6.1
are shown in Figure 6.2. Note that no absolute amplitudes are provided.
Excitation
A machine or structure
can be excited by one or more
vibratory forces.
The force
as
occurs
with
6.3
tools
required
to
for analyzing data include meters, oscilloscopes, tracking filters, data collectors, and FFT
analyzers.
Excitation Devices. Regardless of the excitation device selected, amplitude and frequency must
be controlled. If parameters are being evaluated, a steady or transient excitation is necessary; the
vibration response to the excitation must be measured and analyzed. Sweep excitation can be
induced artificially by mounting shakers on bearing caps or casings. Sweep excitation can be
produced at synchronous frequency by mass unbalance, a bowed rotor, and/or misalignment during startup or coast down; the machine is thus producing its own excitation. Figure 6.6 shows the na-
6.4
Excitations that occur at nonsynchronous frequencies can also be used for condition analysis and identification of parameters. Nonsynchronous frequencies are produced by oil whirl at
subsynchronous frequencies; by excessive misalignment at two and three times synchronous
speed; by transient torsional excitation, which varies from 120 Hz at startup on 60-cycle machines
to zero at operating speed, depending on the slip frequency; and by rolling element bearing excitations, which consist of non-integer multiples of shaft speed. The magnitudes of many of these
excitation sources are difficult to quantify because they arise from faults, but quantification is necessary for modal analysis.
The fact that impact excitations are produced naturally in reciprocating machines is advantageous for parameter identification but bad for normal operating conditions. Engines can be excited at frequencies up to 12 times the operating speed. Half orders can exist in four-stroke cycle
machines.
If naturally-occurring impact excitation is not available in a machine, a hammer or shaker
can be selected to excite natural frequencies. However, it is necessary to exercise caution when
these tools are used to excite operating machines to avoid bearing damage. Figure 6.7 shows a
modally tuned hammer; the sensor in the head of the hammer measures force. Phase information, which is useful in identifying natural frequencies, can be obtained when a hammer is used
in conjunction with a dual-channel analyzer.
Hammers are also used to conduct impact, or bump, tests on structures and piping.
Hammer tips (see Figure 6.7) with various levels of resiliency are used to obtain the waveforms required to provide the necessary data.
If a hammer is not available, a four by four timber about five feet long is adequate for bump
tests. Sledge hammers with steel heads, unless covered by an elastomer, are not ade-
These hammers
bounce off the structure and do not put enough energy into the system. Shakers and calibrated
hammers provide the controlled excitation needed for modal testing.
6.5
Sensors. Sensors are selected to obtain vibration signals that provide information about the dynamic characteristics of a machine or structure. Permanently mounted proximity probes are used
to obtain information about the critical speeds of a rotor. Seismic sensors, velocity pickups, and
accelerometers are placed on the casing, foundation, piping, and other structures to determine
natural frequencies and mode shapes. Moving a sensor to different locations while applying a
fixed force with a hammer or shaker provides information about mode shapes.
Analyzers. The instrumentation used to measure and analyze a resonance or critical speed depends on the goals of the test and the instrument available. Meters and oscilloscopes are adequate for simple structural testing to obtain natural frequencies.
A tracking filter is best for rapid run-up/coast-down tests. Vibration is indicated in the filtered frequency band, which is governed by a reference signal generated by a proximity
probe/notch or optical pickup/reflective tape. Peak vibration levels and phase changes indicate
critical speeds.
The single-channel FFT analyzer or data collector can be used for impact tests in either
the time or frequency domains. Triggering can be free or from a hammer source. Vibration peaks
indicate resonances. During impact tests a uniform (none) window or a Hanning window with a
pretrigger should be used on the analyzer. Some analyzers have special windows for impact
tests.
The dual-channel analyzer is used for impact tests when force and vibration measurements are available. The force (source) is usually recorded in channel one; vibration (response)
is recorded in the second channel. The resulting transfer function of vibration/force is classed as
a mobility (velocity/force). Phase and coherence are generated simultaneously. Changes in
phase provide confirmation that a natural frequency is present at a vibration peak. Coherence is a
measure of the amount of measured force coming from the impact hammer and the relationship
between the source (excitation) and the response (vibration).
Most analyzers have special windows that force the signal to zero at the end of the FFT
sampling process (see Chapter 3) one for the source and one for the response.
