Mounting of Accelerometers

Emerson Process Management - CSI
DoctorKnow® Application Paper

Title: Real-World Mounting of Accelerometers
Source/ S.V.Bowers, K.R.Piety, R.W.Piety
Author:
Product: General
Technology: Vibration
Classification:
Real-World Mounting of Accelerometers for Machinery Monitoring
S. V. Bowers, K. R. Piety and R. W. Piety,

Computational Systems Incorporated, Knoxville, Tennessee
Typical accelerometer mounting configurations were evaluated to determine acceptable

frequency response and resonant frequencies for each mounting. Basic configurations
studied included stud, adhesive, quick lock and magnetic. A hand held probe with varying
stinger lengths was also evaluated. In addition, the effects of bad mounting techniques were
considered. Frequency ranges and resonances were determined for sensor mounts exposed to
rocking, looseness and varying hand pressure.
Predictive maintenance programs that operate in utilities and major industries rely heavily on
manual data collection. The temporary sensor mounting techniques utilized in manual data
acquisition significantly impact data repeatability and reliability and thus the decisions based on
these data. The benefits of these monitoring programs are dependent upon accurate data collection
techniques which affect both detection and diagnosis. Provided vibration instrumentation is
calibrated, accurate vibration measurements are ensured when: (1) sensor mountings do not limit
frequency and dynamic ranges; (2) sensor mass does not change the vibration characteristics of the
test object; and (3) measurement locations are exactly repeated1
The objective of this project was to evaluate the limitations of some common temporary sensor
mounting techniques as they apply to predictive maintenance programs. This was achieved in three
major sets of experiments. For the first set, maximum frequency responses were determined for
common sensor mount configurations. In the second set of tests, misapplication of sensor mounts
was investigated. The third set of experiments explored the effects of several sources of variability
on vibration measurements.
Frequency response limits for a given sensor are defined as the linear range of the frequency
response curve for that sensor. It is desirable for the output from the sensor to be linearly related to
the input over as wide a frequency range and dynamic range as possible.1 A main focus of this
study was to determine sensor frequency response limits resulting from various mount
configurations. In practical applications, the actual frequency response of a sensor depends upon
the dynamic response of the sensor/mount configuration. Vibrations at frequencies near the sensor/
mount configuration resonant frequency will be amplified, whereas at other frequencies the
measured vibration is greatly attenuated. The frequency ranges and resonant frequencies of various
http://www.compsys.com/DRKNOW/APLPAPR.NSF/apw...E03E63DB09436840852565A2006504E6?OpenDocument (1 de 22)05/07/2010 19:13:21

mount fixtures are reported in this article.
The usable frequency range of a mounted sensor can be reduced by rocking and loose connections.
In addition to these factors, the frequency range for hand held probes can be reduced due to
variations in hand pressure. Limitations imposed by these misapplications were evaluated. Because
of improper mounting and irregular or dirty surfaces, rocking occurs, which tends to decrease the
usable frequency range of the transducer.2 Looseness occurs when mount configurations and
sensor are not tightly connected. This problem results in large errors at high frequencies,3 as well
as chattering. Hand held probes can also experience reductions in frequency response due to
chattering because of improper hand pressure? Any action that allows the pickup axis to change
direction also produces incorrect readings.
Other sources contributing to variability in measured results are the mount positions on vibrating
surfaces, varying sensor placement techniques by different people, and individual consistency.
This article presents results relating to these factors and attempts to quantify their effects on
measured values in typical or extreme cases. Tests were evaluated on industrial machines from
data obtained with an accelerometer and a hand held probe using common mount configurations
currently employed in industrial predictive maintenance programs.
Sensor Mount Configurations
During the sensor response tests, accelerometer outputs were measured using a CSI Wavepak
which is a two-channel FFT spectrum analyzer integrated with a personal computer. Each
measurement, unless otherwise noted, was taken with 800 lines of resolution and 200 averages.
The instrumentation employed in this study is indicated in the block diagram of Figure 1. "Pink"
noise was generated and passed through a frequency equalizer to provide a relatively constant
output per unit frequency between 0 and 20,000 Hz. Output of the power amplifier was adjusted to
insure an overall RMS value of 1 g in the measured spectra.
Test data were comprised of transfer function, phase and coherence curves determined from test
and reference accelerometer measurements of each mount configuration. The reference

