+1.616.669.

3028
[Link]
vrsales@[Link]

random vibration testing

John Van Baren | Vibration Research Corporation | Jenison, Michigan

There is some confusion about the various tests

available to the vibration testing engineer. Difficulties
encountered usually center on the differences
between sinusoidal vibration (sine testing) and
random vibration testing.
Strike a tuning fork and the sound you hear is
the result of a single sinusoidal wave produced at
a particular frequency (Fig 1). The simplest musical
tones are sine waves at particular frequencies. More
complicated musical sounds arise from over-laying
sine waves of different frequencies at the same
time. Sine waves are important in more areas than
music. Every structure can vibrate and has particular
frequencies (resonance frequencies) in which it
vibrates with the greatest amplitude. Therefore,
sinusoidal vibration testing is important to help
understand how any structure vibrates naturally. FIGURE 1. Time history of a sinusoidal waveform. Note its
The vibration testing industry has made good use repeatability and predictability.
of sine vibrations to help assess the frequencies at
which a particular device under test (DUT) resonates. dominant resonances of the tested equipment.
These frequencies are important to the vibration Repeating the sine test after otherwise abusing
testing engineer because they are the frequencies at a product should produce the same test results
which the DUT vibrates with the greatest amplitude unless the DUT has been damaged. Any differences
and, therefore, may be the most harmful to the DUT. in the sweeps indicate damage to the equipment –
Because “real-world” vibrations are not sinusoidal, perhaps something as simple as a shift in the natural
sine testing has a limited place in the vibration resonance frequencies that might suggest a few
testing industry. Part of the usefulness of sine testing loose bolts need to be tightened.
is its simplicity, so it is a good point of entry into the
study of vibrations.
RANDOM VIBRATION
Sine testing is used primarily to determine damage Vibrations found in everyday life scenarios (a vehicle
to structures. The best pro-sine arguments are to on a typical roadway, the firing of a rocket or an
search for product resonances and then to dwell on airplane wing in turbulent air flow) are not repetitive
one or more of them to determine modal properties or predictable like sinusoidal waveforms. Consider
and to determine fatigue life associated with each the acceleration waveform (Fig 2) for the dashboard
mode.1 vibration of a vehicle traveling on Chicago Drive near
Aside from testing a product to find and dwell Hudsonville, MI. Note that the vibrations are by no
at its resonance frequencies to determine fatigue means repetitive. So, there is an important need for
life, one might also use sine testing to determine tests that are not repetitive or predictable. Random
damage to equipment. A sine sweep prior to any vibration testing accomplishes this.
shock or random vibration test will identify the Random vs. Sine. Sinusoidal vibration tests are

Page 1 of 5
not as helpful as random testing, because a sine test
focuses upon a single frequency at any one time.
A random vibration test, on the other hand, excites
all the frequencies in a defined spectrum at any
given time. Consider Tustin’s description of random
vibration - “I’ve heard people describe a continuous
spectrum, say 10-2000Hz, as 1990 sine waves 1Hz
apart. No, that is close but not quite correct. Sine
waves have constant amplitude and phase, cycle
after cycle. Suppose that there were 1990 of them.
Would the totality be random? No. For the totality to
be random, the amplitude and starting phase of each
slice would have to vary randomly, unpredictably.
Unpredictable variations are what we mean by
random. Broad-spectrum random vibration contains
not sinusoids but rather a continuum of vibrations.” 1
Advantages of Random Vibration Testing. One of FIGURE 2. Acceleration time history collected on vehicle
dashboard while driving in Hudsonville, MI.
the main goals or uses of random vibration testing
in industry is to bring a DUT to failure. For example,
a company might desire to find out how a particular
product may fail because of various environmental
vibrations it may encounter. The company will
simulate these vibrations on a shaker and operate
their product under those conditions. Testing the
product to failure will teach the company many
important things about its product’s weaknesses and
ways to improve it. Random testing is the key testing
method for this kind of application.
Random vibration is also more realistic than
sinusoidal vibration testing, because random
simultaneously includes all the forcing frequencies
and “simultaneously excites all our product’s
resonances.” 1 Under a sinusoidal test, a particular
resonance frequency might be found for one part of
FIGURE 3. Typical power spectral density vibration testing
the device under test and at a different frequency specification (mean squared acceleration per unit frequency).
another part of the DUT may resonate. Arriving at
separate resonance frequencies at different times (acceleration squared per Hz versus frequency Fig
may not cause any kind of failure, but when both 3). The shape of a PSD plot defines the average
resonance frequencies are excited at the same time, acceleration of the random signal at any frequency.
a failure may occur. Random testing will cause both The area under this curve is called the signal’s
resonances to be excited at the same time, because mean square (g2) and its square root is equal to the
all frequency components in the testing range will be acceleration’s overall root-mean-square (RMS) value
present at the same time. often abbreviated σ.
A random test is conducted by using closed-loop
THE POWER SPECTRAL DENSITY
feedback to cause the random vibration of a single
FUNCTION (PSD)
location (typically the shaker table) to exhibit a
To perform random testing, a random test spectrum desired PSD. The PSD demonstrates how hard the
must be defined. Real-time data acquisition shaker is working, but it does not give any direct
utilizes spectrum-averaging to produce a statistical information about the forces experienced by the
approximation of the vibration spectrum. Generally, DUT. The g2/Hz PSD is a statistical measurement of
the random vibration spectrum profile is displayed the motion experienced at the Control point on the
as a power spectral density (PSD) plot. This plot test object - this is important to remember. Since
shows mean square acceleration per unit bandwidth

