Basic Vibration Primer

Download as pdf or txt
Download as pdf or txt
You are on page 1of 28
At a glance
Powered by AI
The key takeaways are that predictive maintenance using vibration monitoring technology is more cost effective than run-to-failure or preventive maintenance philosophies. Vibration monitoring can predict imminent failures before they occur.

The different types of maintenance philosophies discussed are run-to-failure, preventive maintenance, and predictive maintenance.

Vibration monitoring is used for predictive maintenance and is a vital part of reliability-based maintenance. It gives the most information about the condition of machinery by collecting data while machines are in normal operation.

DoctorKnow® Application Paper

Title: Basic Vibration Primer


Source/Author: Brian Overton
Product: General
Technology: Vibration
Classification: Basic

Source: www.csimeansreliability.com

BASIC VIBRATION PRIMER

By
BRIAN OVERTON,
CSI Training Instructor

In the competitive realm of industry, a company must be able to produce defect-free products while maintaining
a competitive price. In order to achieve this goal, cost-effective maintenance must be performed on the
machinery used. To keep maintenance costs down, unexpected failures must be minimized or eliminated and all
machinery shutdowns for maintenance should be planned. All of these concepts are included in the Reliability-
Based Maintenance TM(RBM)TM philosophy. This philosophy involves the concepts of Preventive, Predictive,
and Proactive Maintenance. This paper will primarily focus on the Predictive maintenance technique.

Before predictive maintenance came on the scene, most maintenance was performed using either the Run-to-
Failure or Preventive maintenance philosophies. These philosophies, however, proved to be less cost effective
than their successors. By allowing a machine to run until failure, the repair costs escalated dramatically. In
general, the repair would involve more repair parts, longer shutdown periods, and more labor to complete.

Preventive maintenance allowed a calendar, or some type of schedule to govern the repair work, whether it is
needed or not. By performing this type of maintenance, some of the repairs that were being performed were
unnecessary. Parts that were in good condition and performing well were being replaced. This caused the repair
costs to escalate because of the amount of repair parts being consumed. Also, some of the work took out good
parts and replaced them with "new" parts, some with preexisting defects that caused problems in a short period
of time.

Predictive maintenance then became the new kid on the block, in particular, the vibration monitoring
technology. This technology has been around for decades, even though many of today's maintenance personnel
tend to think that this is totally new. The use of state-of-the-art digital equipment is relatively new and
continuing to improve as time progresses. The remainder of this paper will focus on the discipline of vibration
monitoring. It will present some of the basics about establishing a vibration monitoring program as well as
discuss some of the basic machinery faults that are detectable with vibration analysis.

What is vibration?
Vibration is the movement of a body around a reference point. This reference may be the center of a bearing, the
center of a shaft, or even the center of a bearing housing. By collecting this information and analyzing the data,
we can predict imminent failures of machine components before they actually occur.

Vibration analysis is a discipline of Predictive Maintenance and thus is a vital part of the RBMTM philosophy.
Other disciplines include thermography, oil analysis, electric motor current monitoring, as well as precision
alignment and balancing.

Of these disciplines and other process variables that can be monitored, vibration gives the most information
about the condition of your machine. An additional benefit is the data is collected while the machine continues
its normal operation.

Historical Information
The earliest use of vibration in industry known to the author would be the creation of the Rathbone Chart (See
Appendix A) by Mr. T.C. Rathbone in the 1930's. Mr. Rathbone was an underwriter of insurance for industry.
Before he would underwrite an insurance policy, he wanted to know the mechanical condition of the machine to
be insured. Based on casing measurements, RPM of the machine, and the overall vibration measurement, he
created the chart that he used to determine the machines condition. The condition of the machine was
determined to be very smooth, good, fair, slightly rough, rough, or very rough. This seemed to work well for
industry in the 1930's.

Just a decade and a half ago, the tools being used in vibration monitoring were analog analyzers. These
analyzers provided the operator with an overall vibration measurement. This overall reading was a simple value
used to indicate the total amount of energy contained in the machine. Only having one number to base your
machines condition on has since been proven to be inaccurate (See Appendix B). At this point in time, this
equipment was state-of-the-art. It did beat using the screwdriver or coins to determine the amount of vibration in
the monitored machine.

The next progressive step taken was to move into the world of digital analyzers. The introduction of the digital
analyzers to the vibration discipline opened up many doors for the natural progression and continuous
improvement of the vibration analyzers as we know them today. The equipment was at one time very bulky and
heavy. Today we have better analytical abilities in our equipment, but at a fraction of the size and weight.

Basics of Vibration
When establishing a vibration monitoring program, a few basic concepts must be understood. These concepts
are the type of transducer to use, where to place the transducer to get the best data, the maximum frequency,
resolution to use for data collection, and what data to collect and store. This means there has to be a working
knowledge about the available transducers, about the components that make up the machinery being monitored,
and what data can be collected.

Transducers
The most common transducer being employed in industry today is the accelerometer. One of the key reasons for
this is the broad frequency response range they have. This range is typically 2 Hz -20 kHz. The maximum
frequency is dependent upon the mounting technique utilized for collection. Specialized accelerometers are also
available for specific applications, such as low frequency data collection.

A large portion of today's accelerometers are called ICP accelerometers. ICP stands for Integrated Circuit
Piezoelectric. These types of accelerometers will be resonant at some given frequency, which affects the
maximum linear frequency. The mounting technique used for the accelerometer changes the resonant frequency
of the accelerometer. We need to keep in mind the three factors that affect the resonant frequencies. Those are
mass, stiffness, and damping. The most common mounting techniques available for transducers are stud
mounting, quick-lock adapters, magnets, or hand-held.

