PeakVue

ONE STEP IN YOUR JOURNEY TO BENCHMARK STATUS

PEAKVUE
WORLD HEADQUARTERS
835 Innovation Drive
Knoxville, Tennessee 37932
Phone: (423) 675-3200
Fax: (423) 675-3205

DETROIT TRAINING CENTER

39830 Grand River Ave. Ste. B-1-A
Novi, Michigan 48376
Phone: (248) 426-0472
Fax: (248) 426-0574

HOUSTON TRAINING CENTER

15425 North Freeway, Ste. 160
Houston, TX 77090
Phone: (281) 873-6000
Fax: (281) 873-6633

SAN DIEGO TRAINING CENTER

4180 Ruffin Road, Ste. 115
San Diego, California 92123
Phone: (619) 571-8882
Fax: (619) 571-8887

ONE

STEP IN YOUR JOURNEY TO BENCHMARK STATUS

Copyright 1997, Computational Systems Incorporated. All rights reserved.

Content for this manual provided by CSI Training Instructor(s).
Formatting and editing by M. Angela West.
0997

Table of Contents
! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

Section

Title

Introduction to PeakVue

Digital Signal Processing

Case Histories and Applications

Appendix A

PeakVue Data Collection Techniques for the RBM

Analyst

Appendix B

An Improved Methodology for Defect Detection in

Gear Boxes

Section One
Introduction to PeakVue
! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

PeakVue Training
Introduction to PeakVue

Objectives
! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

1. Comprehend the different uses of PeakVue techniques.

2. Have a basic understanding of PeakVue processing
! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

Introduction
This section presents the background information and PeakVue techniques. PeakVue
processing is proving to be the preferred technique for detection of machine defects in
traditionally difficult applications. This processing has been effective in both slow speed
and high speed applications, typically far earlier than routine spectral and waveform
vibration data collection.

PeakVue
PeakVue stands for peak value and can be used to detect short duration higher
frequency waves called stress waves. Stress waves are created when metal is bent,
impacted, cracked or scuffed. For example, if a rolling element in a bearing passes
over a defect which may still be below the surface of the inner or outer race, the race
will deflect and then spring back. This motion will create a sine wave. This sine
wave, or Stress Wave will have the following characteristics:

Y
Y

Short term transient - a few milliseconds

High frequency - 1kHz to 50 kHz

PeakVue measures the highest amplitude found in this waveform and holds it as the
highest value during a waveform time length equal to 1 over the sampling rate.
PeakVue also passes the waveform data through a high pass filter to remove
unwanted high amplitude vibration frequencies from the data. The high pass filter
should be set above the conventional Fmax of the spectrum.

1-2

Copyright 1997, Computational Systems Incorporated. All rights reserved.

PeakVue Training
Introduction to PeakVue

Filters
The PeakVue gives you the option of two types of filters:

Band Pass
High Pass

Each one filters data out of a signal, which you may find useful when analyzing
signals with large dynamic ranges. For example, some spectra have both large and
small amplitudes relative to each other. Because of the dynamic range of the
analyzer, however, you cannot analyze the low amplitude vibration in the same plot as
the high amplitude vibration. A filter helps resolve the problem.

Dynamic Range
Dynamic range is defined as the ability of the analyzer to distinguish between the
highest and lowest amplitude signals. It is controlled by the Analog to Digital (A/D)
processor. The 2120 has a greater than 90 dB dynamic range. Decibel (dB) units are
used because the logarithmic scale makes it easier to describe numbers that are very
large and very small with the smaller numbers. Filters can increase the effective
dynamic range by omitting frequencies of high amplitude, allowing the lower
amplitude signals to be shown.

Band-Pass Filter

There are two types of bandwidth filters S Constant Percentage Bandwidth and
Constant Bandwidth. These filters primarily serve the same function. PeakVue can
utilize the constant bandwidth type filters of: 20-150 Hz,50-300 Hz, 100-600 Hz, and
500-1,000 Hz.

Copyright 1997, Computational Systems Incorporated. All rights reserved.

