An Introduction to Time

Waveform Analysis
In recent years there has been a resurgence in the use of time waveform analysis techniques.
Condition monitoring personnel have now come to realize some of the limitations of the FFT process.
Since many find the time waveform analysis process difficult and confusing the technique is rarely
used to its full potential. The key to the successful utilization of time waveform data is knowing when
to use it. This paper examines the limitations of the FFT process and identifies specific applications
where enhancing FFT information with time waveform analysis is appropriate.

The paper also details in practical terms how to set up, acquire and manipulate time waveform data.
Having discussed the acquisition of the data the paper continues to discuss the interpretation of the
data including the time-frequency relationship, symmetry, and pattern recognition of common faults.

Biography

Timothy A. Dunton Vice President Product Development

Born and Educated in the UK, Tim spent 14 years with Shell Tankers UK. Ltd. as a marine engineer,
serving on Oil and LNG tankers in a variety of positions including Cargo Engineer. After emigrating to
Canada Tim brought his practical experience to IRD Mechanalysis as a Consulting Service Engineer,
providing training, software support and on site analysis services throughout Canada.

In 1991 Tim joined Update International as a senior instructor where he conducted various vibration
and skills related courses in both the public and in-plant arenas throughout the world. During his
tenure at Update Tim assisted in development of several seminars in the subjects of vibration analysis,
machinery skills and bearings. Tim was also responsible for upgrading several seminars and
developing award winning multimedia interactive software. Tim is a regular speaker at various
maintenance conferences throughout the world. Now with Universal Technologies, Tim’s primary
responsibility is in the area of multimedia product development but he continues to be active in the
teaching arena.

Introduction

The analysis of time waveform data is not a new technique. In the early days of vibration analysis
time waveform data was viewed on oscilloscopes and frequency components calculated by hand. The
relationship between frequency and time is as follows:

f = 1/p

where: f is the frequency in Hz p is the period in seconds (the amount of time required to complete 1
cycle)
Knowledge of this relationship permits the determination of frequency components from the raw
waveform data.

For example:

The above waveform was acquired from a 1785 RPM pump. The time spacing between the impacts is
0.0337 seconds. From this information the frequency can be determined.

f=1/p = 1/0.0337 = 29.67Hz = 1780CPM

This indicates that the impact is occurring at a frequency of 1 x RPM.

In most situations the time waveform pattern is very complex as illustrated below and therefore the
determination of frequency components is extremely difficult using this method and is not
recommended.

In most situations time waveform data is best utilized by applying the principles of pattern recognition
and if necessary calculating the frequency components of the major events in the waveform pattern.

When to Use Time Waveform

Time waveform can be used effectively to enhance spectral information in the following applications:

 Low speed applications (less than 100 RPM).

 Indication of true amplitude in situations where impacts occur such as assessment of rolling
element bearing defect severity.

 Gears.

 Sleeve bearing machines with X-Y probes (2 channel orbit analysis)

 Looseness.

 Rubs.

 Beats

When NOT to Use Time Waveform

Time waveform can be applied to any vibration problem. In some situations normal spectral and phase
data provide better indications as to the source of the problem without the added complexity of time
waveform data. Examples include:

 Unbalance on normal speed machines

 Misalignment on normal speed machines

Instrument Setup for Time Waveform

The key to successful analysis of time waveform data is in the set up of the instrument. The following
items have to be considered when setting up the instrument

 Unit of measurement

 Time period sampled

 Resolution

 Averaging

 Windows

Units of Measurement

Amplitude measurement units should be generally selected based upon the frequencies of interest.
The plots below illustrate how measurement unit selection affects the data displayed. Each plot
contains 3 separate frequency components of 60Hz, 300Hz, and 950 Hz.
This data was taken using displacement note how the lower frequency at 60 Hz is accentuated.

The same data is now displayed using velocity note how the 300Hz component is more apparent.

The same data is now displayed using acceleration note how the large lower frequency component is
diminished and the higher frequency component accentuated.
The unit of measurement displayed in time waveform data should be the natural unit of the transducer
used. For example if a displacement reading is required, then a displacement transducer should be
used. In most cases where modern data collectors are employed this means that acceleration will be
the unit of choice. If data is gathered from non-contact probes on sleeve bearing machines
displacement is usually used.