Another type of transducer that is available is the Seismic Velocity transducer. The frequency response range of
this transducer is somewhat limited, approximately 10 Hz 2000 Hz. A really big advantage for this transducer is
that it does generate its own power for operation.

The last of the transducers is the proximity probe, also called a displacement probe or eddy current probe. This
transducer is also limited in its frequency response range, about 0 - 1000 Hz.

Each of these transducers have its list of advantages and disadvantages from the others (See Appendix C). They
each also have applications for which each is more suitable. For this reason, the author feels that an outstanding
maintenance program must have more than one transducer available. No one transducer will be able to meet all
the needs of an intensive maintenance program.

Transducer Placement
The location at which the transducers are placed is very important. You must insure that a good transmission
path is available from the possible sources of the vibrator energy to the transducers. Keep in mind that the
sources of some vibration will travel through other material before reaching the transducer. For example, an
inner race defect frequency must travel through the rolling elements, the outer race, and the bearing housing
before it reaches the transducer. By this time, the signal is damped and the amplitude appears very low. This
explains why much of vibration analysis is pattern recognition and not so much the amplitude of the vibration.
Some inner race defects have caused a bearing to fail with amplitudes of .02 - .03 IPS and lower.
Many faults have very discernible patterns in either the frequency domain, time domain, or both. Some of these
faults, if present, you would refer to the amplitude and consider the physical size and application of the
equipment before making your recommendations.

Let's look at an example. If a small, non-critical, 5 Hp motor is out of balance and exhibits a .5 IPS peak at
1xRPM, it may not cause an immediate concern. On the other hand, if a 150 Hp critical motor is out of balance
and has a .5 IPS peak at 1xRPM, a more immediate concern is likely. The reasons for the difference is that one
is critical and of larger physical size. If it is vibrating at .5 IPS, there is obviously a large amount of energy in
the machinery that needs to be corrected.

Another key concept to keep in mind is that the data must be repeatable and collected in the correct plane. Some
of the faults that are diagnosed using vibration analysis will show a higher amplitude in one direction and a very
low amplitude in another. It is highly recommended that two radial readings be collected at each bearing and at
least one axial reading be collected per common shaft.

Knowing where to place the probe is critical to collecting "good" data. You want to avoid the break between the
two halves of the bearing housing if a split housing is used. End caps are not very rigid and do not provide a
good transmission path; therefore, avoid using the end caps of motors.

The data collection points should be marked on the machine in some fashion. Good recommendations are a
paint pen, center punch, permanent studs, quick-disconnects, or use flat washers as "targets". Any of these
methods will help provide repeatability of data collection.
Helpful Hints
• ALWAYS USE CAUTION during data collection to prevent accidents. Data collection requires you to
go into all areas of your facility and monitor many different types of machinery. NO DATA IS
WORTH INJURY.
• Allow 3-5 seconds settling time for accelerometers that have been snapped down into position or
impacted in any way.
• Use a flat piece to span the cooling fins on motors.

Data Collection

Now that you know where to place your transducer to collect your data, you need to understand what data is
being collected, how it is being stored, and that the data is meeting your specific needs for analysis.

The raw data that is coming into your analyzer from the transducer is the time waveform. This signal is a
voltage signal over time. The voltage levels are divided by the sensor sensitivity value that you assign on each
measurement point. This is why it is important to insure that you have the correct sensitivity level programmed
for the transducer that you are using. The results of such mathematical process is the units of display that you
have selected.

The units of display for amplitude levels are acceleration (G's), velocity (in/sec), and displacement (mils). Your
data may be collected and stored using any of these units. You may also convert these units from one to the
other without having to return to the field for more data collection when in the frequency domain. Once a
waveform is stored, it remains in the units that it was stored in when in the time domain. The equations for sine
waves are as follows:

1.
V = 0.0031416 * f * D

2.
A =0.01146 * V * f

3.
A = 0.00003613 * D * f
4.

5.

6.

The amplitude levels may also be expressed as peak, peak-to-peak, RMS, or average values. Typically, the
overall value of the measurement is an RMS value. If you have selected peak, peak-to-peak, or average, then
these are actually calculated from the RMS value using the following equations:

Pk = 0 to A (Peak)
P-P = 2.0 X A (Peak-to-Peak)

RMS = 0.707 X Pk
(Root Mean Square)

Pk = 1.414 X RMS
Avg. = 0.637 X Pk
Note: The conversions above are true only for sine waves.

In the figure of the waveform above, notice that the time it takes for one cycle to complete is designated by "T".
This is referred to as the period of the cycle. This figure also shows the relationship between the peak, peak-to-
peak, RMS, and average values.

One of the great advantages of available software packages is that you can convert from one display unit to
another. For example, you can collect your data in units of acceleration and later display it units of acceleration,
displacement, or velocity. To convert units from acceleration to velocity or displacement requires that the data
be integrated when it is processed. If the conversion is from displacement to velocity or acceleration, then the
data must go through a differentiation process.

The relationship between acceleration, velocity, and displacement measurements are depicted in the following
figure.
The shaded area on the figure above indicates normal operating speed range for industrial machines.

Once the time waveform has been collected, then another mathematical process is performed, and that is called
an FFT (Fast Fourier Transform). This will convert your time waveform to an FFT display, or frequency
spectrum.

When diagnosing a machine, a frequency spectrum is much easier to use than a time waveform. However, the
time waveform still holds some very vital information that is helpful in the diagnostic process. Certain defects
will have distinct patterns in the waveform that may help confirm your diagnosis of a particular fault.