1-3

PeakVue Training
Introduction to PeakVue

Notice the difference between the two types of filters in the example above. As
shown, the filter to choose is the Constant Bandwidth Filter, because it provides the
best resolution between both high- and low-frequency components.

High-Pass Filter

The high-pass filter gives you the ability to filter out low-frequency components for
detailed analysis as illustrated above. This proves useful when low-frequency, highamplitude data swamps the high-frequency, low-amplitude data you want to see. This
situation often occurs when high-frequency events appear in the same plot as Run
Speed and its related harmonics.
Selection of the high pass filter frequency filter is the most important consideration
when using PeakVue. The goal of the filtering process is to remove the rotational
vibration frequencies such as turning speed harmonics, bearing frequencies, multiples
of gear mesh frequency, etc. The high pass filter should be selected to remove these
rotational frequencies. Therefore, select a filter above the highest operational or defect
frequency that you can estimate.
Although the information we are looking for is lower frequency in nature, we want to
use acceleration as our data analysis unit.
1-4

Copyright 1997, Computational Systems Incorporated. All rights reserved.

PeakVue Training
Introduction to PeakVue

Generally, the 1000-Hz high pass filter is a good choice. The PeakVue process alters
the length of short-duration, very high-frequency events occurring in a machines
vibration data. Be sure to set up a measurement point that uses an accelerometer
and does not integrate the data

Measurement Considerations
Keep in mind that the PeakVue reading is really a high frequency measurement even
on low speed equipment. The following are recommendations for making
measurements.

Y
Y
Y
Y

Use an accelerometer with a high frequency response (>5kHz)

.100 volts per G acceleration preferred (.500 volt/G not recommended)
Rare earth, flat magnet or stud mount (2 pole magnet not recommended).
Flat, metal to metal contact between accelerometer and machine being measured ,
no paint or dirt.
Y Use a coupling agent between metal interfaces (bees wax, grease)

Data taken without PeakVue enabled.

Data taken with PeakVue enabled.

Copyright 1997, Computational Systems Incorporated. All rights reserved.

1-5

PeakVue Training
Introduction to PeakVue

2120 Setup
PeakVue can be accessed from the 2120 Acquire Spectrum Menu.

Y FREQUENCY

Chosen as in normal analysis

The actual frequencies are limited to the following in PeakVue:
20 Hz, 50 Hz, 100 Hz, 200 Hz, 400 Hz, 500 Hz, 1kHz, 2kHz and 5khz.

1-6

Y LOW CUTOFF

Normally 0 (zero)

Y LINES

1600 for trending, 3200 for problem diagnosis

Y AVERAGES

4-6 is fine

Y SST CONTROL

Set to NO

Y UNITS/INTG

Acceleration/Analog- This keeps the units in Gs.

The integration method then has no effect
because we are not integrating (acceleration<----acceleration. )

Copyright 1997, Computational Systems Incorporated. All rights reserved.

PeakVue Training
Introduction to PeakVue

Toggle to the third page of the setup menu.

Y DEMODULATE
Y PEAKVUE
Y PREFILTER

Set to NO
Set to YES
Choices are: 500, 1000, 2000, 5000, 10kHz and
20kHz High Pass or 500-1000 Hz Band Pass.

Collection of at least one data point per machine component is recommended using
the PeakVue technology for early detection of bearing and gear defects, shaft to seal
rubs, torsional vibration as well as other machine problems. For machines that run at
less than 600 RPM, PeakVue should begin to replace normal processing for all
readings.

Copyright 1997, Computational Systems Incorporated. All rights reserved.

1-7

PeakVue Training
Introduction to PeakVue

MasterTrend Setup
This can also be activated from a MasterTrend Analysis Parameter Set definition. One
technique is to setup many generic PeakVue sets and modify them as needed.

1-8

Copyright 1997, Computational Systems Incorporated. All rights reserved.

PeakVue Training
Introduction to PeakVue

In this database, nine sets have been created. The names explain the FMAX and High
Pass filter.

Copyright 1997, Computational Systems Incorporated. All rights reserved.

1-9

PeakVue Training
Introduction to PeakVue

This set may be used for a shaft speed less than .5 Hz (30 RPM).