accelerometer (Wilcoxon 726) was always stud mounted on the bottom of the plate and torqued to
20 in-lbs. All test accelerometer measurements were taken from the top of the plate. With the
exception of the hand held accelerometer tests, experiments were conducted using a Wilcoxon
726T accelerometer as the test accelerometer. The manufacturer rated the frequency response of
the 726 accelerometer at 3 to 10,000 Hz (+5%). When stud mounted, the resonant frequency for
this accelerometer is 35,000 Hz. The rated frequency response of the CSI 310 hand held probe stud
mounted and tightened to a vibrating surface by hand is 2 to 2000 Hz.
Test configuration data were normalized by base data in order to reduce or remove measurement
bias resulting from the test procedures employed. Normalization requires complex division of the
test data by the base data. The normalizing process was considered valid provided both the test and
base data had a coherence of 0.95 or above. Base data were taken periodically throughout the study
to verify the consistency of test conditions and results. The base data were acquired with the test
accelerometer (Wilcoxon 726T) stud mounted to the top of the steel test fixture plate to a torque of
15 in-lbs and the reference accelerometer stud mounted to the bottom of this test fixture plate to a
torque of 20 in-lbs.
By normalizing the test data, a transfer function ratio, minimum coherence curve and phase
difference were determined. Any transfer function ratio deviation exceeding 0.90 and 1.10 (greater
than +10%) was considered significant as was a phase difference greater than +-100. These limits
were selected for the purpose of establishing a recommended linear range of application for each
specific mount configuration. Resonant frequencies of each mount configuration were also
determined when possible The resonant frequency of a configuration was noted when a transfer
function ratio peak occurred at a phase difference of approximately 90°·
Results, Discussion and Conclusions. Basic mount configurations considered were: 1. stud
mount, 2. adhesive mount, 3. quick-lock mount, 4. magnet mount, and 5. hand held probe. For the
most consistent and reliable data, it is best that readings from any given measurement point be
taken from smooth surfaces where the path from the sensor to the monitored point is direct. This is
best accomplished by stud mounting a steel disk at desired measurement points. The only
requirement for stud mounting a disk to a machine is that the machine surface be smooth and flat.
A 1 in. diameter, 3/8 in. thick steel disk was stud mounted to the vibrating surface in this study.
The frequency range of the accelerometer stud mounted to this fixture was 3 to 16,125 Hz as
dictated by the phase difference boundary. This range is easily large enough for practical
applications. For cases where the mounting surface is round, the disk may be applied using certain
cement adhesives. Evaluation of a good adhesive for such a use will be discussed shortly.
When mounting disks, or any sensor mount, it is important that the machine surface be smooth
especially when higher frequency measurements are desired. Voids between the measured surface
and the sensor will reduce the accelerometer's frequency range. This is best demonstrated by
comparing experiments conducted with and without wax. After applying wax to a fixture,
frequency ranges were enhanced by as little as 925 Hz to as much as 4,750 Hz. The frequency
ranges were improved because voids between the mount fixture and vibrating surfaces were filled
allowing better transmission between contacting surfaces. In practical measurement situations,
applying wax is not feasible due to surrounding debris and excessive temperatures. However in
situations where higher frequencies must be analyzed, wax can be applied provided the

environment is relatively clean and the temperature is below 200° F (wax melts above 200° F).
As alluded to previously, steel disks can also be bonded to machines via adhesives. The only
requirements are that the surface be clean and smooth (no drills or taps needed). Mounting
procedures for the tests in this study were identical to those required in industrial applications.
Each adhesive was applied using the procedures described by the manufacturer of the particular
bonding agent.
The accelerometer's frequency range resulting from the configuration using the Hottinger Baldwin
Messtechnik X60 adhesive is 3 to 8,775 Hz. When the setup used the Loctite Black Max
cyanoacrylate ester adhesive, the accelerometer's frequency range was 3 to 10,975 Hz. No
resonance was observed for either of the adhesive mount configurations. Of the two bonding
agents, Loctite Black Max was easier to apply. However, this adhesive will only bond surfaces that
are flat and smooth. In situations where one or both of the mating surfaces is rounded (not flat),
Hottinger Baldwin is ideal.
Sometimes, sensors must be isolated from mount configurations. In such situations, electrical
insulators are placed between the sensor and mounting interface. The use of insulators tends to
lower the maximum acceptable frequency. Three common insulators were evaluated in this study.
The maximum acceptable frequency of the stud mounted accelerometer with insulators ranged
from 4200 to 5400 Hz. The transfer function ratios and phase difference curves for these three tests
are plotted in Figure 2.