The frequency spectrum will display the frequency at which an event is


occurring, the amplitude at each frequency, and depending upon the resolution it can separate closely spaced
frequencies. Units of frequency may also be displayed in one of three ways as the amplitude may be. These
three units are hertz (CPS), cycles per minute (CPM), or Orders. When displaying in orders, you are simply
displaying the data as a multiple of turning speed. Orders are always referencing the turning speed at the
measurement point the data was collected.

Spectral plots may be divided into three areas of interest. These areas are Subsynchronous, Synchronous, and
Nonsynchronous.

Subsynchronous references the area of the spectrum below the turning speed of the machine. Some of the faults
that may be found in this area are oil whirl, cage frequency of a Roller Element Bearing, belt frequency, and
indications of rubs.

Synchronous components are those that are phase locked with the rotation of the shaft. Many different
frequencies will be synchronous and fall in this area of interest.

Nonsynchronous components are those that are not an exact integer multiple of turning speed. Some of these
causes are rolling element bearings, another component on the machine, or another machine.

By dividing the frequency domain into these three different areas, the analysis process is somewhat simplified.
Another means of simplification that CSI uses is the Fault Frequency sets. These Fault Frequencies are defined
based on the expected faults for a particular machine. When utilizing these fault frequencies, an overlay is
displayed over the spectral plot. This overlay allows you to quickly determine if any of these expected faults are
suspect. Fault frequency sets are not required, but are a very useful tool for you the analyst.

Consideration is also given to the type of averaging mode to be utilized during the data collection process.
Normal averaging is used for routine data collection. Peak Hold is really not an averaging technique, although
it is considered an averaging mode. This mode will hold the highest amplitude levels attained at each spectral
cell in the frequency domain. Negative Linear averaging provides a means of spectral subtraction to remove
noise from one machinery configuration to another. Synchronous Time averaging will attenuate the amplitude
levels of all frequencies that are not synchronous to the turning of the reference shaft. This averaging mode
requires the use of a phototach or strobe light for the once per revolution signal. The last averaging mode to
discuss is Order Tracking. Order tracking is used when the speed of a machine continuously fluctuates. Order
tracking is capable of tracking a change in speed up to 6% between averages. Without order tracking, the data
would be smeared and the peaks would have broad skirts on them. This mode also requires a tach input for the
once per revolution signal.

The window that is used during data collection is also very important. In the 2115 analyzer you only have two
options, Hanning and Uniform.

Basically, the Uniform window is utilized during transient type events when the beginning and ending data will
be the same. The Hanning window is used with routine data collection that will be performed in an RBM
program.

Overall calculations and integration methods may also be changed in the analyzer. The options are the same for
the overall calculation and the integration methods: Analog and Digital.

Let's discuss the overall calculation first. When using analog, the overall value is calculated from the waveform
and uses all frequencies from 0 - 30kHz. This method is only limited by the maximum frequency of the
analyzer. The digital method is performed in the frequency domain and will be limited by the maximum
frequency selected for the spectral display.

Integration methods are very similar to the overall calculation. When using analog, the data is integrated in the
time domain, an FFT is performed, and the resulting frequency domain display is in the units you selected.
However, the waveform is also displayed in the same units.

When using the digital integration method, the data is integrated in the frequency domain instead of the time
domain. The FFT is performed first, leaving the waveform in the units of the transducer, the FFT is integrated
into the units of choice.

Phase Data

Another diagnostic tool that provides you with vital information is phase data. Phase is simply a time
relationship between two events. In single channel vibration analysis, this relationship is between the peak
amplitude of the vibration signal and the firing of the phototach. The time difference is then used to calculate the
phase angle.

Phase is one of the most important vibration analysis tools an analyst can have at his/her disposal . The analyst
uses phase when trying to balance an unbalanced rotor to locate the heavy spot. Phase is also a useful tool to
determine types of unbalance, misalignment, looseness, soft foot, bearing misalignment, resonance and other
machinery faults.

The illustration shown above provides some insight for understanding phase and the meaning that it has in
industry with respect to machinery diagnostics.

The signal labeled "T" is a tachometer reference pulse. This may be generated with a strobe light, infrared or
photo tach pulse. The signal label DX is the vibration signal representing an unbalanced rotor condition
(1xRPM).
Phase as stated earlier is the relationship between two signals. This relationship is based on the differential time
between the two signals.

The difference is considered the Dt. This time is calculated from the start of the tachometer pulse (the reference
signal) and the time required to see the unbalanced (the comparison signal) condition following the reference
pulse. With this time, the ability to determine the relationship between the two signals becomes much easier
based on the following calculation.

With the ability to draw this relationship in time, the analysts' proficiency in analyzing specific machinery faults
becomes enhanced.

Phase measurements are very important when diagnosing and correcting machine unbalance. However, other
machine faults such as misalignment and soft foot can show up at 1xRPM similar to unbalance. Therefore,
understanding the phase relationships of unbalance, misalignment and looseness and other machine faults can
provide further assistance in the prevention of an improper fault diagnosis.

The following machinery faults are the most common conditions that may be differentiated with the aid of
phase.

Unbalance - Static unbalance will show a zero degree phase shift across the rotor radial to radial or horizontal
to horizontal, a 90° (±20°) 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 that is related to the
heavy spots on each end of the rotor. If the heavy spots are 180° out of phase on each end, then the phase
measurements will also be 180° out of phase.

Misalignment - Angular misalignment will typically show a 180° ( ±30°) phase shift across the coupling in the
axial direction. Parallel misalignment will tend to show a 180°( ±30°) phase shift across the coupling in a radial
direction. Bearing misalignment (cocked bearing) will show a 180°(±30°) hase 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. Often this
shift will be greater than 90°. Also a soft foot would show a phase difference from the other machine feet.