The Signal Processing Parameters tab is where PeakVue is enabled. The high pass
filter is chosen to be above the FMAX . The selection of the HP filter is based on the
desired elimination of any unwanted frequencies to increase dynamic range.

1-10

Copyright 1997, Computational Systems Incorporated. All rights reserved.

PeakVue Training
Introduction to PeakVue

A measurement point must be setup to store the PeakVue data.

The Units Type Code should be set to Accel<---------Accel. Analysis Parameter set 19
is an appropriate choice for this point.
Copyright 1997, Computational Systems Incorporated. All rights reserved.

1-11

PeakVue Training
Introduction to PeakVue

The second tab contains the sensor and signal information

Sensor Sensitivity (V/EU) should be 0.1. A high frequency accelerometer is

recommended. Add the point to a Route and you are ready to collect PeakVue data.

Summary
This technique is very powerful as a diagnostic tool and early detection method. In
this section, we have examined the setup and considerations for PeakVue as well as
methods of enabling this feature from both the 2120 Machinery Analyzer and
MasterTrend.

1-12

Copyright 1997, Computational Systems Incorporated. All rights reserved.

Section Two
Digital Signal Processing
! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

PeakVue Training
Digital Signal Processing

Objectives
! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

Y Relate time waveform length and frequency bandwidth to sampling rate and
sample size.

Y Choose the correct analysis window for each vibration analysis opportunity.
Y Recognize limitations of digital signal processing.
Y Have the ability to use the 2120 analyzer and the Analyze function to
acquire data utilizing PeakVue.

! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

Fast Fourier Transform

The conversion of time domain information to frequency domain information is
the Fast Fourier Transform (FFT).

Often a frequency spectrum is referred to as an FFT. However, the FFT is the

mathematical conversion from the time domain to the frequency domain. Since
the signal that comes into the analyzer is an analog signal as discussed in the
previous section, it must be digitally sampled by the analyzer. Therefore, the
process used by digital analyzers is actually a variation of the FFT, called the
Discrete Fourier Transform (DFT).

2-2

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Digital Signal Processing

For the DFT, the time waveform is recreated in the analyzer by digitally
sampling, and then it is transformed into the frequency domain. The FFT
process works based on the assumption that the signal measured and digitally
sampled is one period of a periodic signal that extends to minus infinity and to
plus infinity. Normally, this is true for most vibrating pieces of equipment.

Instantaneous Sampling-Normal Processing

It is the digital sampling process that makes the signal processing more
complicated. The information here unlocks the mysteries of digital signal
processing without getting bogged down in too much theory.
In order to understand the FFT digital sampling process, you must understand
the relationship between lines of resolution (LOR), maximum frequency (Fmax),
length of time waveform (Tmax), the digital sample size, filters, and unit
conversion.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

2-3

PeakVue Training
Digital Signal Processing

Resolution
Once data has been converted to the frequency domain from the time domain,
view all the frequencies of interest in as much detail as possible. Resolution is
the number of parts of the spectrum, usually called lines of resolution (LOR).
The number of lines of resolution selected are divided into the maximum
analysis frequency (Fmax) to arrive at the bandwidth (BW).
BW = Fmax/LOR

The lines are actually the center frequencies of what are often called bins of
energy. Each bin actually contains an infinite number of frequencies, and all
the energy in the bin is summed and represented by a single amplitude at the
center frequency identified at each line of resolution.
First, identify your frequencies of interest so that enough resolution is chosen to
separate closely spaced frequencies. A common L.O.R. for PeakVue is 1600
lines. Also, be aware that more lines of resolution affect the length of the time
waveform. For normal trending, we have to weigh the pros and cons of higher
resolution. Remember that the time to collect one average is equal to 1 divided
by the bandwidth. As the bandwidth decreases, the data collection time
increases The bandwidth (BW) should be no greater than 5 HZ/Line.