The sampling time required for the FFT process is relatively long. When FFT spectrum
analyzers are used for critical speed tests, rapid coast downs or startups will result in loss of data
the analyzer may not be able to take samples fast enough to obtain sufficient data points for a
continuous vibration versus frequency curve. If the transient time is known, the number of data
points can be calculated.
6.6
Determine the important sources of vibratory forces; e.g., mass unbalance, misalignment,
gearmesh, engine gas, and inertia forces. Important engine forces can be determined from
a star diagram6.2.
2.
Plot the frequency of the vibratory force in CPM or Hz against speed in RPM. For example,
a mass unbalance has a forcing frequency equal to the speed of the rotor, sometimes called
a once-per-revolution (1x) vibration.
3.
Measure or calculate the natural frequencies at a number of different speeds. The machine
can be struck prior to operation to determine the natural frequency when shut down. If the
natural frequency does not vary with speed, the plot is a straight line (see Figure 6.8). If the
natural frequency varies with speed, as is the case with fluid-film bearing, impact tests must
be conducted at various speeds. However, the natural frequency at a critical speed can be
obtained with a coast-down test. If this test is not possible, the natural frequencies are calculated from a validated model. A mass-elastic model6.4 is constructed from drawings or
sketches of the hardware and validated using test results either at the critical speed or zero
speed. The natural frequencies are calculated at a number of speeds so that the curve of
natural frequency versus rotor speed can be plotted.
6.7
The
Determine the vibrations of the structure at a number of known points during operation; see
Figure 6.9. These points should allow an approximate modal description of the vibrations in
the structure. Reference the vibrations to the forces by using a strobe light or key phasor if
possible. If the machine speed is variable, do a waterfall, or cascade, diagram of vibration
versus speed (Figure 6.10) to obtain data for the interference diagram.
2.
Construct an interference diagram from the best available data. The waterfall diagram will
show resonant points.
3.
Strike the structure with a 4 x 4 timber, mallet, or hammer with a soft head in the direction of
the desired mode. If the desired mode is not known, strike the structure in several directions.
The direction shown diagrammatically in Figure 6.9 vertical and horizontal usually provide useful data.
4.
Striking rate is determined by the data acquisition time of the FFT setup. Strike only once per
data acquisition period.
5.
Striking force is determined by the response (damping) in the structure. Do not overdrive the
source strike lightly and increase if required. This means a bit of trial and observation of
the response.
6.
Measure and record the vibration levels at a number of reference points on the structure (see
Figure 6.9). The peaks on the spectrum of vibration levels at various measurement points indicate the natural frequencies of the structure (Figure 6.11). Some natural frequencies are
not seen at all measurement points. These are nodal points.
6.8
Sinusoidal vibrating forces can excite natural frequencies. However, these forces must be swept through
Sensors
forcing frequency that will in turn excite natural frequencies (see Figure 6.8). This phenomenon often occurs in
normal machine operation.
IPS
Y: .452 IPSrms
Figure 6.10. Waterfall or Cascade Diagram
6.9
6.10
system will excite natural frequencies. Figure 6.14 shows a gearbox in which a natural frequency
is excited by all of the random vibration in the system.
Machine motion
causes an impact-like force that excites natural frequencies that might or might not be desirable.
The most popular method for exciting natural frequencies is to strike (excite) a machine or
structure with a timber or a hammer. The range of
frequencies excited is dependent upon the duration
of the impact. Hard-faced steel hammers tend to
Figure 6.14. Gearbox Vibration
6.11
6.12
Storage analog or digital oscilloscopes can be used to observe instantaneously the vibration resulting from an impact to a structure. A trigger level must be set on the oscilloscope in the
single sweep mode. The vibration signal from the impact will be held on the screen of the oscilloscope (Figure 6.16). The natural period is obtained by counting the divisions in one period of vibration and multiplying that number by the time-per division setting on the time base (Chapter 3).
In Figure 6.16 it can be seen that two divisions are contained in the period, or vibration
cycle. The oscilloscope was set at 5 milliseconds per division. The natural period, therefore, is
= 2 div. x 5 m sec./div. = 10 m sec. = 0.01 sec.
The natural frequency is one divided by the natural period.
f n = 1/ = 1/0.01 sec./cycle = 100 cycles/sec. (Hz) = 6,000 cycles/min.
The procedure with a digital oscilloscope is similar except that the period is read directly with a
cursor.
6.13
Set Fmax so that natural frequency of interest will be included in the spectrum.
2.
3.