Resonance - The phase shifts 180° from the frequencies below resonance to the frequencies above resonance. A
90° phase shift will be present right at the resonant frequency.

Composition of Machinery
It is very important that you know the composition of each machine as much as possible. In other words, what
type of bearings are used (sleeve or anti-friction), number of teeth on each gear, number of impeller vanes, etc.
By knowing this information, it will help alleviate some of the guess work when trying to analyze the data. Each
component will typically generate certain frequencies that you can predict. By utilizing this predictability, you
can set up fault frequency sets and very quickly pull up the overlays when analyzing the data. Bearings will
generate certain defect frequencies, gears will generate a Gearmesh frequency, and impellers will generate a
Vane Pass frequency.
If this internal component information is unknown, its like looking for a needle in a haystack. There is too much
left to guess work.

For example, if there is a spectral peak at 5xTS, how do you know if it is an indication of looseness or vane pass
frequency? As you can see, without the correct information, it is difficult to accurately diagnose your data. The
more work you do up front when preparing your database, the better you will be prepared for analyzing the data.
This preparation time may be intensive, but it pays off in the long run.

Imbalance
Imbalance is perhaps one of the more basic faults that is detected with vibration analysis. The distinguishing
characteristics are:

1. The time waveform will be sinusoidal with peaks at 1xTS.

2. The time waveform will be simple, periodic, and non-impacting.

3. If harmonics exist, they will be low in amplitude relative to the 1xTS peak.

4. There will be very little, if any, axial vibration.

With these characteristics now defined, you can see that an imbalance fault is indeed a basic fault. If there are
several harmonics of turning speed and the amplitudes are significant relative to the turning speed peak, then
you should suspect another fault other than imbalance. Imbalance is often times diagnosed as a problem when it
is actually another fault. Another quick check is to change the speed of the machine if possible. Due to the
increase in the centrifugal force, the imbalance amplitude will follow the speed of the machine. If the speed is
increased, then the imbalance amplitude will increase.

Let's stop for a moment and consider some of the causes of imbalance. If a machine has been broken down for
repair and then re-assembled, imbalance could occur if the machine is not assembled correctly or there are parts
missing. The same thing would happen if a part broke during normal operations. With these two causes of
imbalance you would see a sudden increase in the 1xTS amplitude.

What about situations in which there is material buildup on a component, such as a fan? In cases such as
material buildup and normal wear of components, you will have a gradual increase in the 1xTS amplitude. If
you suspect imbalance is a problem with your machine, then take a look at a trend plot of the 1xTS band and see
how the amplitude has changed over time, suddenly or gradually. This will give a better idea of what to look for
once you inspect the machine.

In most imbalance cases, I would recommend cleaning the component first. Often this will alleviate the cause of
the imbalance and a balance job is not necessary. If you balance the machine because of material buildup, then
later the material may fall off and now you are out of balance again. Take a look at the case history below.

The above multi-point plot shows the data collected on the T.A.D. Exhaust Fan. Notice that the vertical
measurements showed a very significant peak at 1xTS. The following figure will show the data from the B2V
measurement point. This single spectrum and waveform will show the sinusoidal waveform and the lack of
harmonics.
The time waveform, even though it is displayed in acceleration, shows the sinusoidal pattern as well as the
higher frequency activity. You can still discern the sine wave even though the acceleration units will diminish
the low frequencies to some extent.

In the figure below, the single spectrum that is displayed shows that there is some level of vibration in the axial
direction, but it is insignificant when compared to the radial vibrations. By insignificant we mean that the axial
vibration is less than half the amplitude of the radial vibration.

This fan was balanced and the vibration levels fell substantially. In the next figure, we show a multi-point plot
very similar to the first, but with data after the maintenance has been completed. Note the spectral plot span has
changed from .4 in/sec to .05 in/sec.
Misalignment
Misalignment is another very common fault found in industry. Over the years there have been several methods
used to "correctly" align equipment. Those methods include the string and straight-edge, flashlight and ruler,
reverse dial indicators and the latest technology of laser alignment. There are arguments to support which is
"better" for different applications, but that is not the focus of this paper.

The characteristics of misalignment as they manifest themselves in a frequency spectrum are as follows:

1. High axial levels.

2. High vibration at 1xTS or 2xTS in the spectrum.

3. A significant phase shift (approximately 180 degrees) in the axial or radial direction across the coupling.

4. The time waveform is repeatable and periodic with one or two clear peaks per revolution.

Misalignment can basically be categorized into three areas: angular, offset, and combination. Each type of
misalignment will have specific characteristics to that type. Angular misalignment shows a high axial
measurement at 1xTS. Offset misalignment shows a high 2xTS measurement radially. As you might expect, the
combination misalignment has both the above mentioned characteristics, high axial and radial measurements.
The combination of offset and angular is the most prevalent in the field. Below is a case history for
misalignment.

This data was collected on a lineshaft turbine. The turbine was misaligned with the gearbox. In the multi-point
plot above, you can see that a substantial 2xTS peak does exist. There are several multiples of turning speed
indicating a possible looseness problem. This looseness may be driven by the misalignment. The looseness may
also be representative of the wear on the sleeve bearings utilized by the turbine. Notice the dominant 3xTS peak
in the horizontal direction. This peak could possibly be indicative of a coupling problem or simply related to the
looseness. Let's take a look at the vertical reading in a single spectrum and at the waveform.
The single plot does show the harmonics of turning speed, as well as the 2xTS peak. Notice the amplitude of the
second harmonic. Also, the waveform shows two distinct peaks per revolution of the shaft. The vertical lines
indicate each revolution of the shaft. Looking at the amplitude of the waveform, you can conclude that there is
very little impacting occurring in this machine. By aligning this machine the problem would go away. There is
no data available after the maintenance was performed.