One mounting system used in lieu of the steel disk is the quick lock. This system consists of a base
permanently attached to a measurement point and a cap connected to the sensor The cap is easily
coupled to the base with 3/4 of a turn. This connection is usually tightened by hand. Results of a
test where the base was stud mounted at 15 in-lbs to the vibrating surface and the remainder of the
quick lock system and sensor were hand tightened to the base show an acceptable frequency
response ranging from 3 to about 4,000 Hz (see Figure 3). While a test was not conducted with the
quick lock base cemented to a vibrating surface, the expected frequency response of such a
configuration would be the same as the stud mounted quick lock system, i.e. 3 to 4,200 Hz. The
frequency range of the quick lock configuration is limited because of the quick lock device, not the
method of permanent connection to the vibration surface.
In many situations, permanent vibration mounting disks are not feasible because of accessibility

and location of measurement points. The best solution in this case is to mount sensors on magnets.
Three different magnets were evaluated in the experiments. For each mount, the accelerometer was
stud mounted at 15 in-lbs to the magnet being tested. The frequency range of the accelerometer
mounted on a rare earth magnet was the widest and ranged from 3 to 3,350 Hz. The frequency
response of the Balmac magnet and CSI 906 super magnet were similar with ranges from 3 to
1,775 and 3 to 1,475 Hz, respectively. An example of transfer function ratios from the Balmac
magnet configurations with and without wax are plotted in Figure 4. Even though the rare earth
magnet had a better response, this magnet can only be used on flat surfaces because the face of the
magnet is flat. The Balmac magnet is capable of holding onto flat and slightly rounded surfaces,
while the design of the CSI 906 super magnet allows it to be used on practically any rounded or
flat surface.
The final mount technique studied utilized a CSI 310 hand held probe and stinger configuration.
Hand held probes should be used when measurement points are inaccessible by any other method.
Typical stinger lengths are 2 and 8.5 inches. The maximum acceptable frequency for these
configurations was much less than any of the other tests. For the hand held probe with a 2 in.
stinger attachment, the maximum acceptable frequency was 525 Hz and for the 8.5 in. stinger
attachment it was 350 Hz. It is not recommended that hand held probes be mounted on magnets
because cantilever action can occur which will produce erroneous data.
The most common mount configurations of those evaluated are ranked in Table 1 with respect to
the valid frequency response in decreasing order. The suggested frequency ranges and resonant
frequencies are liberal estimates and will differ slightly under various conditions and for different
sensors. The value observed for the recommended frequency range was influenced by the shape of
the transfer function ratio curve for each mount. If the curve was smooth and sloping gently
upward, the recommended value was chosen objectively at the base of the resonant curve where
the graph was generally flat (linear relationship). Therefore, it is possible for the recommended
maximum value to be larger than that observed.
Table 1 also shows that different mounting configurations tend to amplify measurements about
their resonant frequencies to varying degrees. In general, hand held probe measurements amplify
signals about the resonant frequency less (6x) than do the other configurations. One can expect a
mounting system employing insulators to increase readings about the resonant frequency by about
12 to 16x the actual vibration. Magnetically mounted sensors will amplify vibration around the
resonant frequency by 13x when using a rare earth magnet to over 20x when using a CSI 906 or
Balmac magnet. Finally, when using a quick lock configuration, vibration can be amplified about
the resonant frequency by 10x. These data suggest that measurements should not be taken at or
around resonance when actual amplitudes are desired because the true vibration magnitudes will be
amplified.
Poor Sensor Mounting Practices
Various misapplications associated with sensor mounting were investigated. The equipment and
procedures used for tests in the previous section were applied to the experiments in this section. A
description of the mount configurations tested is outlined in Tables 2 through 4. The maximum
acceptable frequency with respect to the transfer function ratio and phase difference along with the
resonant frequency for each mount are also recorded in Tables 2 through 4.

Results, Discussion and Conclusions. In most "real world" applications, the mounting
configuration can be far from ideal. For example, rocking occurs when the vibrating surface and
the mount or sensor do not fully and completely mate. Undesirable debris accumulation between
the two surfaces tends to rock the sensor. To evaluate the severity of this prob-lem, ideal and
rocking modes were measured for stud mount and magnet mount configurations. Rocking was
induced by randomly placing drill press residue between the test fixture plate and the contacting
surface of several mounting configurations. Extra rocking was introduced into the magnetic mount
by piling metal shavings in a row which created a "seesaw" effect.
The effect rocking has on a sensor's frequency range was apparent. Without the presence of
rocking, earlier results revealed the frequency range for a basic stud mounted accelerometer to be 3
to 17,450 Hz. With rocking, the frequency ranges observed were 3 to 6000 Hz for the sensor
torqued to 5 in-lbs, 3 to 2150 Hz for the sensor finger tightened and only 3 to 775 Hz when the
sensor was loose (Table 2). These results showed reductions in frequency ranges of 11,450, 15,300
and 16,675 Hz, respectively. Degradation in frequency response can be partially attributed to loose
sensor mounting.
Table 1. Maximum acceptable measurement frequency limits and mounting resonances.
Maximum Mounting
Acceptable Resonance
Sensor Mount Configuration* Frequency (Hz)** Freq(Hz) Ampl(g)
Sensor stud mounted to 3/8"

steel plate stud mounted to vi-
brating surface 16,200 >20,000 N.O.
Sensor stud mounted to steel

