Time Domain Analysis of

Vibration Data
Using Time Domain Signals to Further Diagnose
Machinery Health

Summary
This article provides valuable information on using time domain
signals to further diagnose machinery health and confirm
problems detected with frequency domain analysis. Moreover,
time waveform analysis may be the only method available to
detect certain types of problems, such as cracked and broken gear
teeth or rolling element bearing defects on low speed
applications.

Jason Michael Mais

10 pages
September 2002

SKF Reliability Systems

@ptitudeXchange
4141 Ruffin Road
San Diego, CA 92123
United States
tel. +1 858 244 2540
fax +1 858 244 2555
email: [email protected]
Internet: www.aptitudexchange.com

Use of this document is governed by the terms

and conditions contained in @ptitudeXchange.
 Time Domain Analysis of Vibration Data

Introduction particular problem that may otherwise have

been a best guess scenario.
Studying time domain analysis is very useful
to the overall understanding of what is There are some applications where the need
occurring in the frequency domain. Time for time domain data is absolutely mandatory.
domain is the actual data received from the Some applications that are mandatory
machinery, and further processed through concerning time waveform analysis are:
Fourier Transform formulas to arrive at the
frequency domain. • Cracked, broken, or deformed gear teeth in
gearboxes.
There are many useful components of time
domain that give insight into machine health. • Rolling element bearing defects on very
The time domain discerns actual frequencies low speed machines (<10 RPM).
and amplitudes of components within a
machine. • Motor startup transient problems that
result in bearing deterioration and winding
In general, the time domain is nothing more problems.
than a record of events as they happen. If we • Reciprocating compressors: short-lived
setup a signal generator in the lab and produce impact type vibration problems such as
a sine wave we can see it in the time domain piston slap, loose rod and main bearing
spectrum just as it is seen on the screen of an and inlet or discharge valve problems.
oscilloscope.
This paper illustrates three cases: two
Complications arise when the technology is gearboxes and a low speed bearing. There are
applied in the real world. A sine wave, simply many documented cases where, without time
stated, is a single frequency, and therefore a domain information, the problem could not be
solitary signal. In reality, a machine does not solved or time and cost to repair a machine
exhibit a solitary signal. For example, an would have been extended.
operating motor connected to a gearbox, and
then to a compressor produces thousands or Several arguments arise in which a
millions of signals that add and subtract to / timesavings or the ability to store the data is
from each other based upon their relationships sighted as the cause for not collecting the time
and the influence of external forces. That is all domain signal. Saving time is certainly a valid
discerned from the time domain signal, which point to consider, but the key to analyzing
at first glance appears to be a series of machinery is to collect data that helps resolve
squiggly lines of various amplitudes and the problem. If needed data is not collected
frequencies (Figure 1). when analyzing a machine, the time required
to recollect the data or make a decision based
Time domain analysis is much more than a upon a lack of key data can cost even more
series of squiggly lines. It gives us the resources and time. And with the use of
opportunity to detect several problems that modern computer storage systems, the few
can only be seen when collecting and bytes of additional information needed do not
analyzing this type of data. Time domain impact the storage capacity of the system
analysis can also add a factor of confidence to negatively. Therefore, storage capacity is not a
the analysis of data in the frequency spectrum. valid argument.
In some instances it can help confirm a

© 2002 SKF Reliability Systems All Rights Reserved 2

 Time Domain Analysis of Vibration Data

Figure 1. Typical time domain signal of an Enveloped Acceleration reading with various frequencies coming
together to display the overall signal of the machine at a particular point in time.