Bearings

Bearings are a major concern to most personnel involved in maintenance. In general, there are several reasons
why a bearing might fail. Some of these reasons are: pre-existing flaw from manufacturing,
improper installation, improper lubrication (over or under), excessive external vibration, and utilizing a bearing
housing as a ground for welding. These are not all inclusive but does give us some idea that there are several
reasons why the bearings fail. You can perform post-analysis on bearings and determine the cause of the failure.
This technology falls more into the category of Proactive Maintenance. Using this, you could determine why it
failed and eliminate the reason instead of continually replacing a by-product of the real fault.

As mentioned earlier, bearings will emit certain calculable defect frequencies. When these frequencies are
present, there is a defect in the bearing. The different frequencies that may be calculated
are:

FTF - Fundamental Train Freq.


BSF - Ball Spin Freq.
BPFI - Ball Pass Freq. Inner Race
BPFO - Ball Pass Freq. Outer Race

These frequencies are calculated based on the geometrical design of the bearing. Specifications for the pitch
diameter, ball diameter, contact angle, and the number of balls or rollers are used for these calculations. The
FRQCAL program in MasterTrend can be used to look up the bearing of interest and then display these
calculated values. This is very helpful when analyzing data. You can also create a fault frequency set using this
bearing information.
Some of the characteristics that you may see for a bearing defect are:

1. Nonsynchronous peaks and harmonics of these peaks, there may also be sidebands of 1xTS around these
peaks.

2. There may be a presence of broad band energy mounds. This is an indication of late stages of failure.

3. There will be a presence of impacting in the time waveform. Any impacting greater than 2 g's total swing is
cause for concern, unless the bearing is associated with a gearbox which will have high levels of impacting due
to the meshing of gears.

4. In the early stages of the defect, the amplitudes will be low.

The environment in which the bearing will be operating plays a very important role in the life expectancy of the
bearing. Each bearing is designed to last for a certain amount of time, but if the maintenance is not performed
they will fail prematurely. The bearings must be installed correctly and lubricated at the correct intervals, and
the correct lubricant must be used. Each bearing is also designed for a particular application. If a bearing is used
for the wrong application, it will fail prematurely.

In order to maximize the life that one can get from a bearing the load and speed may be adjusted. This is
typically done when a defect has been diagnosed and a planned outage is very near. By changing the load or
speed, the bearing may last until the maintenance can be planned. Load has a cubed effect on the bearing life
and the speed has a direct correlation.

In the characteristics of a bearing failure, it was mentioned that broadband energy may be present, especially in
the later stages. What is broadband energy? Broadband energy occurs when there is much impacting in the
waveform. This impacting is exciting all the frequencies in the bearing causing the amplitudes to increase. It
will be expanded over a wide range of frequencies, hence the name broadband. It may also be referred to as a
raised noise floor.
The following case history is from a paper mill that manufactures tissue paper. The data came from a primary
washer used in the process. The first plot is a multi-spectra plot of the same measurement point showing the
progression of the defect over time. The amplitudes did not vary that much, but the defect did warrant
replacement of the bearing.

Notice that there are several nonsynchronous peaks in all of the spectra except the last. This particular defect is
on the outer race, which may last for a long time before complete failure. Notice that the amplitudes are very
low, .07 in/sec is the highest peak of all the spectra shown. The amplitudes of an outer race defect will be
dependent upon where the defect is physically located. If the defect is in the load zone, the amplitudes will be
higher and the rate of progression will be faster. On the other hand, if the defect is not in the load zone, then the
amplitudes will be lower and the rate of progression will be slower. A bearing that has an outer race defect out
of the load zone may last for months, as did this one, or even longer before replacement is required.

The above figure shows the single spectral plot and waveform for the data collected on 8 July 1993. It is very
easy to see that there are 15 harmonics of the frequency at 10.33 orders (10.33xTS). This frequency is the
primary BPFO defect frequency for this bearing.

Usually this primary defect frequency may not be as discernible as it is in some of these plots. Often times the
other vibrations will be of such magnitude that the lower amplitude of the primary bearing defect frequency may
not be manifested in the spectrum. Notice also that there is about 2-3 g's total swing in the waveform for this
bearing. The next figure will show the data after the bearing has been replaced.
Note the amplitude levels after the bearing has been replaced. The levels are less than .01 in/sec.

The term "sidebands" has been used during this discussion on bearing defects. What are sidebands?

Sidebands are frequencies that are displayed around another center frequency. These sidebands are a result of a
true modulation taking place in the machine or as a result of beat frequencies.

Mechanical Looseness

Looseness falls into one of two categories.

1. structural looseness
a). base mount
b). split casings
c). bearing caps
d). bearings supports

2. rotating element looseness


a). impellers
b). fans
c). bearings
d). couplings

Characteristics of looseness:

1. Presence of a large number of harmonics of turning speed

2. Often directional in nature; horizontal and vertical amplitudes may differ greatly

3. Occasional occurrence of half-harmonics

4. Random, non-periodic time waveform

With mechanical looseness the waveform is very distinct in that it is random and has no repeating pattern. This
characteristic is what distinguishes looseness from other faults such as misalignment.

As mentioned earlier, phase data is very useful in determining a looseness condition and its location, but other
"tricks" may be used also. One such method is the use of a strobelight.
A mechanical looseness case history follows.