disk bonded to vibrating sur-
face with Loctite Black Max
cyanoacrylate ester 10,975 N.O. N.O.
Sensor stud mounted to steel

disk bonded to vibrating sur-
face with Hottinger Baldwin
Messtechnik X60 9,000 N.O. N.O.
Sensor stud mounted with a:

bakelite insulator 10,000 14,650 12.0
nylon insulator 9,000 13,750 13.5
fibre insulator 9,000 12,075 15.5
Sensor stud mounted to rare

earth magnet 7,500 12,075 13.0
Sensor mounted to quick con-

nect stud mounted to vibra-

ting surface (hand tightened) 6,000 10,150 9.0
Sensor stud mounted to Bal-

mac magnet 3,000 6,325 21.0
Sensor stud mounted to CSI

906 super magnet 2,000 5,250 22.0
CSI 310 probe using a pointed

2 in. steel stinger † 800 1,475 6.5
CSI 310 probe using a blunt

8.5 in. aluminum stinger † 500 1,075 6.0
* All stud mounts torqued at 15 in-lbs unless otherwise noted.

** Maximum acceptable frequency chosen objectively at the base of the transfer function curve
where the slope of the line begins to increase towards the resonant peak but is still generally flat
(linear relationship).
Freq Mounted resonance frequency.
Ampl Normalized amplitude at the mounted resonance frequency.
N.O. None Observed
† Stingers tightened to accelerometer at 15 in-lbs. Medium pressure applied to hand held probe.
Table 2. Evaluation of rocking sensor mounts.
Maximum Mounting
Sensor Mount Configuration Frequency (Hz) Frequency (Hz)
Metal shavings between contact

ing surfaces of the sensor, stud
mounted to the vibration pad for:
normal mount* 17,450 >20,000
5 in-lbs torque 6,000 14,350
finger tightness 2,150 7,350
loose connection 775 2,725

earth magnet:
normal mount* 2,975 8,875
metal shavings between magnet
and vibration pad 1,800 6,125
metal shavings piled to one side 1,200 5,075
Sensor stud mounted to Balmac

magnet:


and vibration pad 1,200 4,000
Sensor stud mounted to CSI 906

super magnet:
and vibration pad 750 2,650
*Torqued at 15 in-lbs.
Table 3. Evaluation of loose sensor mounts.
Maximum Mounting
Sensor stud mounted to the vibra-

tion pad for:
normal mount* 17,450 >20,000
loose mount 1,050 2,450

earth magnet:
finger tight 2,200 6,725
Sensor stud mounted to Balmac

magnet:
loose mount 500 3,375
CSI 310 handheld probe using a

pointed 2 in. steel stinger:
medium pressure 425 1,210
loose stinger 385 1,045
*Torqued at 15 in-lbs.
Table 4. Results of hand pressure applied to a hand held probe.
Maximum Mounting
CSI 310 hand held probe using a

0.25 in. stud as a stinger at

pressures:
light 395 1,050
medium 550 1,420
heavy 605 1,630
CSI 310 handheld probe using an

8.5 in. aluminum stinger at
pressures:
light 280 750
medium 320 875
heavy 415 1,070
*Torqued at 15 in.-lbs.
Tests conducted with magnet mounts revealed reductions infrequency responses due to rocking as
well. The effects of rocking reduced the maximum acceptable frequencies from as little as 500 Hz
to as much as 1,775 Hz. An example showing the effects of rocking on a magnet mount is seen in
the transfer function ratios of Figure 5.
Results suggest that measurement point surfaces and sensor interfaces be thoroughly cleaned and
connections tightened in order to minimize the influence interface dynamics can have on data,
particularly when data are being taken above 500 Hz. Furthermore, in a situation where a magnet is
walking around on a surface or intermittently losing contact, low frequency motion and impacts
(which are not generated inside the machine but due to a poor interface) will raise the noise floor
of measured vibration spectra.
Mounting looseness occurs when any connection between the sensor and a vibrating machine is
not securely tightened. Looseness often appears after moving from measurement point to
measurement point without periodically checking the sensor and mount connection. Hand contact,
vibrations from the equipment being monitored, and temperature fluctuations can eventually
http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (10 de 22)05/07/2010 19:13:21

