — a spectrum representing the relative magnitude and phase of two
signals. For two signals A and B, the transfer function from A to B is the ratio of the
cross spectrum from signal A to signal B, divided by the autospectrum (power spectrum)
of signal A. Its phase is equal to the phase of the cross spectrum.
Glossary of Terms
Transient — a time-varying signal of finite duration, i.e. having a definite start and
finish. May refer to an impulsive signal, such as a hammer blow or the vibration signal
from a machine coastdown or spinup. Such signals have finite energy, unlike periodic
signals.
Trigger — the control signal for starting and stopping data acquisition. May be based
on an incoming measured signal, an external pulse signal, or an internal clock.
Uniform window — a type of window used for analyzing a signal without shaping.
Subject to leakage and amplitude errors if the frequency components are not centered on
a line in the spectrum. Suitable for transient signals wholly contained within the analysis
time record length and when using external sampling. Also known as a Bartlett, Boxcar,
or Rectangular window. Compare Hanning window.
Velocity — the rate of change of displacement of a mechanical system. Units are inches
per second (in/s or ips) in English units and m/s, cm/s or mm/s in SI units. Can be
measured directly with a velocity pickup or by integrating an acceleration signal from an
accelerometer.
Vibration — the oscillatory motion of a mechanical system about a mean position.
Waveform — analog or digital representation of a signal displayed as a plot of level

against time.
Windowing — a process of applying a weighting to a waveform signal before computing

the FFT in order to minimize leakage and/or the picket fence effect that gives misleading
spectrum levels. See also Hanning, Uniform, Exponential, Force/exponential.
Zoom — a frequency analysis at higher resolution than the baseband spectrum over a
limited frequency span in order to see more detail. There are two types: nondestructive
zoom and real-time or true zoom. The latter involves re-analyzing the signal
(destructive), using frequency translation and digital filtering to obtain the results.
Nondestructive zoom involves acquiring more samples in the first instance, giving a
potentially higher resolution anywhere in the baseband frequency range. More detail can
then be seen merely by expanding the frequency scale in the region of interest.

Vibration Intermediate

Uploaded by

Copyright:

Available Formats

Vibration Intermediate

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

Vibration Intermediate

Uploaded by

Copyright:

Available Formats

What are the steps to properly prepare for vibration analysis?

What are the steps to properly prepare for vibration analysis?

What are some common types of faults that can occur in rotating machinery?

What are some common types of faults that can occur in rotating machinery?

INTERMEDIATE

DETROIT TRAINING CENTER

HOUSTON TRAINING CENTER

SAN DIEGO TRAINING CENTER

ONE STEP IN YOUR JOURNEY TO BENCHMARK STATUS

Two Review of Vibration Basics

Three Effects of Amplitude and Phase

Seven Antifriction Bearings

Eight Data Averaging

Nine Digital Signal Processing

Ten Waveform Characteristics

Eleven Peak/Phase Analysis

Thirteen Electrical Defects

Sixteen Pump Vibration Basics

< Recognize the importance of predetermining potential defect frequencies.

< Identify the difference between synchronous, nonsynchronous and

Preparing for Analysis

Ô Collect machine information.

Õ Determine appropriate measurement points.

Ö Calculate potential fault frequencies.

× Determine alarm criteria.

Ø Set priority of potential faults.

Ù Determine possible fault causes.

Ú Establish preventive actions.

Û Establish information feedback.

Vibration Analysis Flow Chart

Potential Fault Analysis Diagnostic Analysis

Set Priority Define

Determine Compare Actual

Correct most Contingent

Steps to Solving Vibration Problems

< What part of the machine has the vibration problem?

Õ The machine geometry should be understood as completely as possible.

< Sketch the entire machine train.

< Determine belt information.

< Drives: motors, engines, turbines

Ö Gather available maintenance history and any previous vibration data.

< Has vibration data been collected on this machine before?

× Determine the best points to collect data.

< At the bearings

• Coastdown or startup data

Ø Determine as many as possible forcing frequencies before taking data.

< Check for beats by watching the instantaneous spectrum or by

Blade/Vane Pass . . . . . . . . . . . . . # of blades/vanes x TS . . . . . . . Radial

Antifriction Bearing Loose in Housing

Cage or Train Frequency of Antifriction Bearing

Primary Belt Frequency

Defective Tooth-to-tooth Repeat Frequency

Ignition or Fuel Problem on Four-Cycle Recip

Check horizontal and vertical amplitude and phase at each bearing. If

If the horizontal-to-vertical phase shift is not close to 90E, consider possible

If the phase relationships side-to-side in both the horizontal and vertical

If the readings indicate an unusually high unbalance, look for a possible

If the misalignment is primarily angular, the top-to-bottom and the side-to-side

Any truncated function will produce harmonics in the spectra so that a

One or more system resonances may be excited by one or more of these

ONE STEP IN YOUR JOURNEY TO BENCHMARK STATUS

T = .07031 sec. - .03144 sec.