In the example above, peaks from 1XTS to 6XTS indicate looseness. Motor Inboard Vertical (MIV) appears to
have the most energy.

The plot above provides an expanded view of the preceding page. Motor Inboard Vertical (MIV) clearly shows
the highest peaks. Note the relative amplitudes of the five measurement points. This comparison enables you to
predict the type of looseness present on the machine.

The spectrum above from MOH shows a 1XTS peak with a hump of energy from 3XTS to 5XTS. The overall
amplitude remains relatively low at 0.1275 ZIPS, and harmonics exist to 10XTS. The non-periodic, erratic time
waveform below lacks a repeatable pattern even though small impacts are visible. This waveform pattern
indicates looseness.
Gears
Gearboxes may appear to be very complex when you first sit down to analyze one. Just remember to keep it
simple and not make a mountain out of a mole hill. The basic thing to keep in mind is the calculation for
gearmesh. Gearmesh is a simple calculation that is the product of the rotational speed of the gear and the
number of teeth on the gear.

GMF = (# teeth) X (TS)

For example, a 256-tooth gear runs at 3600 RPM or 60 Hz.

GMF (Hz) = 256 X 60 Hz


= 15.360 Hz

GMF (RPM) = 256 X 3600 RPM


= 921. 600 RPM

You usually calculate GMF in Hz. When you analyze a multiple gear train, you may need to calculate additional
parameters to identify which gear(s) show(s) defects.

Unlike other defect frequencies, gearmesh is always present in the spectrum. Its amplitude will change with the
amount of load on the gears. Here are a few of the concepts that need to be remembered when dealing with
gears.

1. Gearmesh frequency

a). Appears regardless of gear condition


b). Amplitude changes significantly with load

2. Sidebands

a). High level sidebands indicate problem


b). Sidebands indicate which gear is bad by the spacing between peaks

3. Gear natural frequency

a). Gear natural resonance excited by gear defects


b). Good indicator of a problem
The gearmesh frequency appears at 14XTS of the pinion. Note that this peak clearly dominates all three spectra
shown above.

Sidebands provide the key to gearbox analysis. The spectrum above shows sidebands of the gearmesh frequency
modulated by 1XTS of the output shaft.

Sidebands modulated by the speed of the input shaft appear on the spectrum above. Because sidebands of the
output shaft are also present, it might prove more difficult to determine which shaft has the defect. The input
shaft is the most likely culprit because of the higher amplitude of its sidebands.

Belts
A unique characteristic about belt defects is that the primary belt frequency is below the turning speed of the
components of the machine. The belt primary frequency falls in the subsynchronous area of the spectral display.

Another rather unique characteristic is that there are generally two sheaves in a belt driven configuration. This
means that if there is a belt defect, such as a crack in the belt, then it will strike both sheaves and cause the two
times belt frequency peak to be more dominant. Typically, this two times belt frequency peak will be very near
the turning speed of one of the components on the machine. Depending upon the resolution and the maximum
frequency used when collecting data, you may not be able to differentiate the two peaks.
This equation is used to determine the resolution of each spectral cell.

The equation below is used to estimate the amount of time needed to collect the data. This time does not include
autoranging.

Another common problem with belts is the alignment of the two sheaves. You could have offset or angular
misalignment, much like the misalignment of two shafts. Again, a high axial reading would be a good indicator
of this problem.

A helpful hint is to collect one of the radial readings in line with the belt. This will help in analyzing eccentricity
of the sheaves.

The spectrum above comes from the vertical direction on the drive motor. The turning speed of the belt
calculated at roughly 24.5 Hz. Turning speed means the speed at which the belt actually moves around the two
sheaves. The spectrum may show a peak at the primary belt frequency. Note that in this case, the amplitude of
the 2 X belt frequency peak is higher, because the defect on the belt hits both sheaves during each belt
revolution. The peak caused by the turning speed of the motor appears at 59.2 Hz. A peak caused by the turning
speed of the fan appears at 113 Hz. The vibration caused by the defective belt clearly adds a lot of energy into
the system.

The spectrum above comes from a permanently mounted transducer located on the fan housing. The plot shows
a very high 2 X belt frequency. The 2 X belt frequency shows a peak amplitude similar to the peak caused by
the unbalance of the fan. Removing the belt problem would eliminate almost half of the energy going into the
system.
The time waveform from the motor position shows a modulated pattern of vibration. This type of pattern
commonly occurs on belt driven equipment. The large cycle represents the difference in frequency between the
turning speed peak on the motor and the 2 X belt frequency peak.

Electrical Faults

Electric motors can experience many of the mechanical problems presented elsewhere in this manual such as
unbalance, misalignment, looseness, eccentricity, and bearing defects. Each mechanical problem exhibits the
characteristics covered in the section that specifically deals with that defect. Vibration transducers easily detect
such mechanical problems.

Purely electrical problems, however, occur because of the electromagnetic fields associated with electric motors.
The various electrical defects also present distinguishing characteristics. Unequal electromagnetic forces act on
the rotor or stator to cause vibration, so vibration transducers can detect many of these defects. This section,
therefore, looks at the vibration characteristics of defects in electric motors.

You can measure electrical defects with several different transducers. CSI also offers the Model 341 Current
Probe for use in further diagnostics, but this section does not review the use of that probe.

Sources of rotor vibration include:

1. broken or open rotor bar that shows predominant vibration at shaft turning speed with sidebands spaced at a
frequency equal to the number of poles on the motor multiplied by its slip frequency

2. loose iron or slot that shows predominant vibration at 2 X electrical line frequency (2XLF) and rotor bar pass
frequency (the latter frequency has sidebands spaced at 2XLF)

3. eccentric rotor that appears at 1XTS with sidebands spaced at slip frequency and/or 1XLF or 2XLF

Note:

slip frequency = magnetic field freq. - rotor freq.

rotor slot frequency = # of rotor slots X rotor freq.

