Sei sulla pagina 1di 4

6

Vibrations Vol 28 No 3 September 2011


T
e
c
h
n
i
c
a
l
A
r
t
i
c
l
e
frequency response curve supplied by the manufacturer is
needed to choose a suitable hammer and tip for the proposed
impact test.
Recommended Practices for a Bump Test
These practices for implementing a successful bump test result
from extensive research and discussions with practicing
vibration specialists.
When the hammer is struck, avoid middle or end positions
of the structure. They may be nodal points of the lower
fundamental natural frequencies.
Do not strike too hard; try to stay within the linear elastic
regime of the structure. Strike gently at first, then gradually
harder if necessary.
Check the impact force time waveform; a double hit will
appear as a double peak.
Check the impact force spectrum; it should be flat within
the frequency span of interest. A 20 dB drop from the initial
Although the bump test is one of the most fundamental methods
used to determine the natural frequencies of a structure, not
much literature is available regarding the detailed procedures
and techniques required for an accurate reading of natural
frequencies. A series of recommended practices for conducting a
successful bump test are summarized in this article. An example
illustrates their application for the accurate assessment of natural
frequencies.
If an exciting frequency matches a structural natural
frequency, resonance occurs. At resonance, vibration amplitude
can be controlled only by damping; however, artificially-
introduced damping is not preferred because it consumes
energy. In general, resonance should be avoided in any system.
In addition, structural natural frequencies should be separated
from potential exciting frequencies. A 15% separation margin is
commonly recommended [1] although a 20% margin is
advisable in some specific applications; e.g., the special purpose
gear unit in API Standard 613 [2]. An accurate reading of
structural natural frequencies of a system with a bump test is a
prerequisite to any code-compliant system design or design
modification.
Instrumentation for a Bump Test
The following instrumentation is required to carry out a bump
test.
FFT (Fast Fourier Transform) vibration analyzer with
cross-channel functions such as transfer function and
coherence features [3].
Instrumented hammer (a hammer with a force sensor
embedded inside) with multiple tips of different size and
hardness.
Vibration sensors; e.g. accelerometer.
Careful selection of a suitable hammer and tip is important for
a successful bump test. In general, a softer (often also larger)
hammer tip provides higher amplitudes of exciting frequency
components of the impact force, a narrower force spectrum, and
a lower exciting frequency. Figure 1 contains a series of sample
frequency response curves for an impact hammer with tips of
different size and hardness. For a specific hammer, a detailed
Recommended Practices for a
Bump Test
Jianke Wang, Ph.D., P.E.
National Oilwell Varco
Conroe, Texas
Jianke.wang@nov.com
Summary. Procedures for conducting an impact test
are reviewed. They include hammer selection,
impacting location and force, and analyzer setup. An
example provides details for interpreting test results,
especially with regard to phase.
Figure 1. Frequency Response Spectra for a Hammer
with Various Tips.
-20
-15
-10
-5
0
10 100 1000 10000
d
B