Factors for Interpreting the Time As the Fmax increases, the time span is
Domain Spectrum decreases, and as the Fmax decreases, the time
span increases. For example, when collecting
Once a time domain spectrum can be readily a 400 line spectrum, if the Fmax is set to 10,000
identified, the next step is to become familiar RPM the time span captured is 2400
with its components and interpret their various milliseconds, or 2.4 seconds. When Fmax
meanings. The time domain spectrum is made increases to 20,000 RPM, the time span
up of two axis planes: the X-axis is time and adjusts to 1200 milliseconds or 1.2 seconds.
the Y-axis is amplitude. Conversely, if we decrease the Fmax to 5000
RPM, the time span adjusts to 4800
Let’s first define the X-axis. The X-axis can milliseconds or 4.8 seconds. So, at an Fmax of
be equated to a process of taking a photograph 5000 RPM we have a picture of the events
of what is occurring to a machine at any happening in the machine for the last 4.8
particular point in time. This time span, or seconds.
photograph in time can be adjusted by
controlling the upper frequency, or Fmax of the Another way to look at this is to remember
frequency spectrum. This is accomplished by how the carnival shooting gallery works.
controlling the time factor. Usually, targets are on a belt and appear at the
right side of the gallery. As they are drawn
Frequency = 1/T across the target area another target appears on
the right and begins to move. In a time
T = Time
domain, events or “targets” appear on the right

© 2002 SKF Reliability Systems All Rights Reserved 3

 Time Domain Analysis of Vibration Data

side of the screen and move to the left. As Enveloped Acceleration, as the enveloping
other events occur in the machine, they appear process takes place prior to the development
on the right side and move across the screen. of the time domain display.
At any given moment, what we see on the
screen is a moving summation of all the A final point: energy is the same in both the
events that occurred in the past 4.8 seconds. If time domain spectrum and the frequency
we freeze the collection and save the data we domain spectrum. Other than the True peak
have a block of events from the past 4.8 and RMS derived peak differences, they
seconds. We can examine this data for contain the same amount of energy. The
machine condition information. Fourier function allows us to see the time
domain rotated 90 degrees, and see each sine
The Y-axis is the amplitude of the signal or wave within the same plane. It’s as if a book
the amount of energy causing the squiggly were printed on clear plastic pages. Viewing
lines. For some types of analyzers the this book with the cover facing you, you see
amplitude of the signal in the time domain is all the letters in the book. This perspective, as
shown in true peak, or an actual 1-to-1 it relates to vibrational data, is the time
relationship of the signal shown compared to domain. Viewing the book by tipping it up 90
what is occurring on the machine. This is in degrees, you see the edge of each page, or the
contrast to the frequency domain, which is leading edge. This relationship, as it relates to
displayed in RMS derived peak. As many vibrational data, is the frequency domain.
experts note, the two can be extremely
different. Case 1: Ball Mill Gearbox
At this point I do not propose to delve into The following case history is a classic
RMS vs. True peak other than to say that if a example of using time waveform analysis to
reading of 10 g overall occurs in the time detect gearbox problems. The system contains
domain, and a reading of 2 g overall occurs in a gearbox that drives a very large ball mill
the frequency domain, the resulting 2 g overall used to crush ore. The owner drove the unit
signal in the frequency domain is due to the clockwise for 6 months, then reversed the
relationship of a small number of 10 g signals motor leads and drove the unit counter
that average out when processed by RMS. clockwise for 6 months to even out the wear.
Periodically the mill broke a tooth on the drive
Moreover, the amplitude of a signal in the or pinion gear. This occurred often enough
time domain is an acceleration amplitude. In that the company kept a spare pinion gear on
other words, the signal is not integrated. hand at all times. Gear mesh frequency at
Therefore, for example, a velocity spectrum elevated levels was present in the gearbox’s
isn’t actually a velocity measurement although frequency domain (Figure 2). The owner
the user collected the data in velocity. assumed that a broken tooth or teeth occurred
Actually, it is an integrated signal from the on the pinion gear and planned to replace the
time domain. However, if the user selects worn pinion during the outage scheduled for
Enveloped Acceleration, the time domain is in two weeks later.

© 2002 SKF Reliability Systems All Rights Reserved 4

 Time Domain Analysis of Vibration Data

Figure 2. Elevated levels of the frequency related to the pinion gear in a large ball mill. The elevated level of this
spectrum as compared to the trended level is an indication of a broken tooth on the pinion.