Sources of stator vibration include:

1. loose stator laminations that appear at 2XLF and which may also show harmonics of 2XLF

2. open or shorted windings that appear at 2XLF and which increase in vibration as motor temperature climbs

3. insulation breakdown that shows up at 2XLF

4. unbalanced phase that appears at 2XLF

When you suspect an electrical problem, check the vibration on the motor the instant after you turn off the
electric power. (If using the CSI Model 2110/21 15 Machinery Analyzer, use Monitor Waveform.) If the signal
drops instantly, you have an electrical defect. Note that electrical problems appear in the radial direction except
on motors away from their magnetic centers.

If a three phase motor is single phasing, then the indications will be a significant peak at two times line
frequency with sidebands spaced at 1/3 line frequency. This is a serious problem and should be corrected
promptly.

Journal Bearings

Excessive clearance, improper bearing load, and improper lubrication can each result in high vibration levels in
journal bearings.

A journal bearing with excessive clearance allows a small excitation force, such as a slight unbalance or
misalignment, to cause significant vibration in the bearings. The predominant frequency of the vibration can
occur at 1XTS, 2XTS, 3XTS, or even higher harmonics, depending on bearing design and application. Collect
data in the radial and axial directions. Radial readings usually provide the best information on plain bearings.
Compare both vertical and horizontal readings. The vertical reading usually gives the best indication of
excessive clearances in a journal bearing. Axial readings are best for thrust bearings
.
Oil whip or whirl occurs when the oil film in pressure-lubricated, sleeve-type bearings exerts a force that pushes
the shaft around within the bearing. During spectral analysis, you can detect oil whip at less than one-half the
shaft speed. Under normal operating conditions, the shaft rides up the side of the bearing on the oil film wedge.
Because of friction, the oil film speed approximates only 42% to 47% of the shaft speed. The oil film force,
however, usually remains very small when compared to normal forces in the machine. If you perfectly center a
shaft in a bearing, the machine will not become susceptible to oil whip. Of course, the shaft can become
eccentric within the bearing because of improper bearing design, improper loading, or excessive bearing wear.
The oil film force can then become the dominant force within the machine.

Bearing wear can also make the machine more susceptible to oil whip, because the shaft rides farther away from
the bearing center. Look for sleeve bearing wear at 1XTS, 2XTS, or higher multiples. Tilted pad bearings
usually show wear at the frequency equal to the number of pads multiplied by the shaft speed. Changes in either
lube oil pressure or viscosity likewise increase susceptibility to oil whip. You can often correct oil whip either
by properly loading the bearing or by changing one or more of the following: bearing design, oil viscosity, oil
pressure, or the oil injection point. Several types of sleeve bearing designs minimize the effect of oil whip.
These bearings include the axial grooved bearing, tilted pad bearing, and lobed bearing. These bearings have
surfaces that form multiple oil film wedges in an attempt to center the shaft within the bearing.

The peak hold averaging mode is very beneficial when dealing with oil whirl in a journal bearing. The oil whirl
frequency will fluctuate over time and if normal averaging is used, then it gets averaged over this period and
may not appear to be of any serious concern. However, if the highest amplitude is maintained and compared to
the amplitude of the turning speed, you may find that the oil whirl is significant, just as in the following case.

Oil whirl peaks appear prominently at 0.4XTS on the turbine outboard bearing positions--TOH and TOV. The
highest vibration levels appear at 1XTS on the inboard bearing positions of the turbine--TIH and TIV. Note that
1XTS peaks normally have the highest amplitudes. The 0.4XTS peak also appears on the generator positions--
GIH and GOH. Some peak should normally show up at about 0.4XTS, but it should remain low and steady in
amplitude. Watch for steadiness by watching a live time spectral display.
The spectra above show the effects of peak-hold averaging on the outboard turbine points. The first and third
(from the bottom) come from the regular route mode. The second and top spectra come from the use of peak-
hold averaging. Peak-hold averaging keeps the highest value measured among all the averages for each line of
resolution. This method reveals that the amplitude for the oil whirl peaks surpassed that for the 1XTS peaks.
Even though the amplitude at both frequencies are relatively low, but 0.4XTS peaks that exceed 1XTS peaks
indicate a significant problem.

The spectrum above shows data for the turbine outboard horizontal point--TOH. The data comes from routine
route data collected with normal averaging. The height of the oil whirl peak causes concern, because its
amplitude matches that of the 1XTS peak. When viewed in live time, this oil whirl peak appears very erratic in
amplitude. It sometimes appears much higher than 1XTS, and at other times it almost disappears. An erratic
amplitude characterizes an oil instability problem.

The plot above shows the time waveform for the turbine outboard horizontal point--TOH. The vertical lines
denote the time required for each revolution of the shaft to occur. It basically shows a non-repeating pattern.
Every two to three revolutions of the shaft, a peak caused by the whipping motion of the oil becomes visible.
The RMS value of the time waveform is 0.4613 mils peak-to-peak, but the true peak-to-peak value of the time
waveform appears to exceed 0.7 mils.
Resonance
Every mechanical structure has at least one characteristic frequency (and sometimes more than one)
which is called its critical or resonant frequency. When excited by an external force, the mechanical structure
tends to vibrate at its resonant frequency. Less damping occurs at the resonant frequency than at other
frequencies. Therefore, vibration that occurs at this frequency becomes amplified. You can often observe higher
levels of vibration at a machine's resonant frequency than at other frequencies. These vibration levels, however,
decrease over a machine's operating lifetime.