d
r
o
p

f
r
o
m

t
h
e

p
e
a
k

r
e
a
d
i
n
g
,

d
B
Frequency, Hz
Hard tip
Medium tip
Soft tip
Super soft tip
143096 Mag OP r5:109847 Vibration mag (11/03) 9/13/11 8:24 AM Page 6
7
Vibrations Vol 28 No 3 September 2011
low-frequency peak is generally sufficient. If the spectrum
is not flat, change to a harder hammer tip (or even to a
bigger hammer if necessary) to assure a flat force spectrum
within the frequency span of interest.
Check the response (vibration) time waveform. The
response should taper off to zero within the time span. If
necessary, adjust the FFT lines and frequency span to
extend the span of the time waveform.
Try a rectangular/no window first. If the response time
waveform is too noisy, use a force or exponential window
for the original signal collected.
Auto ranging is not recommended. Always check the
signal; no over-loading (clipping of the time waveform) of
the FFT analyzer should occur. If overloading exists,
increase the range manually.
Take several averages (three is often recommended) to
reduce spurious noise.
Mark and reference the sensor and hammer positions,
especially when multiple positions have been tried or a tri-
axial sensor has been used.
For a system under ideal conditions with one natural
frequency, a 180 phase shift occurs across the natural
frequency. However, single-frequency systems do not occur.
Even for the simplest single spring-mass system, many natural
frequencies of the spring itself exist. In addition, it is impossible
or not practical to eliminate all environmental noise during a
bump test. Therefore, in reality, not all peaks in the magnitude
plot of the force/response transfer function (inertance, mobility,
or dynamic compliance) are actual natural frequencies of the
system measured.
Several practices are recommended to confirm that the peak
frequency in the plot of the selected transfer function is an
excited system natural frequency.
A minimum phase shift of 30 to 45 in the phase plot of
the transfer function is recommended.
A minimum coherence of 0.8 between the input (impact
force) and output response (vibration) at that frequency is
recommended [4]. If the coherence is less than 0.8, the
peak could be a consequence of either background or
electrical noise if the test has been carried out correctly.
If necessary, the phase plot of the transfer function can be
flipped from the wrapped phase (within [-180, 180], the
default setup in most FFT analyzers) to the unwrapped one.
This may occur when two consecutive peak frequency
components are very close to each other.
According to Baxter [4] coherence in a linear system is a
measure of the causality between the input and output/response
signals. It represents the fractional portion of the output signal
power that results from the input signal at a specific frequency
component. For a linear and ideal system with a single input and
single output, the coherence between two signals will be equal
to one. For two signals that are completely unrelated, the
coherence will be zero. If the coherence is less than one but
greater than zero, either noise is entering the measurements or
the system has been struck too hard and attained a nonlinear
regime.
Three types of transfer functions between the input signal
(force) and output response (vibration) are commonly
recommended to identify the natural frequencies of a system.
They include inertance (also called accelerance, which is the
ratio of acceleration response over impact force input), mobility
(velocity response over impact force input), and dynamic
compliance (displacement response over impact force input).
The selection of the transfer function is based on the strength of
the measures (acceleration, velocity, or displacement) at
different frequencies and the value of the dominant exciting
frequency. Dynamic compliance is usually recommended for
low-frequency applications (less than 20 Hz). The transfer
function of inertance/accelerance is used for relatively high-
frequency applications (greater than 1,000 Hz). For intermediate
exciting frequencies, the transfer function of mobility is
preferred [3].
Example Results and Discussion
An example illustrates the application of the techniques to
accurately assess the natural frequencies of a system.
Figure 2 shows a setup for a shale shaker bump test. An
instrumented hammer, a tri-axial accelerometer, and an FFT
analyzer with cross-channel features are used to collect the data
of the impact force input and vibration response. The
accelerometer and hammer have been moved to obtain the best
signal. The operating frequency of the shale shaker is 30 Hz;
therefore, mobility (the ratio of velocity response to force input)
is utilized to identify the natural frequencies of the excited
system. Single integration of the acceleration measured is used
to obtain the velocity signal. Frequency resolution was set to
0.125 Hz with a frequency span of 200 Hz, 1,600 FFT lines, and
a uniform window (Editors note: Resolution is calculated
without the Vibration Institutes recommended factor of two.)