Page 2 of 5 What Is Random Vibration Testing

the PSD is the result of an averaging process, an
infinite number of different time waveforms could
have generated such a PSD. The idea that an infinite
number of real-time waveforms could generate a
particular PSD can be seen from the graphs shown
in Figures 4-7 produced from data collected at
Vibration Research Corporation, June 28 and 30,
2005. Note that the PSD curves (Figures 6 and 7) are
virtually identical although they were generated from
entirely different waveforms.

THE PROBABILITY DENSITY FUNCTION

(PDF)
An examination of the acceleration waveforms
of Figures 4 and 5 will indicate that much of the
random vibration acceleration values are nearly the
same (±5g). However, some of the acceleration values FIGURE 4. Sample acceleration time history of excitation applied
to Lightbulb-4 test, #1330.
are quite large compared to the most common
values. To help illustrate the range of acceleration
values, another statistic, the probability density
function (PDF) is required. A PDF is an amplitude
histogram with specific amplitude scaling. Each
point in the histogram is a count of the number of
times the measured signal sample was found to be
within a corresponding small range (an amplitude
“bin”) of amplitude.
The PDF shown in Figure 8 conveys the probability
of the signal being at a particular g-value at any
instant in time. Its vertical units are 1/g and the area
under this curve is exactly 1. Thus, the area under the
PDF between any two (horizontal axis) g values is the
probability of the measured signal being within that
amplitude range. Note that Figure 8 is drawn using
a logarithmic vertical axis. This serves to make the
extreme value “tails” more readable. FIGURE 5. Sample acceleration time history of excitation applied
to Lightbulb-4 test, #1110.
Various weighted integrals (moments) of this
function are determined by the signal’s properties.
For example, the first moment is the signal’s mean
(μ), the averaged or most probable value; this is
necessarily equal to zero for a random shaker Control
signal. The second moment is the signal’s mean
square (σ2) and it is equal to the area under the
previously discussed PSD. The third moment is the
signal’s skew, an indication of bias towards larger
positive or negative values; this is always equal to
zero for a random Control signal. The fourth moment
is called the Kurtosis and it measures the high-g
content of the signal.
The horizontal axis of an acceleration PDF has
units of gpeak (not RMS). This axis is often normalized
by dividing the g values by the signal’s RMS value s.
Many signals will exhibit a symmetrical “bell-shaped”
FIGURE 6. PSD spectrum for trial #1330.
PDF with 68.27% of the curve’s area bounded by