When you strike a bell or a tuning fork, it rings at its resonant frequency. Likewise, a machine "rings" at its
resonant frequency when a force such as misalignment or unbalance occurs. Therefore, an impact test to
determine a machine's resonant frequency is also sometimes called a "ring test".

Stiffness, mass, and damping combine to determine a machine's resonant frequency. Changing any one of these
three factors modifies the machine's resonant frequency. In turn, this alteration may help solve a resonance
problem on the machine.

By increasing the mass of an object, the resonant frequency is decreased. Increasing the stiffness will increase
the resonant frequency. Changing the damping has no effect on the resonant frequency itself, but it does change
the amplification factor.

When passing through the resonant frequency of a machine, the phase data will change by approximately 180
degrees (+/- 30). Another characteristic is that the amplitude levels will increase due to the amplification factor
of the resonant frequency.

Some rather simple field tests may be performed to determine the resonant frequency of a machine. One has
already been mentioned, and that is an impact test. While the machine is not running, you would impact the
machine while collecting your vibration data. The impact excites all the frequencies, but only the resonant
frequency will continue to ring.

Another test is to perform a coast down or startup on the machine. While doing this, you monitor the peak and
phase for the increase in amplitude and phase shift. This will indicate where the resonant frequency is located.
You could also capture transit data or cascade data. The output of this data would indicate by the increased
amplitudes where the resonant frequencies are.

By changing the averaging mode, you can perform an impact test while the machine is still running. This is a
powerful tool since operations do not always want to shut down the machines. The impact test is performed in
the same manner, insuring that the machine is impacted once per average. Once the set number of averages has
been collected, then stop impacting the machine and subtract the operating data from the other. You will be left
with the impact data only.

Conclusion

The maintenance philosophies utilized today have evolved over time. Much of the technology that we use today
has been around for some time, we are just fine tuning the equipment. Vibration analysis is primarily limited by
our imagination. This technology can be applied to electric motors, pumps, turbines, fans, compressors, chillers,
etc. At this point in time, all of this may seem like magic, but as you gain valuable experience it will all become
more logical. Good luck!
Appendix A

Rathbone Chart

The chart above was made based upon casing measurements taken on heavy, slow speed machines.
Appendix B

Overall versus Selective Bands

About a decade and a half ago, the vibration equipment that was available supplied only an overall reading,
much like the trend above. This overall reading was a single number used to try and determine the condition of
the machinery. As was mentioned in the text of the article, this has been proven to be inaccurate. The overall
level is still utilized in the analysis process as an indication of the amount of energy contained in the system. In
the above figure, the overall level trend shows that this machine passes all the preset alarm levels and should be
in good condition. Is it?
In this figure, you can see the collected data for the last four collection periods. This PDM program is monthly-
based. Note the high 1xTS peaks in all the spectra. The level at 1xTS was higher in the first reading and dropped
in amplitude over the next three sets of data. Looking at the last set of data collected in October of 1989, you
can see that there has been a drastic increase in activity since the previous reading. This alone should be proof
enough that the overall value by itself is not very accurate in determining the condition of the machinery.
Having the ability to divide the overall value into selected frequency bands for more discrete alarming and
analysis provides a very powerful tool for vibration analysis. The data shown in the figure above reflects the
selective band for the Subharmonic & 1xTS band. Each individual band has its own alarm limits established.
You can see that the alarm levels displayed above and those on the previous trend plot are different. Here, the
1xTS peak has decreased in amplitude over the four months of data collection. This peak is a result of
misalignment, and as the bearings wear and relieve the misalignment the 1xTS peak diminishes.

In the above trend plots, the bearing bands are being displayed. Here you can see that there has been some
significant changes take place. The level of energy has drastically increased since the last months data was
collected. In both plots, note that the levels of vibration have triggered alarms.
The actual single spectrum from October 1989 is displayed above. There is significant energy at the higher
frequencies. Several of the peaks have 1xTS sidebands and the peaks are nonsynchronous. All of these
characteristics point to a possible bearing defect (which it was) yet the overall trend shows that everything
should be okay. You be the judge!

Appendix C

Transducer Advantages and Disadvantages

Displacement Transducers

Advantages:

1. Reads displacement.
2. Measures static displacement.
3. Measures relative motion.
4. Operates over a wide frequency range.
5. Inexpensive.
6. Small and lightweight.
7. Research and development is ongoing.

Disadvantages:

1. Reads displacement.
2. Requires a stable datum plane.
3. Requires an outside power supply.
4. Output affected by material.
5. Output affected by magnetic spots.
6. Dynamic signal rides on DC voltage.
7. Limited at high frequency.
8. Double differentiation to cross all vibration parameters.
9. Should be calibrated each time for the target material.
Velocity Transducers

Advantages:

1. Reads velocity.
2. Self-generating.
3. Excellent signal-to-noise ratio.
4. Rugged.

Disadvantages:

1. Large and heavy.


2. Limited frequency range.
3. Limited temperature range.
4. Some effect from magnetic fields.
5. Output varies with measurement position.
6. Relatively expensive.
7. No research to improve design.

Accelerometers

Advantages:

1. Reads acceleration.
2. Small.
3. Light weight.
4. Relatively inexpensive.
5. Wide frequency range.
6. Research to improve.

Disadvantages:

1. Reads acceleration.
2. High input impedance to charge amplifier.
3. Limited signal-to-noise ratio.
4. Requires outside power.
5. High frequencies can swamp element.
6. Double integration required to cross all vibration parameters.

You might also like