Take data collector off autoranging and approximate range required to not over or under
load the analyzer.
4.
5.
Set the trigger on the vibration use a 1/3 sample pretrigger, if the Hanning window is
used.
6.
Select one to four (4) averages depending on the noise on the data.
7.
8.
Set autoscaling off select scales for displays during impacting or processing of recorded
data.
9.
10.
Readjust setup and/or striking force as required after initial data are analyzed.
Set the FFT frequency span, Fmax, wide enough to capture the natural frequency of inter-
est and any other natural frequencies in the vicinity generally FMAX = 2 x FN. Select the number of lines that will provide adequate resolution. If the test is complete and resolution is not adequate to separate natural frequencies, it will have to be repeated unless it has been tape recorded.
Set the range of the analyzer so it will not overload or have insufficient dynamic range. Do
not use autoranging. Just because the response is low in amplitude does not mean that it is not
important.
Select a uniform or Hanning window. If the vibration does not ring down to zero at the end
of the data sampling time, leakage will occur and a uniform window is not appropriate. If a Hanning window is used select a pretrigger of 1/3 to 1/2 of the data sampling time (Figure 6.11). Otherwise the Hanning window will destroy the data. In Figure 6.17, an impact test on a vertical
pump was conducted with a uniform window. Note the time waveform is intact.
Set the trigger on the vibration signal. Use a pretrigger with a Hanning window otherwise data will be lost (Figure 6.18). The reduction of amplitude by a Hanning window if the data
6.14
are not suitably positioned in the buffer are shown in Figure 6.19. The upper trace is not windowed. The lower trace is windowed this is what the FFT uses to calculate the spectrum.
6.15
6.16
Figure 6.19. Impact Test Results with a Hanning Window. Upper Trace not Windowed
The interval of impacting should be greater than the data sampling time so that one impact
is recorded per FFT sample. Otherwise sideband noise will occur (Figure 6.20). The choice of
using single or multiple-averaged impacts depends on the noise within the data. The choice of
averaging has to be made on a case by case basis. Usually four (4) averages provide a good
compromise. The level of impacting should be just enough to get response. Do not try to kill it!
Figure 6.21 shows data from a dual channel impact test on a fan support structure. The
trigger was on the instrumented hammer. An exponential response window (Chapter 3) was used
on the velocity signal and a force window was used on the hammer signal. The upper plot is the
transfer function mobility (IPS/lb) and the lower plot is phase. Approximate 180 phase changes
in the vicinity of transfer function peaks are used to identify natural frequencies. The real portion
(transfer function multiplied by the cosine of the phase) is used to develop mode shapes. The imaginary part of the transfer function (transfer function multiplied by the sine of the phase) is used
to evaluate damping.
6.17
Y: 33.302
6.18
Select one or more appropriate sensors to measure the vibration. Proximity probes measure relative shaft displacement and are preferred if they are permanently installed. Otherwise, mount seismic sensors either velocity sensors or integrated accelerometers as
close to the bearing as possible in the horizontal and vertical directions. If casing resonance
is suspected use casing mounted sensors.
2.
For a permanent or temporarily mounted trigger use a proximity probe or a magnetic pickup
adjacent to an indentation or mark on the rotor. A combination of an optical pickup and reflective tape or paint (hot surfaces) may also be used.
3.
Wire the vibration pickups and trigger to a tracking filter, tape recorder, or FFT analyzer.
4.
If the vibration pickups and trigger are permanently mounted, a coast-down test of the machine can be performed directly.
5.
Run the machine at 10% to 15% overspeed if possible, then cut the power and allow the
machine to coast down from normal operation as data are being recorded. Otherwise
measure and record data prior to and during the time the power is cut.
6.
If either sensors, trigger, or both must be mounted, record the startup. Run the machine until it is thermally stable before cutting power for the coast-down data.
7.
Process the data and identify the critical speeds. They will be peaks in the spectrum. Depending on the plot they will be peaks in the spectrum from an FFT analyzer, peaks from the
tracking filter in a Bod plot, or loops from the tracking filter in a polar plot.
8.
It is a good idea to record coast down and startup data prior to and after an outage.
speed. The natural frequencies of rotors mounted on rolling element bearings are constant and
not speed dependent because the stiffness of the bearings is relatively constant it varies slightly with load.