Time Period Sampled

For most analysis work the instrument should be set up to see 6-10 revolutions of the shaft being
measured. The total sample period desired can be calculated by this formula

Total sample period [seconds] = 60 x # of revolutions / desired RPM

The following table Illustrates common time period in seconds by machine speed

Time period for 6 Time period for 10

Machine RPM
revolutions (secs.) revolutions (secs.)
3600 0.1 0.167
1800 0.2 0.333
1200 0.3 0.5
900 0.4 0.667
300 1.2 2.0
100 3.6 6.0

Some instruments do not permit the setting of time period data when acquiring time waveform data.
With these instruments it is necessary to set an equivalent FMAX setting. The appropriate FMAX
setting can be calculated by the following formula

FMAX [CPM] = Lines of Resolution x RPM / # of Revolutions desired

The following table Illustrates the common FMAX settings for 1600 lines of resolution by machine RPM

Machine RPM FMAX for 6 revolutions FMAX for 10 revolutions

3600 960kCPM 576kCPM
1800 480kCPM 288kCPM
1200 320kCPM 192kCPM
900 240kCPM 144kCPM
300 80kCPM 48kCPM
100 26kCPM 16kCPM

Resolution
For time waveform analysis it is recommended that 1600 lines (4096 samples are used). This ensures
that the data collected has sufficient accuracy and key events are captured.

Averaging

In most data collectors averaging is performed during the FFT process. Unless synchronous time
averaging is invoked the time waveform presented on the screen will be the last average taken even if
multiple averages are selected in the instrument setup. It is normal therefore to take a single average.
Overlap averaging should be disabled. Synchronous time averaging can be used to “synchronize” data
acquisition to a particular shaft. This can be useful on gears where broken teeth are suspected to
assist in the location of the defective teeth relative to a reference mark. It is also useful on
reciprocating equipment to “time” events to a particular crank angle.

Windows

Various windows can be applied to the time waveform prior to performing the FFT. The purpose of
these windows is to shape the spectrum and minimize leakage errors. Some instruments can apply
these windows to time waveform data as well. This would force the data to zero at the start and end
of the time sample potentially losing data. To eliminate this effect a uniform or rectangular window
should be applied.

Interpretation of Waveform Data

Unbalance

The classic sine wave illustrated above is rarely seen in acceleration time waveform. This is because
acceleration emphasizes the higher frequency components that are almost always present in the
vibration signal. This de-emphasizes the underlying lower frequency signal.

The waveform below is more representative of sinusoidal vibration when viewed in acceleration. Note
the high frequency components superimposed on the lower frequency.
Misalignment

Although the classic symptoms of misalignment are M and W shapes in the time waveform, these
symptoms cannot be relied upon. The relative phase angle between the 1 x RPM and 2 x RPM
components determines the shape or pattern of the plot.

The pattern above illustrates the classic pattern of misalignment. In the pattern below the relative
phase between 1 x RPM and 2 x RPM was changed 90 degrees resulting in a very different pattern.
The pattern below originates when the 1 x, & 2 x, vibrations are 0 degrees apart.

Amplitude Symmetry

When observing time waveform data symmetry above and below the centerline axis is important.
Symmetrical data indicates that the machine motion is even on each side of the center position. Non
symmetrical time waveform data indicates the motion is constrained possibly by misalignment, or
rubs.
This waveform pattern is symmetrical above and below the zero line.

The following waveform pattern is non-symmetrical above and below the zero line. The amplitudes
below the line are significantly higher than those above the line. In this case a misalignment condition
was the source. The markers on the plot indicate 1 x RPM.

Symmetry of the Time Axis

When the previous time waveform is observed with 1 x RPM markers present it can be noted that the
waveform pattern although complex is repetitive with 1 x RPM. This indicates that the vibration is
synchronous to RPM.
The waveform pattern below indicates a non-repetitive pattern characteristic of non- synchronous
vibration.

This example, is of two frequency sources that are not harmonically related. (58 Hz and 120 Hz) This
is the kind of signal that could be created when a 2 pole motor has an electrical hum problem.