Figure 3 contains a set of results from the shaker setup in
Figure 2. In Figure 3, x represents vibration response in the
longitudinal direction; y is vertical direction, and z is lateral
direction. The impact force is represented by f; T is the time
waveform, and S represents the spectrum. H refers to the
transfer function between two signals. C is the coherence
between two signals.
In Figure 3 the impact force, a flat force spectrum, and the
classical (exponentially decaying) time waveforms of the impact
responses in the longitudinal, vertical, and lateral directions
verify that the impact test results are valid.
Figure 2. Shaker Bump Test Setup.
Accelerometer
Hammer
143096 Mag OP r5:109847 Vibration mag (11/03) 9/13/11 8:24 AM Page 7
10
Vibrations Vol 28 No 3 September 2011
Figure 3. Results of the Shaker Test Setup in Figure 2.
L
a
t
e
r
a
l
L
o
n
g
i
t
u
d
i
n
a
l
V
e
r
t
i
c
a
l
143096 Mag OP r5:109847 Vibration mag (11/03) 9/13/11 8:24 AM Page 10
11
Vibrations Vol 28 No 3 September 2011
In some cases, a wrapped phase plot within [-180, 180]
might cause confusion with the phase shift around a natural
frequency. In Figure 3, the phase plot of the transfer function in
the longitudinal direction has been flipped from the wrapped
phase within [-180, 180], which is the default setup in most
FFT analyzers, to the unwrapped one.
Figure 4 shows that an unwrapped phase plot is preferred in
the longitudinal direction. The top of Figure 4 is a wrapped
phase plot. Abrupt changes occur from -180 to 180 around the
peak amplitudes, especially at the peak frequencies of 40.88 Hz
and 48.75 Hz. They could cause misreading of the phase shift.
In this example, the unwrapped phase plot shown in the bottom
of Figure 4 is preferred.
On the other hand, although many abrupt changes occur in
the wrapped phase plot in the lateral direction (see Figure 3), an
unwrapped phase plot results in even poorer resolution of the
phase reading due to the huge phase angle span ranging from
-800 to 180.
In Figure 4, the three peak frequencies of 5.000 Hz (#7),
5.375 Hz (#8), and 6.125 Hz (#9) are not considered excited
natural frequencies of the shaker basket because the phase shift
around each peak is less than 30. The peak at 0.125 Hz is
exactly on the first non-zero bin of the FFT spectrum; near-zero
Hertz digital integration issues are present in most FFT
analyzers. In addition, the lower frequency limit of the
accelerometer response is off by 3 dB at 0.5 Hz; the 0.125 Hz
therefore cannot be confirmed as a natural frequency of the
shaker basket.
According to the recommended practices described in this
article, the first eight fundamental natural frequencies of the
shaker basket shown in Figure 2 are measured as 0.5 Hz,
1.25~1.50 Hz, 2.0~2.125 Hz, 2.875 Hz, 3.5~3.625 Hz,
4.125~4.250 Hz, 40.88 Hz, and 48.75 Hz.
The finite element method (FEM) in ANSYS was used in a
pre-stressed modal analysis of a simplified shaker basket model
(no fillet welds are included) to show that the first eight
fundamental natural frequencies are 0.30 Hz, 0.46 Hz, 1.31 Hz,
2.93 Hz, 4.05 Hz, 4.34 Hz,, 38.74 Hz, and 43.80 Hz [5]. The
test results match the predicted ones. If all fillet welds are
included in the shaker basket model, the slight differences
between the predicted and measured natural frequencies will be
even smaller [5]. The reason is that the fillet welds increase the
local stiffness of each weld [6]. In other words, with all fillet
welds included in the FEA model, the natural frequencies of the
rigid modes will decrease and those of the flexible modes will
increase [5].
Conclusion
This article presents a series of recommended practices that are
crucial to an accurate reading of the natural frequencies of any
structure or system. An example illustrates the application of
these recommendations so that the excited natural frequencies of
the system can be differentiated from environmental or electrical
noise.
Acknowledgment
The author wishes to thank NOV Brandt R&D Department for
their help during the bump and related testing.
References
1. Eshleman, R.L., Vibration Control, Advanced Vibration
Analysis, Vibration Institute (2006).
2. American Petroleum Institute, Special Purpose Gear Units for
Petroleum, Chemical, and Gas Industry Services, API Standard
613 (2003).
3. Eshleman, R.L., Basic Dual Channel Testing, Machinery
Vibration Analysis: Diagnostics, Condition Evaluation, and
Correction, Vibration Institute (2002).
4. Baxter, N.L., Field Application of Dual Channel and
Synchronous Time Averaging Analysis Techniques, Advanced
Vibration Analysis, Vibration Institute (2006).
5. Wang, J.K., Virtual Prototyping plus Physically-Tested
Validation Delivers State-of-the-Art Shale Shakers, Proceedings
of NAFEMS World Congress 2011, Boston, USA (2011).
6. Det Norske Veritas, Fatigue Design of Offshore Steel
Structures, DNV Recommended Practice DNV-RP-C203
(2010).
Figure 4. Reduction of Outboard Spindle Bearing Spike
Energy after Lubrication.
Wrapped Phase Plot
Unwrapped Phase Plot
.
143096 Mag OP r5:109847 Vibration mag (11/03) 9/13/11 8:24 AM Page 11

Potrebbero piacerti anche