Você está na página 1de 4

T

e
c
h
n
i
c
a
l
A
r
t
i
c
l
e
Modal Testing.
Part I: Introduction and Impact Testing
Excitation of natural frequencies (resonance) in machinery
and structural supporting systems is a common cause of
excessive vibration. Modal testing techniques such as
impact and coast-down testing can be used to identify
natural frequencies. Experimental modal analysis (EMA)
can be used to determine the mode shapes associated with
each natural frequency. This information is crucial to
successful modification of machinery and structures in order
to change their natural frequencies. In some cases, operating
deflection shape (ODS) testing can be used to develop the
mode shapes of excited natural frequencies.
Resonance, which is the result of excitation of a natural
frequency, is a common root cause of excessive machine
and structural vibrations. Operating machinery that produces
dynamic forces at frequencies close to natural frequencies
can significantly amplify vibration.
Basic Theory
Figure 1 is a simple approximation of a structural-
mechanical vibrating system. It consists of the mass (m) of a
machine supported by a structure approximated by two
springs, each having stiffness (k). The total stiffness of the
structure would be 2k. This approximation is presented as an
academic exercise in order to illustrate the difference in the
response of machine structures to static loading versus
dynamic loading. Most structural-mechanical systems are
too complex to be treated as such a simple vibrating system.
If the force (F
static
) produced by the machine is statically
applied to the structure, the resulting deflection is:
Deflection = Force/Stiffness = F
static
/2k Equation (1)
The response of the structure to a dynamic load is more
complex. It depends upon the ratio of the frequency of the
dynamic force (F
dynamic
) to the natural frequency (f
n
) of the
structure. The vibration amplitude that will occur as the
result of dynamic loading is:
Amplitude = MF (F
dynamic
/2k) Equation (2)
This is similar to the calculation for static deflection except
that it contains a magnification factor (MF) that accounts for
the proximity of the frequency (f
d
) at which the dynamic
force is applied to the natural frequency of the structure. For
a simple single-degree-of-freedom system with damping
(), the magnification factor is:
MF = 1/ [{1-(f
d
/f
n
)
2
}
2
+ {2(f
d
/f
n
)}
2
]
1/2
Equation (3)
Most structural-mechanical systems are lightly damped.
Neglecting damping further simplifies the magnification
factor:
MF = 1/ [1-(f
d
/f
n
)
2
] Equation (4)
Figure 2 illustrates the effects of resonant amplification.
The amplification increases as the frequency ratio increases
from the origin toward 1.0. At a frequency ratio of 1.0, the
magnification factor without damping is infinite. The
magnification factor then begins to decrease as the
frequency ratio increases beyond 1.0. At ratios above 1.414,
vibration is attenuated instead of amplified.
Robert J. Sayer, PE
Applied Structural Dynamics
Medina, Ohio
Summary. Part I of this three-part article on modal
testing introduces basic theory, test equipment
requirements, and proper procedures for modal
testing. Parts II and III include case histories to
illustrate the use of modal testing in root cause
analysis and considerations for shop testing versus
field testing.
Figure 1. Spring-Mass System.
5
Vibrations Vol 26 No 2 June 2009
132237 Mag r2:109847 Vibration mag (11/03) 6/9/09 4:40 PM Page 5
Very stiff structures will have higher natural frequencies
and lower frequency ratios (usually significantly below 1.0).
Thus, stiff structures will be subject to some resonant
amplification, albeit in most cases the magnitude of the
amplification is minimal. In addition, because the vibration
levels of a very stiff structure that is not subjected to
resonant amplification are usually very low (unless the
dynamic forces are very large), the resultant amplified
vibration is typically acceptable (unless the frequency of the
dynamic force is almost coincident with the natural
frequency).
The design of structural support members and foundations
should include a calculation of natural frequencies and mode
shapes. Natural frequencies in the neighborhood of
frequencies of dynamic forces produced by a machine
should be altered by structural modification to preclude the
excitation of resonance.
The evaluation of structural natural frequencies is even
more important for equipment operating with variable
frequency control devices. Dynamic forces will be produced
by equipment over a wide range of frequencies instead of at
discrete frequencies. This increases the probability of
resonance unless the structures are properly evaluated.
Impact Testing
Impact testing is commonly used to determine the natural
frequencies of machines and structures. An impact test can
be performed using a single-channel analyzer; however, this
paper is limited to a discussion of multi-channel impact
testing. A multi-channel impact test requires an instrumented
force hammer, at least one response transducer, and a multi-
channel analyzer with cross channel functionality.