2-4

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Digital Signal Processing

This will give adequate resolution for identifying trend changes and reasonable
data collection time.
One popular way of setting Fmax is to use an order based set based on turning
speed of the shaft being monitored. Lets take a look at the effect of RPM on
the Sample Rate with a typical 70 x Turn Speed Rolling Element Bearing
Set:

The drawing above represents the sampling of instantaneous values to represent

a sine wave. This is fine when the sampling rate is high enough relative to the
frequency being sampled. As the bandwidth or Fmax is lowered, the sampling
rate decreases and can not accurately represent a high frequency, transient
event such as a stress wave.
PeakVue samples a time interval for a maximum value at the maximum
sampling rate independent of the bandwidth. Even very low Fmax settings are
still able to capture a peak event from stress waves.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

2-5

PeakVue Training
Digital Signal Processing

RPM

x 70
= Fmax

* 2.56
=Sample
Rate

60 HZ (3600 CPM)

4,200

10,752/sec
.

20 HZ (1200 CPM)

1,400

3584

70

179

1 HZ (60 CPM)

If the time window is held constant at .01 seconds, an Fmax of 4200 hz would
give 107 samples (10,752 100). PeakVue would take the above time
increment, sample it at 100,000 samples per second, independent of the Fmax.

Sampling Rate Limitations-Normal Processing

The highest value would be captured for processing. Transient events, like
those produced in stress waves from defects or impacting in metal are easily
displayed.

2-6

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Digital Signal Processing

Sampling Rate Example - PeakVue Processing

PeakVue, through some elaborate signal processing, captures high frequency,

transient events, extends them in time and processes this signal for the time
waveform and for FFT processing of the spectrum. The real amplitude is
captured, so unlike other forms of processing, like Demodulation, this value
can be trended.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

2-7

PeakVue Training
Digital Signal Processing

Time Record Length

Calculate the time record length of the time waveform, Tmax, from the following
basic relationship.
Tmax = 1 / BW
or
Tmax = LOR/Fmax
At face value, this is a simple and often used equation. However, to understand
the limitations of some analyzers, it is important to more fully investigate the
relationship between the Fmax, the LOR, and the Tmax.
Tmax = sample size / sample rate
Sample size = 2.56 * Lines of Resolution
Sample rate = 2.56 * Fmax

These terms have already been defined, but be aware that some analyzers have
an upper limit on the sample size. In the 2120, this number is 4,096.
Therefore, a 400-line spectrum would require 2.56 * 400 = 1,024 samples, a
800-line spectrum would require 2.56 * 800 = 2,048 samples, a 1600-line
spectrum would require 2.56 * 1600= 4096 samples and a 3200-line spectrum
would require 2.56 * 3200= 8,192 samples. However, if the analyzer is limited
to 4,096 samples, then the 3200-line spectrum will be created from 4,096
samples since it is the upper limit of the analyzer. This is important when
discussing the Tmax in the time waveform, because, in general, raising the Fmax
decreases Tmax, and raising LOR increases Tmax to the point that the multiple of
2.56 * LOR reaches the sample limit in the analyzer. In this case, the sample
size for anything greater than 1600 lines is forced to be 4,096.

2-8

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Digital Signal Processing

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

2-9

PeakVue Training
Digital Signal Processing

The table below demonstrates how this limitation affects the Tmax for a
maximum 4,096 sample size analyzer.

Tmax = sample size / sample rate

Fmax

Sample Rate (Sr) =

Fmax*2.56

Sample Size (Ss)=

L.O.R.*2.56

Time (sec.)=
Ss Sr or
LOR Fmax

400

1024

400

1024

1.00

400

1024

800

2,048

2.00

400

1024

1600

4,096

4.00

400

1024

3200

8,192
( 4,096 stored by
MT))

L.O.R.

8.00
(4.00 stored
by MT)

The last entry in the table may seem incorrect, but remember 4,096 is the
maximum sample size, and anything that would be greater than 4,096 is forced
to be 4,096 for storage in MasterTrend. This is the reason your time waveform
is not affected when following the raise-the-LOR-to-lengthen-the-timewaveform rule. You must be aware of the upper limit of the sample size and
the number of lines of resolution to which this number corresponds. To
increase the amount of time in the time record, it will become necessary to
adjust the Fmax to a lower value. The following chart show the effect on the time
record of various Fmax settings.
Fmax (Hz)

Sample Rate (Sr) =

Fmax*2.56

L.O.R.