loosen sensors from their mounts. Looseness allows the sensor to vibrate randomly. This reduces
the usable frequency range of the sensor and increases the probability of acquiring spurious data.
Two degrees of looseness were evaluated: "finger tight" and "loose mount." The finger tight mount
was achieved by twisting the accelerometer to the mount with the fingers only. Thus, the sensor
was snugly mounted, but not to the degree of hand tightness. The degree of looseness was defined
as "loose mount" when movement between the sensor and the mount could be described as a
random motion.
The usable frequency range measured for the "loosely" stud mounted accelerometer is 3 to 1050
Hz (Table 3), a reduction of 16,400 Hz from the basic stud mount setup. In the experiment where
the sensor on a rare earth magnet was tightened with only the fingers, the frequency range was
reduced by 775 Hz so that the adjusted frequency range was 3 to 2200 Hz. The frequency range
with the accelerometer "loosely mounted" on a Balmac magnet was 4 to 500 Hz, a reduction of
1900 Hz to the regular Balmac magnet mount frequency range. The effect of this loose mount
setup was further observed by evaluating the transfer function ratio and phase difference curve of
Figure 6. The small peaks appearing at various frequencies indicate that the sensor was chattering.
Because of chattering, high erroneous readings can be obtained. Such readings, especially at low
frequencies, could mislead the analyst to conclude that mechanical problems are present.
With regard to the hand held probe, a loose connection produced the sensor frequency range of 3 to
385 Hz, a reduction in frequency response of 40 Hz relative to the preferred hand heldprobe
configuration. While reductions due to mounting looseness were not as severe as found in other
mount configurations, the fact that tests were performed in a clean environment had some effect.
Even though the stinger attachment was loose, the pressure applied to the probe eliminated some of
the problems normally caused by slack joints and chatter. In field conditions, the technician
experiences less than ideal conditions and vibrations occur in the vertical, horizontal and axial
directions (not just vertically). A loose stinger connection in a field environment is likely to
deteriorate the output signal more than it did in our laboratory tests because of the additional
vibrations present in the horizontal and axial directions.

The hand pressure applied to hand held probes will vary between different people and among
individuals. Therefore, tests were conducted to study the effect force variations have on hand held
probes. For this experiment, two stingers, each connected to a CSI 310 hand held probe, were
evaluated at three different degrees of force: light, medium and heavy (see Table 4 for results). Of
the two stingers, the first stinger considered was a 10-32 UNF stud threaded tightly into the sensor
until it extended 0.25 in. beyond the probe's tip. The frequency ranges for the light, medium and
heavy pressure measurements were 3 to 395 Hz, 3 to 550 Hz and 3 to 605 Hz, respectively. For
each decrement of hand force with respect to the heavily applied pressure measurement, the
frequency ranges decreased 55 and 210 Hz. This changing value will vary with increases and
decreases in hand pressure. Figure 7 compares the transfer function ratios for each of the three
hand pressures.
When using the 8.5 in. aluminum stinger attachment, the acceptable frequency response ranged
from 3 to 280, 3 to 320 and 3 to 415 Hz for the data collected using the hand held probe with light,
medium and heavy hand forces, respectively. The frequency range decreased 95 and 135 Hz with
decrease in vertical force.
The results in Table 4 show that the resonant frequency and therefore frequency range of the hand
held probe varies pro-portionally with hand pressure. For measuring vibrations in the upper portion
of the sensor's frequency range, consistent hand pressure is essential. Otherwise, erroneous
readings may be obtained.
Data Collection Inconsistencies
Variability can occur due to measurement location, various people collecting data and individual
inconsistency. The potential for introducing variability in the measured data was evaluated on
bearing housings of rotating machinery utilizing typical data collection methods. A sketch showing
the loca-tions of measurement points for the two bearing housings studied appears in Figure 8. By
collecting cross spectral data, all test data from each bearing housing could be compared and
normalized with respect to its base data. The reference accel-erometer used on the first bearing

housing, which had a rounded surface, was a Wilcoxon 726T stud mounted (15 in-lbs) to a CSI
906 super magnet and placed at position 0 as seen in Figure 8a. For measurements conducted on
bearing housing number 2, which had a flat surface for mounting, the reference accelerometer
(Wilcoxon 726T) was mounted with a rare earth magnet at the position indicated in Figure 8b.
Variability Due to Measurement Location. Data variability due to measurement point location
was evaluated on bearing housing number 1 which contained a shaft turning at a speed of
approximately 4 Hz. A Wilcoxon 726T accelerometer, also mounted on a CSI 906 super magnet,
was used to collect the test measurements. A single spectrum was obtained at each measurement
point over a frequency range of 0 to 2000 Hz by collecting twenty-five averages with 200 lines of
resolution.