Impact hammer and tip selection. Instrumented force
hammers are available from several manufacturers. The
most important criteria in the selection of an impact hammer
are the capacity and frequency range of the hammer. The
hammer should be capable of providing an impact force that
will produce a ring-down response significantly greater
than the background vibration present during the test. The
impact hammer must be big enough for the intended test and
provide force across the entire frequency range of concern.
It is recommended that the time waveform of the response
transducer and the frequency spectrum of the applied force
be checked to insure that these criteria are met.
It is essential that the impact hammer excite the mechanical
system over the entire frequency range of concern. This is
controlled by the hardness of the hammer tip.
The maximum frequency to which an impact hammer can
excite is a function of the time duration of the impact.
Figure 4 shows the calculated Fourier transform for the
impact force shown in Figure 3. The frequency range of
excitation is related to the inverse of the time during which
the impact force is applied. The frequency range of
excitation increases as the impact is applied over a shorter
duration period. The harder the tip, the faster the hammer
will rebound from the machine or structure that is excited.
The time duration of the impact decreases as the hardness of
the hammer tip increases. Thus, the frequency range will
increase as the hardness of the hammer tip increases.
Note that the amount of force (~12 lbs) available at any
particular frequency is much less than the total impact force
(~2,000 lbs) provided by the hammer. The total force is
spread out across the frequency range of the impact. This
reduction in available excitation force is an important
consideration when testing large massive machines and
structures. If the force is not sufficient to produce a ring-
down response significantly greater than the background
vibration present during the test, an alternate test method
may have to be considered.
Response transducer selection. A response transducer
must be selected that does not significantly affect the natural
frequencies and mode shapes of the structural-mechanical
system being tested. Since natural frequency is affected by
6
Vibrations Vol 26 No 2 June 2009
Figure 2. Magnification Curve.
Figure 3. Waveform Impact Force.
Figure 4. Frequency Spectrum.
132237 Mag r2:109847 Vibration mag (11/03) 6/9/09 4:40 PM Page 6
mass, the weight of the transducer should be as light as
possible so as not to contribute to the modal mass of the
structure or machine being tested. Larger transducers are
more appropriate for very massive machines and structures.
For very small machine components, a non-contacting probe
may have to be employed so as not to adversely affect the
results of the test. The author has used microphones to
obtain natural frequencies by measuring the acoustic
pressure emitted from a vibrating ring-down response to an
impact force.
The frequency range of the response transducer should
match that of the test. Commercially available transducers
are supplied with a calibration certificate that defines their
frequency range. It is important that the transducer is
capable of accurately measuring the ring-down response of
a machine or structure to the lowest frequency of concern
and to the highest frequency limit set for the impact test.
Small high-frequency transducers are available to measure
the response in the high-frequency range. Higher frequency
transducers generally have limited lower frequency range.
Seismic accelerometers are available to measure low
frequencies but are relatively massive. Capacitive
accelerometers are lightweight and have the ability to
measure acceleration down to 0.0 Hz. However, capacitive
accelerometers have a limited upper frequency range.
Another consideration in the selection of a response
transducer is that displacement probes accentuate response
at lower frequency and accelerometers accentuate response
at higher frequency. Figure 5 is a frequency spectrum of
impact test data reported in units of displacement (mils).
Figure 6 is a spectrum of the same data reported in units of
acceleration (gs). The upper spectrum clearly accentuates
the peaks in the lower frequency range; the lower spectrum
is dominated by the higher frequency peaks.
Frequency range and resolution. The setup within any
spectral analyzer requires setting the frequency range (f
max
)
for data acquisition. Since some maximum frequency is
defined, the data represent a truncated sample of the
response of a mechanical system to the impact force.
Truncation of data can affect the accuracy of any modal
analysis if the frequency range of data acquisition is not
properly defined.
Two rules-of-thumb are suggested. First, set the frequency
range of the analyzer so that all natural frequencies of
concern are identified by the impact test, including natural
frequencies that currently may not be excited but could be
influenced by any subsequent structural modification.
Second, define an f
max
that is significantly greater than the
largest natural frequency of concern.