Figure 6.22b shows the variation in fluid-film bearing stiffness with rotor speed (vertical
lines) superimposed on the variation in natural frequency of the rotor with bearing stiffness (horizontal lines). Rotors with large overhung impellers have gyroscopic effects that stiffen the rotor
as speed increases, thereby raising the natural frequency6.5.
6.20
Rotors are often lifted out of a machine and rung to obtain natural frequencies. In such
cases only the natural frequencies of the rotor are obtained, not those of the system. The effects
on natural frequencies and critical speeds of fluid-film bearings, pedestals, and supporting foundations would be accounted for as they are in Figure 6.22.
The natural torsional excitation of
motors and engines is often used in resonance and critical speed testing. The interference diagram would show no variation of
natural frequency with speed (see Figure
6.23) for an example of an interference diagram with no variation in natural frequency).
In synchronous motors an excitation at
twice slip frequency is obtained during
startup that excites the first and second torsional natural frequencies of the unit. The
result is a sweep of a single frequency. Because the excitation varies from 120 Hz (60
Figure 6.23. Synchronous Motor Start-Up
ure 6.23). Engines are usually tested by sweeping through the speeds from low to high idle, so
that the many engine frequencies will excite the natural frequencies of the system. These frequencies are
then identified using an interference diagram (Figure
6.24).
An
interference
diagram
for
an
en-
gine/generator is shown in Figure 6.24. The first torsional natural frequency can be excited at engine
speeds of 1,050 RPM, 1,400 RPM, and 1,800 RPM at
6, 4.5, and 3.5 orders of excitation respectively. The
torsional test data (Figure 6.25) show that the 4.5x order torsional vibration excites the first natural
Figure 6.24. Calculated Torsional
Critical Speeds of an Engine/Generator Unit
6.21
These
machine.
modal properties of the machine (Chapter 8). For example, the noncontact proximity probe is
used to assess vibration modes of the rotor; seismic sensors on the casing, pedestals, or foundation are used to determine structural modes.
6.22
data on critical speeds. However, the frequency range selected must be high enough to track the coast down. The peak hold feature
holds and displays the peak values of all data after each spectrum is computed (figure 6.28a).
The acquisition time of the block of data analyzed is dependent on the frequency span selected.
The lower the frequency span, the longer the acquisition time.
N
f
N is the number of lines, usually 400. The number of samples is a measure of analyzer resolution. The acquisition time can be decreased by reducing the lines of resolution. In addition, use of
overlap processing also reduces data acquisition time. With overlap processing data from the
previous sample are used in calculating the spectrum. In typical machine coast-down tests the
process of computing the FFT is faster than the data acquisition.
6.23
If a 400-line FFT analyzer is set on a frequency span of 100 Hz (6,000 CPM), four seconds (400/100) are needed to acquire a sample. Because many samples are necessary to plot a
smooth curve during a start-up or coast-down test without data drop out, the frequency span of
the analyzer must be evaluated carefully prior to data collection. Resolution is lost with too wide a
frequency span. Too narrow a span may prevent the observation of the critical speed because
the data acquisition time is excessive. Table 6.5 illustrates the procedure
tained hardly sufficient for a curve. Increasing the frequency span to 400 Hz
(24,000 CPM) and decreasing the number
of lines to 100 means that a sample is
taken every 0.25 second (100/400 =
0.25);
12
second
span/0.25
sec-
36 Data
Points
6.24
sine of phase angle) and the imaginary (amplitude times sine of the phase angle) at the various
speeds.
Figure 6.29 shows an example polar diagram. The generator is operating in the coun-
ter clockwise direction therefore phase lags are measured in the clockwise direction. The small
6.25
loop6.6 in Figure 6.29 identifies the first critical speed of the generator as 1,000 RPM. There is a
phase change of 90. The large loop is the second critical speed, 2,250 RPM. It shows the operating speed to be well above the second critical.
DAMPING
Damping information can be used to identify problems that might occur when a machine passes through or dwells on a critical speed. If significant damping is present, it is not necessarily
harmful to operate a machine close to a critical speed, but inadequate damping could result in a
disaster.
Evaluating machines subject to mass unbalance during operation as well as those operating close to a critical speed requires data on damping. Damping for a single mode can be estimated from the time domain, the frequency domain, or by phase. Modal testing techniques are
used when a system has more than one mode (Chapter 8).
Damping Ratio and Logarithmic Decrement
The time domain of the vibration of a steam turbine
casing subjected to an impact is shown in Figure 6.30. The
damped natural period, which is obtained directly from the
plot, is equal to 0.0125 second per cycle. The frequency,
which is the reciprocal of the period, is equal to 80.0 Hz.