Comparisons of spectra computed from reference and test accelerometer measurements, taken
from the negative side of bearing housing number 1, are displayed over the range of 0 to 400 Hz in
Figure 9 and 1200 to 2000 Hz in Figure 10. Spectra (not shown) produced from the reference and
test accelerometer readings were acquired on the positive side of the bearing housing yielding
similar results. The frequency range between 400 and 1200 Hz is not displayed because data in this
range were small in amplitude relative to the other frequencies.
Results showed that the magnitude for any given frequency was different for each change in
measurement location. The largest magnitude at any given frequency occurred in the vertical and/
or horizontal positions; i.e. some vibrations proved to be prominent vertically, while others were
predominant horizontally. It was noted that a mere 2 in. displacement of the sensor on a housing
approximately 40 in. in circumference would result in changed magnitudes ranging from 0.001 to
0.3 g depending on the frequency range evaluated. On other machines, the output change depends
on the size of the vibrating surface. Larger surfaces may allow for more liberal movement of
sensors while small machines may require a more specific measurement location in order to avoid
collecting invalid data.
Measurement points should be chosen and marked in the vertical, horizontal and/or axial
directions. The best marking method would be to mount (stud or adhesive) a washer or mounting
disk onto the machine surface. Not only will these mounts be permanent, but the measuring surface
will be flat. Another method that will improve the reliability of data is to mark the machine

housing for each measurement location with a spot of paint or indention. In general, the more time
and money put into marking measurement points, the more reliable the resulting data. One must be
careful when choosing the position for measurement on a housing surface. Measurement points
should be at positions having solid metal paths from the sensor to the source of vibration.
Interfaces and internal voids can attenuate vibration and provide unreliable data.
Variability Resulting in Measurements by Different People. The variability in vibration

readings taken by different people and the variability in repeated measurements by the same
individual were evaluated on bearing housing number 2 which was running at a speed of 1440 rpm
(24 Hz). All measurements were taken from the same marked point on the bearing hous-ing in
order to eliminate changes in measurement location as a factor contributing to the variability of the
data. Base data were obtained using Wilcoxon 726T accelerometers. Both the reference and test
accelerometer were stud mounted (torque of 15 in-lbs) on rare earth magnets.
All measurements taken from the bearing housing in Figure 8b were evaluated by calculating
differences in magnitude of the reference accelerometer minus magnitude of the test ac-
celerometer for prominent frequency ranges of concern. By comparing differences, the variance
due to motor speed and reference sensor location was eliminated.
Variability introduced when different people collect data was assessed on the basis of the two tests
described below:
1. Five individuals acquired acceleration measurements using a hand held probe (CSI 310) with a
pointed 2 in. steel stinger attachment. Spectra with a frequency range of 0 to 2000 Hz were
obtained with six averages and 400 lines of resolution.
2. Five individuals took acceleration measurements with the same Wilcoxon 726T accelerometer
mounted on a rare earth magnet. Spectra with a frequency range of 0 to 2000 Hz were obtained
with six averages and 400 lines of resolution.
In the first test, reference measurements were very similar and are not plotted. However, spectra
calculated from test measurements are displayed in Figure 11. Energy within each spectrum was
not evenly distributed over all frequencies, but was dispersed into groups within various frequency
ranges. To compare the results of each person's measurements, the spectra were divided into major
frequency ranges (or groups) which contained energy significant for evaluation. These groups are
indicated by the vertical lines in Figure 11. Within each frequency range, an overall RMS
acceleration value in g was obtained.

The largest spread in magnitude difference for any frequen-cy group between 0 and 990 Hz was
about 0.3 g. This value was for the range of frequencies between 885 and 990 Hz. Above 990 Hz,
hand held probe measurements approached the resonant frequency of the sensor system. This
phenomenon was characterized by large deviations between the reference accel-erometer readings
and individual readings.
The previously reported results revealed the amount of variability (or non-repeatability) between
readings by each individual. To determine the accuracy of the test sensor (hand held probe), the
magnitude differences calculated for each measurement within a particular frequency group were
subtracted from the magnitude differences calculated for the base data of the same frequency
range. The base data were considered to be the "actual" or "true" vibration measurements. A graph
showing the results of these calculations is seen in Figure 12.

The hand held probe resonant frequency changed with each individual's hand pressure. One
measurement (person 2, Figure 12) was substantially different from the others. The resonant
frequency for this reading was approximately 1175 Hz. Other resonant frequencies, for the same
hand held probe, were in the neighborhood of 1510 Hz. These results demon-strate a problem that
can be encountered when taking data with a hand held probe, i.e. applied pressure to the probe will
vary. The more it varies, the less repeatable and reliable the readings. Varying pressures applied to
hand held probes are likely when different people take the same measurements. Even though the
outlying point between 885 and 990 Hz (Figure 12) was present, the magnitude was only 0.4 g.
Therefore, measurements taken by various individuals are accurate for frequencies up to 1000 Hz
when using a hand held probe, as long as they are taken from the same measurement position.
The second experiment included a test accelerometer (Wilcoxon 726T) mounted on a rare earth
magnet. Spectra calculated from the test accelerometer measurements are plotted in Figure 13.
Frequency ranges were evaluated as shown by the vertical boundary lines present in Figure 13.