Sample Size (Ss)=

L.O.R.*2.56

Time (sec.)=
Ss Sr or
LOR Fmax

1000

2560

1600

4096

1.6

400

1024

1600

4096

200

512

1600

4096

100

256

1600

4096

16

10
(low freq.
limit)

25.6

1600

4096

160 (maximum
stored by MT)

10

25.6

3200

8192

320 (160 stored

by MT)

Integration and Differentiation

2-10

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Digital Signal Processing

The vibration data that is the input signal into the analyzer is a time-varying
voltage proportional to the vibration measured by the transducer. In other
words, an accelerometer produces a voltage that varies over time relative to the
acceleration measured by the transducer. The voltage amplitude in the time
waveform is converted to the desired amplitude units based on the sensitivity
and conversion factor of the transducer.
Most analyzers have the ability to convert from the measurement units of the
transducer to either of the other two units in the time domain or the frequency
domain. At CSI, integration of the time signal is called analog integration and
integration of the frequency domain is called digital integration.
Integration is a process of converting from acceleration to velocity or
displacement, or converting from velocity to displacement.
Differentiation is the process of converting from displacement to velocity or
acceleration, or converting from velocity to acceleration.
How are these unit types related, mathematically?
D (Displacement) = distance traveled by vibrating object
V (Velocity) = change in Displacement/change in Time
A (Acceleration) = change in Velocity/change in Time
These terms are often represented with the following equations:
D = X
V = X/T
A = X/T/T = V/T
Therefore, if any one of these terms has been measured, integration and
differentiation allows any of the other terms to be calculated, provided the
analyzer or software used is capable of this conversion process.
One drawback to integration is a flare-up of the lower frequency data caused by
the integration process. This effect is often called integration noise or a skislope effect. This very noticeable when integrating from acceleration to velocity
in PeakVue and tends to affect only the lower 1% of the frequencies. This may
cause the overall vibration level to be higher than usual, and it may be excluded
from the calculation of the overall vibration level.

Summary
Copyright 1997, Computational Systems, Incorporated. All rights reserved.

2-11

PeakVue Training
Digital Signal Processing

This section has introduced some signal processing basics. A clear understanding
of signal processing may help the analyst when making decisions on how to setup
a vibration data collection point. PeakVue is a unique process, with great power in
many applications.

2-12

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

Section Three
Case Histories and Applications
! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

PeakVue Training
Case Histories and Applications

Objectives
! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

Y Use PeakVue in multiple applications.

Y Choose the correct setup for each vibration analysis opportunity.
! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

Case History - Chipper Bearing

This equipment is in a large papermill. The chipper is fairly slow speed and runs
24 hours a day. The equipment was on a normal vibration route.

Wood Chipper
SKF 23156C double row sperical roller bearings

900 HP
300 RPM
Synchronous Motor

3-2

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

PK Velocity in In/Sec

The normal data is shown below. The relatively low amplitudes in both the
spectrum and waveform would not have alarmed automatically.
2WYD - #2 CHIPPER CH-1
132-01-02 -CIA CHIPPER INBOARD AXIAL

0.036

ROUTE SPECTRUM
04-APR-97 10:02
OVRALL= .0455 V-DG
PK = .0455
LOAD = 100.0
RPM = 301.
RPS = 5.01

0.030
0.024
0.018
0.012
0.006
0

Acceleration in G-s

100

200
300
Frequency in Hz

400

500

0.5
0.4

WAVEFORM DISPLAY
04-APR-97 10:02
RMS = .0516
PK(+) = .3759
PK(-) = .2328
CRESTF= 7.28

0.3
0.2
0.1
-0.0
-0.1
-0.2
-0.3
0

100

200

300

400
500
Time in mSecs

600

700

800

The analyst suspected a bearing problem and decided to setup to take PeakVue
data. Notice the impact level in the waveform is only .6 Gs Pk-Pk. The real level
will be shown later. With a 500 hz Fmax the sample rate is 1,280 samples/sec.
Utilizing the Analyze function, additional data was stored.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-3

PeakVue Training
Case Histories and Applications

PeakVue Setup
The setup of PeakVue is flexible, so the analyst should not feel intimidated when
using it. Many incorrect setups will still give good data. The choice of Fmax and
the High Pass Filter follow some basic rules.
Setting Fmax For rolling element bearings:
Y
Use 3-10 times the inner race frequency (BPFI)
Y
If you do not know the BPFI, 10 times running speed is a good
estimate for large (> 6 diameter), in this case 50 HZ.