The damping ratio can also be calculated form Figure
6.30. The damping ratio (C/CC) is a measure of modal viscous damping. It is defined as the ratio of damping in the system to critical damping. Critical damping is the lowest magnitude of damping that will not allow vibration. The damping
ratio varies from zero (no damping) to one (no vibration).
Figure 6.30. Time-Domain Data from an Impact Test on
a Turbine Casing
Typical damping ratios are 0.001 for steel; 0.05 for rubber;
0.025 for machines with rolling element bearings; and from
0.03 to 1.0 for machines with fluid film bearings.
The API
(American Petroleum Industry) recommendation is that machines passing through a critical speed should have a damping ratio greater than 0.0625.
6.26
The level of vibration of a damped system typically decreases with time at a rate that can
be used as a measure of damping. This measure is called the logarithmic decrement. It is defined in equation form as
xo
1
n
n
xn
=
=
logarithmic decrement
number of cycles of vibration measured
xo
xn
n
=
=
The logarithmic decrement is related to the damping ratio by the following equation.
2 C / Cc
2
1 C / Cc
C
C
for
0.5
Cc
Cc
C C 2
C
Cc
logarithmic decrement
damping ratio
CC
fn
natural frequency, Hz
The logarithmic decrement for the dominant pedestal mode at 83 Hz in Figure 6.30 is (1/7)
6.27
Q AF
fn
f2 f1
or
1
C
2
C
c
C
1
C c 2 AF
quality factor
AF
amplification factor
fn
f1
f2
The half-power method can be used to calculate a damping ratio of 0.16 from the turbine data
shown in Figure 6.31. The critical speed is 5,800 RPM, the first half-power point is 5,200 RPM,
the second half-power point is 7,100 RPM, and the amplification factor is 3.05.
Coast-down data can be used to calculate the damping of a rotor (Figure 6.32). The rotor
is part of a grinding machine that is mounted on a structural platform. The peak vibration at 0.245
inch per second can be used to identify the critical speed, which is 1,238 RPM.
The half-power points are used to obtain the amplification factor. At 0.707 times peak vibration, or 0.17 inch per second, the speed N1 and N2 at the half-power points are 1,350 RPM
and 1,140 RPM (see Figure 6.32). The damping ratio for a calculated amplification factor of 5.9 is
0.084. This relatively high value is surprising because the rotor is supported in rolling element
bearings.
The actual damping value can be calculated from single-degree-of-freedom theory. The
rotor weighs 525 pounds. The critical damping is calculated as follows.
Cc 2 m n
m
natural frequency
= 20.6 Hz x 6.28 = 129.57 radians/sec.
CC
The structural natural frequency (28 Hz) of the platform could not be excited by a bump test but
was seen during a coast-down test.
6.29
N1
AF
AF
amplification factor
NC
d/dN =
Nc d
360 d N
rate of change of phase angle with respect to speed around critical speed
The rate of change d/dN is approximated by the finite difference /dN. From Figure 6.31, N is
equal to 5,800 RPM and the finite difference is (458-380)/(6250-5250), which is 0.078. The amplification factor is
AF = (5,800/360)(0.078) = 3.95
6.30
REFERENCES
6.1.
Eshleman, R.L., Resonance and Critical Speed Testing. Part I: Basic Concepts and Instrumentation, Vibrations, 6(3), pp 3-7 (1990).
6.2.
Den Hartog, J.B., Mechanical Vibrations, McGraw-Hill Book Co. (1968) [Dover Books].
6.3.
Eshleman, R.L., Resonance and Critical Speed Testing. Part II: Resonance Testing Techniques, Vibrations, 6(4), pp 3-7 (1990)
6.4.
Eshleman, R.L., Modeling Turbomachinery for Design and Analysis, Rotor Dynamics and
Balancing Course, Syria, VA, May 15-19, 1989, Vibration Institute, Willowbrook, IL (May
1989).
6.5.
Eshleman, R.L. and Eubanks, R.A., On the Critical Speeds of a Continuous Shaft-Disk
System, J. Engrg. Indus., Trans. ASME, 89 (4) (November 1967).
6.6.
Halloran, J.D., Vibration Analysis of Rotating Machinery Using Polar Diagrams, Vibrations,
1(4), pp 2-10 (1986).
6.31