Looking at the spectrum in Figure 2, the teeth on the spare pinion, no other information
elevated gear mesh frequency shows a is available. We should also note that the
problem in the gearbox. However, it is speed, 600 RPM, is set as a marker on the
difficult to determine the source of energy. spectrum. With the cursor on the gear mesh
The two known facts regarding the gearbox frequency (GMF), a note in the upper left-
are: hand corner of the data window explains that
the “orders” of running speed at which the
• The input shaft speed is operating at 600 cursor is positioned is 36.05 orders. This tells
RPM. us that there are 36 teeth on the pinion gear, as
the GMF is the number of gear teeth, times the
• The output shaft speed is 210 RPM. rotational speed.
One can deduce, based upon the input and
output speeds, that the gear ratio is about 3:1.
However, without counting the number of

© 2002 SKF Reliability Systems All Rights Reserved 5

 Time Domain Analysis of Vibration Data

Figure 3. Time domain spectrum of the same data in Figure 2. The use of time spaced overlays is helpful in
calculating the relationships between peaks in the spectrum and the frequency that is related. The time spacing is
0.2874 seconds, which translates to 208.8 CPM.

If we further analyze the data we can resulting jerk or jump is captured as a burst of
determine more information about the gearbox energy. The time waveform analysis was
(Figure 3). The X-axis tells us that we necessary, as the GMF is not always that
collected data for 3 seconds. In that time visible in the vibration spectrum.
period, 11 bursts of energy, or targets
appeared on the screen. Using the harmonic Condition monitoring was able to save this
markers in the time domain, two cursors were customer over one million dollars by
generated. The left cursor is set on a providing information about the condition of
prominent energy spike. The right cursor is set his gearbox. The company did not have a
on the next prominent burst of energy. A spare bull gear on hand. If they found the
simple calculation of the spacing between problem during the outage, they could have
energy bursts determines the interval between lost the 2 ½ days of production required to
bursts is 0.2874 seconds, which is equal to a obtain another bull gear.
frequency of 208.8 CPM (F=1/T). Thus,
whatever is causing the bursts of energy is Case 2: High Speed Gearbox
occurring at the same rate as the rotation Located in a petrochemical plant, this
speed of the output shaft on the bull or driven compressor was connected to a gearbox that
gear. If the energy bursts occurred once every increased the input speed from 3600 RPM to
0.100 seconds the source would be the 600 7546 RPM. The operators complained about
RPM input shaft (one rotation every 0.1 sec). the increasing noise and vibration that began
shortly after the bull or driven gear was
This spectrum is a perfect example of a gear
replaced. Figure 4 depicts the gear mesh
with damaged teeth (singular or plural). As the
frequency on the high-speed gearbox of this
damaged area comes in contact at mesh, the
petrochemical plant.

© 2002 SKF Reliability Systems All Rights Reserved 6

 Time Domain Analysis of Vibration Data

Figure 4. A zoomed in view of the gear mesh frequency of the gear on the gearbox in a petrochemical plant. The
center frequency of the gear mesh is 401,322 CPM. The spacing of the higher amplitude sidebands is 7513 RPM,
which is the output shaft speed. The smaller sidebands are spaced at 210 CPM, which is also the output shaft speed.

For clarity, we zoomed in at the gear mesh gear attached to that shaft. If examined
frequency in order to easily measure the closely, the frequency spectrum in Figure 2
spacing of the sidebands. In this case, we can also exhibits very small sidebands located
use the band cursor and determine that the around 210 RPM, which is the same speed as
sidebands are equal to the output shaft speed: the output shaft.
around 7513 RPM. If desired, the lines of
resolution can be increased to improve the Figure 5 shows the time signal of the suspect
accuracy of the determined frequency. gearbox. As noted in the Y-axis, this data was
However, increasing the number of lines collected using enveloped acceleration, gE.
increases the time required to collect data. The These energy bursts appear to only go
preferred method is to accept the slight upwards from 0 amplitude, unlike the data
inaccuracy for improved speed and collect all collected in acceleration, which somewhat
data at 400 lines of resolution. An increase can symmetrical above and below the 0 line. The
be implemented if needed. Note that the above display is normal for some enveloped
presence of shaft speed sidebands around the spectra, as it is the result of the enveloping
gear mesh is a sign of serious damage to the process and is of no concern.