Variations between the two major peaks around 1000 and 1100 Hz were observed in Figure 13.
Smaller variations were also present at the frequencies ranging from 1300 to 1750 Hz.
Since the reference and test accelerometers were mounted in the same configuration, most
variations are due to differences in individual placement of the magnet on the marked spot, the

offset location of the test accelerometer from the reference accelerometer and/or the inducement of
rocking because of improper placement.
The largest magnitude difference for any individual measurement was 0.7 g which occurred
between 1140 and 1750 Hz. With respect to accuracy, the frequency range between 1140 and 1750
Hz had the greatest deviation of 0.8 g (see Figure 14). This amount was not large enough to
warrant limitation of the frequency range because 0.8 g is a relatively small change over the entire
610 Hz frequency range (1140-1750 Hz). Therefore, in "real world" applications, measurements
taken by different people with a magnetically mounted accelerometer can be expected to vary 0.1 g
for high energy frequency intervals less than 1000 Hz. Magnitude variation within high energy
frequency segments may be as much as 0.6 g for measurements up to 2000 Hz
Variability Resulting from Multiple Measurements by One Person. In industry, one person
will often be responsible for collecting data. After hours of such measurements, fatigue can be
transferred into carelessness which can then affect data reliability. Therefore, a test was
administered to determine how repeatable and accurate a single person can measure vibrations over
a period of time. Variability in repeated read-ings collected by the same individual was based on
20 mea-surements with a hand held probe (CSI 310) at a marked posi-tion on the bearing housing
(see Figure 8b). Between each reading, the hand held probe was removed. For each measure-ment,
a spectrum using six averages and 800 lines of resolu-tion was acquired with a frequency range of
0 to 4000 Hz.
Spectra from the test measurements are plotted in Figure 15 for the base and first seven tests. The
remaining 13 replications produce spectra with similar results and are not plotted. Data above 2300
Hz were not considered because very little energy was observed in this frequency region.

The largest spread in magnitude difference below 960 Hz was 0.3 g. As the resonant frequency of
the hand held probe was approached, the measurements were not as accurate nor repeatable.
However, when using a hand held probe (CSI 310) and measuring vibrations from a given point,
data are repeatable and accurate from 0 to 960 Hz. Therefore, the results suggest that a person can
collect valid data between 0 and 1000 Hz at scheduled intervals provided machines are properly
marked for repetitive data acquisition. Furthermore, when collecting field data with a hand held
probe, one can expect a measurement variation of 0.3 g for high energy frequency groups in the
range of 0 to 1000 Hz. Employing more than one person to collect data can increase the variability.
Recommendations
Throughout this study, many mount configurations have been evaluated. Variations which can
affect the response of sensors used in these configurations were investigated as well. The main
theme derived from the results of these tests is to be consistent and methodical when acquiring data
for spectral analysis. Below is a list of recommendations which will enhance the chances of
obtaining accurate results. Recommendations are:
General
● Limit the measurement frequency range to the linear region of the mounted sensor system
frequency response.
● Limit the number of different individuals recording data.
● Allow only one person to collect data for any given route.
● Always use the same sensor and mount configuration for a particular measurement point.
● Make measurements on clean, smooth and flat surfaces.
● Take measurements in horizontal, vertical and/or axial directions.
● Mark measurement points on machinery to eliminate variation due to measurement location.
● If possible, mount a steel disk on the machine at the measurement point.
● Tighten sensors (preferably to 15 in. lbs torque) to the mount configuration to avoid looseness.
This rule applies to

● any configuration whether it includes a steel disk, a magnet, a quick lock or hand held probe.
● The more random the speed and load of a machine, the more averages required per
measurement. For normal situations, 4 to 6 averages are sufficient. For vibrations that are more
random, as many as 12 averages may be required.
● Avoid measurements where rocking may be induced. Adjust sensor position to eliminate
rocking.
Magnet Mounting
● Keep magnet bottom (feet) clean to avoid rocking and insufficient contact with the machine
surface.
● Do not mount hand held probes on magnets because of possible cantilever action.
● Use as light a weight of cable as feasible. The heavier and stiffer the cable, the greater likelihood
of cantilever action.
Hand Held Probe Use

● Keep a constant hand pressure for every measurement taken.
● The shorter the stinger attachment, for a particular vibrating structure, the better the frequency
response.

● Do not allow a hand held probe to bounce while collecting data. Bouncing will be observed as
chattering in the resulting spectra.
● Avoid tilting the probe. Take measurements so that the length of the probe is perpendicular to
the surface of the machine being analyzed.
References
1. Serridge, M. and Licht, T. R., Piezoelectric Accelerometer and Vibra-tion Preamplifier
Handbook, Brüel & Kjaer, Denmark, pp. 28-29, 88-103, 1987.
2. Carlin, J. B., "A Survey of Factors Which Affect the Measured Vibra-tion Spectra of Machines,"
IRD Mechanalysis, Inc., Columbus, OH.
3. Advanced Training Manual, IRD Mechanalysis, Columbus, OH, pp. 61-66, 1980.
All contents copyright © 1998 - 2006, Computational Systems, Inc.