Page 3 of 5 What Is Random Vibration Testing

±σ and 99.73% within ±3σ. Such signals are said to
be “normal” or Gaussian. A Gaussian signal has a
Kurtosis of 3 and the probability of its instantaneous
amplitude being within ±3σ at any time is very nearly
100% (actually 99.73%).
There are many “real-life” situations where there
are more high acceleration values than a Gaussian
distribution would indicate. Unfortunately, most
modern random control techniques assume the
Control signal should be Gaussian with the majority
of the instantaneous acceleration values within
the ±3σ range. This assumption removes the most
damaging high peak accelerations from the test’s
simulation of the product’s environment and “under
testing” results. In fact, a Gaussian waveform will
instantaneously exceed three times the RMS level
FIGURE 7. PSD spectrum for trial #1110.
only 0.27% of the time.
When measuring field data, the situation can be
considerably different, with amplitudes exceeding
three times the RMS level as much as 1.5% of the
time. This difference can be significant, since it has
also been reported that most fatigue damage is
generated by accelerations in the range of two to
four times the RMS level.2 Significantly reducing the
amount of time spent near these peak values by
using a Gaussian distribution can therefore result in
significantly reducing the amount of fatigue damage
caused by the test relative to what the product will
experience in the real world. So Gaussian distribution
is not very realistic.
Hence, present-day methods of random testing
may be unrealistic for some simulations because they
fail to account for the environments most damaging
content. Furthermore, random testing with Gaussian FIGURE 8. Probability density function for a lightbulb test using
distribution will result in a longer time-to-failure Gaussian distribution (k = 3).
because the higher accelerations responsible for
failure have been omitted. Therefore, Gaussian
random testing, for all its advantages over traditional
sine testing, has its own disadvantages, and a better
method of testing products is called for. Vibration
Research Corporation has developed such a method.3
It is called Kurtosion®. Kurtosion allows a random
vibration test to be run with a user-specified kurtosis
of 3 or greater. This amounts to using feedback to
force the Control signal’s PDF to have broader “tails”.
That is, more intense peak accelerations are included
more often than using a Gaussian controller. This
method permits the adjustment of the kurtosis
level while maintaining the same testing profile
and spectrum attributes. With this new technique,
a random vibration test is best described by a PSD
and an RMS versus time schedule and an additional
FIGURE 9. Probability density function for
lightbulb test using Kurtosis Control (k = 5).

Page 4 of 5 What Is Random Vibration Testing

kurtosis value. The required kurtosis parameter can
be easily measured from field data. This is similar to
current random tests but is one step closer to the
vibrations measured in the field.
In Figure 8, the data set has a kurtosis value of three
(Gaussian distribution) and is a smooth distribution
with few large amplitude outliers. However, Figure
9 shows a data set with a kurtosis value of five.
Note how the “tails” extend further from the mean,
indicating a large number of outlier data points. The
contrast between the PDFs of a Gaussian distribution
and a higher kurtosis distribution is clearly seen in
Figure 10.
Therefore, the fundamental difference between a
Gaussian and controlled-kurtosis distribution is that,
although the two data sets may have the same mean,
standard deviation and other properties, the Gaussian
FIGURE 10. A comparison of kurtosis values 3 and 7. Note how the
data set has its data points closely centered on the higher kurtosis value includes higher peak accelerations.
mean, while the controlled-kurtosis distribution has
larger “tails” further from the mean. random airloads and engine harmonics. Random-
on-random tests superimpose narrow-band random
SOME OTHER TESTING OPTIONS
noise on broadband random noise. Such tests are
Modern test and measurement systems are blessed
claimed useful to simulate aircraft gunfire reactions.
with inexpensive memory. In recent years, it has
Both types of tests are designed to model a very
become feasible to record a long-time history and
specific class of environment and both are “tricky” to
then play it back as a shake-test Control reference.
set up.
Vibration Research pioneered such Field Data
Replication (FDR) testing several years ago and just
recently introduced their VR Observer™, a highly
CONCLUSIONS
Overall, random vibration testing is an excellent
portable four-channel recorder in support of this and
general-purpose tool for environmental vibration
other testing purposes. While FDR is the preferred
simulation. It is more efficient more precise, and
solution for many cases, it is not a replacement for
more realistic for this purpose than sine testing.
random vibration testing. FDR provides an exact
Although random vibration testing is not perfect,
simulation of one instance of the environment.
the testing industry should use the techniques
Random provides a statistical average of that
extensively in their screening and qualification
environment. Where FDR might exactly capture what
procedures.
one driver experiences while driving a prescribed
route, random represents the average of thousands
of different drivers trying to follow the same course.
While an FDR recording uses massive amounts of
memory, a random reference requires very little.
Certain mixed signal testing can also provide REFERENCES
1. Tustin, Wayne, Random Vibration & Shock Testing, Equipment
random signals with high kurtosis, but they are Reliability Institute, Santa Barbara, CA, 2005.
in no way competitive with Kurtosion. Sine-on- 2. Connon, W. H., “Comments on Kurtosis of Military Vehicle Vibration
Data,” Journal of the Institute of Environmental Sciences, September/
random tests may prove useful to simulate a specific October 1991, pp. 38-41.
environment with random and tonal components, 3. Van Baren, Philip D., System and Method for Simultaneously
Controlling Spectrum and Kurtosis of a Random Vibration,
such as an aircraft package-shelf experiencing both US Patent Office #7,426,426 B2.

Page 5 of 5 What Is Random Vibration Testing

+1.616.669.3028
[Link]
vrsales@[Link]

© 2020 Vibration Research Corporation is a registered trademark in the United States and other countries. | 062020