Higher frequency component “Riding” on other wave

It can be seen that the higher frequency wave does not always start at the same part of the lower
frequency cycle and therefore appears to “ride” on the other wave causing symmetry to be lost.

Care must be exercised when determining symmetry of the time axis. 1 x RPM markers are available
in most software programs and should be used to avoid confusion.

At first glance this waveform appears to have large impacts occurring with somewhat similar spacing.
The horizontal axis is scaled in time units.

By using RPM as the horizontal axis and applying 1 x RPM markers the major impacts can be seen to
be occurring at approximately the same part of the revolution. However closer inspection reveals that
the spacing is not exactly synchronous In this case the problem was a single large defect on the inner
race of a bearing. the change in amplitude of the defect was due to the defect coming in and out of
the load zone
This is the FFT taken from the above machine note the highest amplitude at BPFI is< 0.05 ips !
(15g pk @ BPFI = 2.13 ips)

Beats & Modulation effects

Another excellent application for time waveform is the observation of beat frequencies and modulation
effects. Often these phenomena are audible. The time span for data collection should be set to capture
4-5 cycles of the beat.

The time period between the beats on the above waveform is 0.5 s. From this information the
frequency of the beat is calculated at 120 CPM. This represents the frequency difference between the
two source frequencies In this case the beat was caused by interaction between a 2 X RPM vibration
source and a 2 x fL vibration source on an induction motor.

Impacts

When the FFT process is applied to a signal that contains impacts the true amplitude of the vibration is
often greatly diminished. The following time waveform was taken from an 1800 RPM machine. It
shows several random impacts with magnitudes over 6 g pk. The cause of this signal was a failed
rolling element bearing. The shape of the waveform often appears to be a large spike followed by a
“ring down”.

The plot below was a velocity spectrum taken from the same bearing note the amplitude of vibration is
less than 0.04 ips!

Care must be exercised when assessing the amplitude severity of 1 x RPM impact vibration using
spectrum. This is the spectrum of a machine with where the key is impacting the coupling guard. The
amplitude scale indicates amplitudes of less than 0.02 ips.
This is the time waveform from the above machine the amplitude of the impacts exceeds 0.15 ips.

In this case severe damage had occurred to the key and shaft of the machine in question.

Conclusions

In conclusion, how is this information practically applied in a condition based maintenance program?
  Time waveform analysis is an analysis tool. The writer would not recommended that it be
taken on all measurement locations on a regular basis. This would add significantly to the time
required and data storage requirements.

 Use Time waveform for the following selected analysis situations to enhance FFT information

 Low speed applications (less than 100 RPM).

 Indication of true amplitude in situations where impacts occur such a assessment of
rolling element bearing defect severity.
 Gears.
 Sleeve bearing machines with X-Y probes (2 channel orbit analysis
 Looseness.
 Rubs.
 Beats
 Impacts

 Use an appropriate measurement unit

 Rolling element bearings, gears, looseness, rubs, impacts ...acceleration

 Sleeve bearing machine with x-y probes…displacement

 Initially set up to observe 6 -10 revolutions of the shaft in question

 Study the following symptoms in the waveform

 Amplitude
 Amplitude Symmetry
 Time Symmetry (use RPM markers)
 Beats / Modulation
 Impacts (shape and amplitude)

Remember use time waveform to ENHANCE not REPLACE spectral data.

 Time Domain Plots

The dreaded time domain plot -

a powerful tool, but a number of questions for the analyst:

 How is it interpreted ?

 How is one set up ?

 Should I use it everywhere ?

 Why do I need it when I have the spectrum?

This section of the manual will provide you with information on how to use and interpret the time domain plot. The time domain, of
course, is where the reading begins - an analog measurement of how the surface is moving. This analog signal is fed from the
transducer to the analyzer where it is converted to a digital signal - it goes through an A/D converter. The result of this process can
be seen above in the plot above.

It is important to realize that it is experience (i.e. practice) that creates a 'comfort' level for the analyst in interpreting the time domain
plot. Experience in setting it up properly and experience in being able to recognize what you are seeing - the pattern of what you are
seeing. Looking at Figure 1, what do you see ? A fuzzy caterpillar (from about 1 inch away) - looks like a rough winter is coming. But
let's get closer and see what else we can see.