All Rights Reserved.

Mounting of Accelerometers

Uploaded by

Copyright:

Available Formats

Mounting of Accelerometers

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mounting of Accelerometers

Uploaded by

Copyright:

Available Formats

Emerson Process Management - CSI

DoctorKnow® Application Paper

Real-World Mounting of Accelerometers for Machinery Monitoring

S. V. Bowers, K. R. Piety and R. W. Piety,

Typical accelerometer mounting configurations were evaluated to determine acceptable

http://www.compsys.com/DRKNOW/APLPAPR.NSF/apw...E03E63DB09436840852565A2006504E6?OpenDocument (1 de 22)05/07/2010 19:13:21

mount fixtures are reported in this article.

Sensor Mount Configurations

http://www.compsys.com/DRKNOW/APLPAPR.NSF/apw...E03E63DB09436840852565A2006504E6?OpenDocument (2 de 22)05/07/2010 19:13:21

http://www.compsys.com/DRKNOW/APLPAPR.NSF/apw...E03E63DB09436840852565A2006504E6?OpenDocument (3 de 22)05/07/2010 19:13:21

http://www.compsys.com/DRKNOW/APLPAPR.NSF/apw...E03E63DB09436840852565A2006504E6?OpenDocument (4 de 22)05/07/2010 19:13:21

http://www.compsys.com/DRKNOW/APLPAPR.NSF/apw...E03E63DB09436840852565A2006504E6?OpenDocument (5 de 22)05/07/2010 19:13:21

Poor Sensor Mounting Practices

http://www.compsys.com/DRKNOW/APLPAPR.NSF/apw...E03E63DB09436840852565A2006504E6?OpenDocument (6 de 22)05/07/2010 19:13:21

Table 1. Maximum acceptable measurement frequency limits and mounting resonances.

Sensor stud mounted to 3/8"

Sensor stud mounted to steel

Sensor stud mounted to steel

Sensor stud mounted with a:

Sensor stud mounted to rare

Sensor mounted to quick con-

http://www.compsys.com/DRKNOW/APLPAPR.NSF/apw...E03E63DB09436840852565A2006504E6?OpenDocument (7 de 22)05/07/2010 19:13:21

ting surface (hand tightened) 6,000 10,150 9.0

Sensor stud mounted to Bal-

Sensor stud mounted to CSI

CSI 310 probe using a pointed

CSI 310 probe using a blunt

* All stud mounts torqued at 15 in-lbs unless otherwise noted.

Table 2. Evaluation of rocking sensor mounts.

Metal shavings between contact

Sensor stud mounted to rare

Sensor stud mounted to Balmac

http://www.compsys.com/DRKNOW/APLPAPR.NSF/apw...E03E63DB09436840852565A2006504E6?OpenDocument (8 de 22)05/07/2010 19:13:21

normal mount* 2,400 8,525

Sensor stud mounted to CSI 906

Table 3. Evaluation of loose sensor mounts.

Sensor stud mounted to the vibra-

Sensor stud mounted to rare

Sensor stud mounted to Balmac

CSI 310 handheld probe using a

Table 4. Results of hand pressure applied to a hand held probe.

CSI 310 hand held probe using a

http://www.compsys.com/DRKNOW/APLPAPR.NSF/apw...E03E63DB09436840852565A2006504E6?OpenDocument (9 de 22)05/07/2010 19:13:21

CSI 310 handheld probe using an

http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (10 de 22)05/07/2010 19:13:21

http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (11 de 22)05/07/2010 19:13:21

Data Collection Inconsistencies

http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (12 de 22)05/07/2010 19:13:21

http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (13 de 22)05/07/2010 19:13:21

http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (14 de 22)05/07/2010 19:13:21

http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (15 de 22)05/07/2010 19:13:21

Variability Resulting in Measurements by Different People. The variability in vibration

http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (16 de 22)05/07/2010 19:13:21

http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (17 de 22)05/07/2010 19:13:21

http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (18 de 22)05/07/2010 19:13:21

http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (19 de 22)05/07/2010 19:13:21

http://www.compsys.com/DRKNOW/APLPAPR.NSF/ap...03E63DB09436840852565A2006504E6?OpenDocument (20 de 22)05/07/2010 19:13:21

● Make measurements on clean, smooth and flat surfaces.

● Take measurements in horizontal, vertical and/or axial directions.

● Mark measurement points on machinery to eliminate variation due to measurement location.