Resolution becomes important when several modes of
natural frequency are close to each other. For these cases,
the resolution (number of lines or samples of data) must be
increased to properly separate the mode shapes associated
with each natural frequency. The resolution of the data is
defined as:
Resolution = f
max
/Lines of Data Equation (5)
Data acquisition time is a function of the reciprocal of the
resolution according to the following relationship:
Acquisition Time per Average = 1/Resolution Equation (6)
For example, the following data acquisition times will result
from the listed resolutions.
Resolution = 1 Hz Time = 1/1 = 1 sec
Resolution = 0.5 Hz Time = 1/0.5 = 2 sec
Resolution = 0.25 Hz Time = 1/0.25 = 4 sec
It may become necessary to increase the resolution of the
data and thus, data acquisition time for very flexible,
lightly-damped mechanical components. Flexible mechanical
systems continue to ring-down for a longer time after the
application of an impact force. Even though closely coupled
natural frequencies may not exist, the resolution may have to
be increased when testing flexible systems to assure that the
response to the impact force is close to termination by the
end of data acquisition. Figure 7 is a ring-down waveform
that is truncated far before the response has had a chance to
decay. This impact test required increased resolution, not to
distinguish closely coupled modes, but to capture more of the
ring-down event.
Figure 6. Acceleration Spectrum.
Figure 5. Displacement Spectrum.
Figure 7. Truncated Ring Down Response.
7
Vibrations Vol 26 No 2 June 2009
132237 Mag r2:109847 Vibration mag (11/03) 6/9/09 4:40 PM Page 7
8
Vibrations Vol 26 No 2 June 2009
Force-exponential window. A force-exponential data
conditioning window is recommended when an impact test
is performed. The setup of this window in most
commercially available FFT analyzers requires the
following inputs:
Trigger Delay Time
Start Time of Force Window
Width of Force Window
Decay Constant of Exponential Window
Figure 8 contains a force window superimposed upon the
time waveform of an impact and an exponential window
superimposed upon the resulting response waveform. The
trigger delay is set so that the force falls inside the window.
As an example, a trigger delay equal to 10% of the data
acquisition time and a window delay of 9% of the data
acquisition time assure that the force occurs after the start of
the application of the window. The width of the force
window must be sufficient to assure that the window is not
terminated before the force is applied.
The definition of the exponential decay constant differs
among commercially available analyzers. Reference
should be made to the manual provided with the analyzer
for instruction on the proper setup of the exponential
decay constant.
Test results (transfer function and coherence). Transfer
function and coherence are multi-channel functions. It is
not possible to obtain these functions with a single-channel
analyzer. The transfer function is a measure of the
response of the mechanical system per pound of dynamic
force over the frequency range of concern. It is obtained
by dividing the average cross power spectrum of force and
response by the average spectrum of the force hammer.
Therefore, the transfer function is valid only over the
frequency range over which the force hammer provides
excitation to the test subject.
Because the transfer function is a measure of the response
per pound of force, it provides an accurate assessment of the
structural sensitivity of a mechanical system versus
frequency. By definition, the response of a structure to a
given dynamic load is greater at resonance. Therefore, a
peak response on the transfer function curve indicates the
probable existence of a natural frequency.
The coherence function is a way to evaluate the
probability that any peak response in the transfer function is
associated with a natural frequency and is not due to
background vibration or other sources of interference. It
provides one indication of the validity of the impact test.
Coherence is the ratio of the coherent output power to the
total output power. In other words, it provides a comparison
between the response obtained for a single sample of impact
data to the average response of all the impact data. If the
mechanical system is linearly elastic, the structural response
should be identical for each impact, resulting in a coherence
of unity (1.0). A coherence of 1.0 indicates a coherent
(good) impact response. A coherence of 0.0 indicates an
incoherent (bad) response.
The coherence is always 1.0 for a single impact because
the sample response and the average response are identical if
only one sample exists. Therefore, multiple samples should
be obtained. A minimum of four averages is recommended.
Some reasons for poor coherence are double impacting,
electrical problems with cables, power supply problems,
highly nonlinear systems, high levels of background noise
(low signal/noise ratio), and faulty instrumented hammer or
response transducer. Figure 9 is an example of poor
coherence. Figure 10 is an example of good coherence.
Both were obtained from impact tests performed in an
industrial setting.
Figure 8. Force-Exponential Window.
Figure 9. Example of Poor Coherence.
Figure 10. Example of Good Coherence.
132237 Mag r2:109847 Vibration mag (11/03) 6/9/09 4:40 PM Page 8

Você também pode gostar