Time Domain Plots and The FFT

Let's go back to that nice clean sine wave you see all the time - at least all the time you are sitting in a training class. That is what
you see in Figure 1. Performing an FFT on Figure 1 would generate the plot you see in Figure 2 - a single peak at 1x rpm.
 Figure 1

Figure 2
The FFT process sees a simple sinusoid and calculates the period (time required) for a single cycle. In Figure 1, the period is 40
msec (0.040 seconds). Then, do the following:

 Period = 40 msecs per cycle [0.04 secs / cycle]. We have seconds per cycle - we want cycles per second (Frequency).

 Frequency = 1 / 0.040 = 25 cycles per second [25 cps or 25 Hz]. If we want cycles per second:

 Frequency = 25 Hz x 60 seconds / minute = 1500 cycles per minute [CPM].

The FFT is created with a peak at 1500 cpm or 25 Hz. The amplitude shown will be based on the Window type shown and whether
you have a signal detection of RMS, peak, peak to peak or true-peak.

Unfortunately, in the mechanical world there are only two problems that cause such a pure sinusoid to occur (and it will only be pure
if they are the only problems present):

 Unbalance

 Resonance

Time Domain Plots and Frequency Modulation

But could there be another explanation for the signal shape seen on those pages ? Let's return to our discussion of the actual, real-
life vibration signal we looked at a few pages back. We discussed how there can be some variation in the free rotation of the shaft -
a momentary 'binding' action that can occur for a number of reasons. In that situation, we considered the possibility of the gears
being improperly set. That would create more resistance to rotation when the teeth were bottomed out than opposite that point. It
would momentarily slow down the rotation.
 Let's examine the 'frequency modulated' signal shown in Figure 1:

 Let's first examine the positive going peak. The bottom of the cycle (the '-' peak) occurs at about 19 msec and the '+' peak at
about 32 msec - a total of about 13 msec. The reciprocal of the period will give us the frequency: 1/(0.013 x 2) x 60 = 2308 cpm.

 Now let's examine the negative going peak. From the '+' peak at 32 msec, the signal descends to a '-' peak at about 53 msec - a
total of 21 msec. That's a frequency of 1/(0.021 x 2) x 60 = 1429 cpm.

 The total for one cycle (peak to peak) is from 19 msec to 53 msec - about 34 msec. That calculates to about 1764 cpm.

This is frequency modulation. What is happening here may or may not be evident if we were to analyze the time domain signal. But
remember, the question we are discussing here is: How will the FFT treat this phenomenon ?
 Time Domain Plots and Amplitude Modulation

We have discussed
frequency modulation and its impact on the spectrum plot - namely, it creates harmonics. But we have also touched on amplitude
modulation - now let's cover it more in-depth. Amplitude modulation is a increase and decrease in the amplitude of a particular
frequency at a different frequency. That's simple enough, right ? But what effect does it have on the FFT ?

Let's look at some examples. What do you make of the time domain plot taken on a gearbox shown here:

This is a case of a low frequency cycle (occurring 15 times over the time sample) and a high frequency (occurring many times for
each of the low frequency cycles). This is known as 'a high frequency riding a low frequency'. For analysis, let's zoom in:

This shows only about one of the low

frequency cycles. The high frequency could be a gear meshing frequency. The low frequency is at 1x rpm. How many teeth are on
the gear ? This is another advantage of using time domain on gearboxes - you can actually obtain detailed internal information that
you can only guess at on the spectrum. Count the small peaks from the top of one low frequency peak to the next. There are 23
teeth. What does the spectrum look like ?

Both the amplitudes and the frequencies are constant - there is no modulation in either - so you only get the peaks that are actually
being generated. Now let's consider some variations on this "perfect" gearbox.

Applications where everyone should use time domain

Obviously, time domain is a powerful tool. But what kinds of problems and situations are better analyzed with the time domain plot ?

 Slow Beat  Chipped / Broken Gear

 Frequencies Teeth

 Impacts  Amplitude Modulation

 Transients  Frequency Modulation

 Rubs  Slow Speed Bearing

Defects

Additionally, you can gather information related to the machine condition such as:

 Misalignment  Eccentricity

 Looseness  Waveform
Shape
 # of Teeth on a
Gear  Instabilities