© 2002 SKF Reliability Systems All Rights Reserved 7

 Time Domain Analysis of Vibration Data

Figure 5. Further exploration into the time domain of suspect gearbox. Again, the spacing markers are displayed to
assess the time difference between peaks. In this instance, the time frame is 0.0079 seconds, which relates to a
frequency of 7559 RPM. This is the speed at which the output shaft rotates.

As before, we place the left marker on one increases. The GMF frequency is still
spike and the right marker on the adjacent calculated the same way: the number of teeth,
similar spike. The resultant time interval is times the shaft speed.
0.007937 seconds, which is equal to 7559
RPM. We now have two indications that there However, amplitude changes as damage
is a problem with the bull gear in this box. increases. Therefore, a trend chart may be the
first indication that a problem is developing.
Examination of the box reveals that 21 of 53
teeth are damaged on the bull gear. Further Case 3: Low Speed Bearings
investigation into the manufacturing process
Low speed bearings can be nearly impossible
of the gear revealed that the teeth were
to analyze if using only the frequency domain.
incorrectly cut when this gear was
The following case study involves a unit that
manufactured.
turned at 0.50 RPM, in which the operators
Note that the gear mesh frequency does not noted an increased noise level at the bearing.
change as the number of damaged teeth Fmax was chosen so 500 RPM gives us a
collection time of 48 seconds.

© 2002 SKF Reliability Systems All Rights Reserved 8

 Time Domain Analysis of Vibration Data

Figure 6. Data collected in the frequency domain of a failing bearing in a machine turning at 0.50 CPM. As can be
seen above, there are no predominant peaks in the spectrum that indicate a failing bearing.

What can be discerned from the spectrum in Using enveloped acceleration, these repetitive
Figure 6? Nothing! The very low amplitudes signals were enhanced and displayed. Using
of gE are normal for low speed applications. the harmonic cursors, the time interval was
The Ball Pass Frequency Outer Race, (BPFO) determined as 10.23 seconds, which gave us a
at this speed for this bearing was 5.9 RPM, frequency of 5.86 RPM (Figure 7). The
though there is no indication of this frequency bearing database calculated that at 0.50 CPM
in the spectrum. The collection limits were this bearing would have a BPFO of 5.9 RPM.
controlled operationally due to the speed of Considering the data obtained and the method
the application. Therefore, a reduction of the of analysis, it was agreed upon that the
Fmax or increase in the number of lines for bearing was damaged, and the
better resolution was not conductive to the recommendation to order a replacement was
operation. We arrived at an impasse. made. The replacement cost for this particular
bearing was $250,000. Confirmation of
Therefore, time domain data was examined. In bearing damage on the outer race was given
short, as each roller passed over the flaw in after the replacement bearing was installed.
the bearing, a burst of energy was generated.

© 2002 SKF Reliability Systems All Rights Reserved 9

 Time Domain Analysis of Vibration Data

Figure 7. Time domain signal of the 0.50 CPM application shown in figure 6. The time differential marker is shown
on the spectrum at a spacing of 10.23 sec. This translates to 5.86 RPM. The outer race of the bearing has a
frequency of 5.9 RPM.

Summary analyze the collected. Time domain data and

analysis is useful in applications such as
Time domain data is a valuable analysis tool gearboxes and low speed bearing applications.
to determine some of the factors affecting Using the time domain signal, the analyst can
your machinery. The methodologies of time determine components that are failing or
domain analysis and several case studies that processes that may otherwise go undetected
exhibit real world applications of time domain until machinery failure. In some cases, this
technology were explained. It is as important failure can cause significant financial
to collect the proper type of data, as it is to detriment to a process or plant

© 2002 SKF Reliability Systems All Rights Reserved 10