In this case the running speed is 300 RPM or 5 HZ. Then, 3-10 times would yield
an Fmax of anywhere in the range of 150-500 HZ. The correct Fmax would have
been 500 HZ. Remember, dont get too worried about being exact, PeakVue is
forgiving!
Setting the HP Filter:

Y
Y

3-4

Choose the next filter above Fmax

Set filter above high G level vibration (running speed harmonics,
cavitation, gear mesh etc.)

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

The analyst chose a 100 HZ Fmax . The high pass filter should be the next filter
available above Fmax . The filter could have been 500hz, 1Khz, 2Khz, 5Khz,
10Khz, or 20Khz. He chose 2Khz. With the filter we are trying to increase the
dynamic range of the equipment by filtering out any high G levels that would
interfere with stress wave processing. If there are high G levels near 1000hz, set
the filter at 2000hz.The results are shown.
2WYD - #2 CHIPPER CH-1
132-01-02 -CIA CHIPPER INBOARD AXIAL

0.45

Analyze Spectrum
04-APR-97 10:04
(PkVue-HP 2000 Hz)

0.40

RMS = 1.26
LOAD = 100.0
RPM = 300.
RPS = 5.01

RMS Acceleration in G-s

0.35
0.30
0.25
0.20
0.15
0.10
0.05
0
0

20

40
60
Frequency in Hz

80

100

Freq: 2.221
Ordr: .443
Spec: .311

The highest amplitudes in the axial direction are at the Fundamental Train
Frequency (FTF). The cage is impacting, but how high is the real amplitude?
The waveform reveals what the normal processing was missing.
Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-5

PeakVue Training
Case Histories and Applications

Approximately 20 revolutions of the shaft are show (5 rev/sec. Times 4 seconds).

Impacts as high as 18 Gs are shown.
2WYD - #2 CHIPPER CH-1
132-01-02 -CIA CHIPPER INBOARD AXIAL

20

Waveform Display
04-APR-97 10:04

18

RMS = 1.55
LOAD = 100.0
RPM = 300.
RPS = 5.01

16

Acceleration in G-s

14

PK(+) = 17.26
PK(-) = .8885
CRESTF= 11.16

12
10
8
6
4
2
0
-2
0

3-6

0.5

1.0

1.5

2.0
2.5
Time in Seconds

3.0

3.5

4.0

Time: 2.246
Ampl: -.409

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

If we look closer, we can see multiple impacts at every other revolution of the
shaft.
2WYD - #2 CHIPPER CH-1
132-01-02 -CIA CHIPPER INBOARD AXIAL

20

Waveform Display
04-APR-97 10:04

18

RMS = 1.84
LOAD = 100.0
RPM = 300.
RPS = 5.01

16

Acceleration in G-s

14

PK(+) = 17.26
PK(-) = .8885
CRESTF= 11.16

12
10
8
6
4
2
0
-2
0.4

0.6

1.0
1.2
Time in Seconds

1.4

1.6

Waveform Display
04-APR-97 10:04
RMS = 2.09
LOAD = 100.0
RPM = 300.
RPS = 5.01

16
14
Acceleration in G-s

Time: 1.094
Ampl: 17.33

2WYD - #2 CHIPPER CH-1

132-01-02 -CIA CHIPPER INBOARD AXIAL

20
18

0.8

PK(+) = 17.26
PK(-) = .8885
CRESTF= 11.16

12
10

>SKF 23156C
C=BSF
: 20.51

8
6
4
2
0
-2
800

900

1000

1100
1200
Time in mSecs

1300

1400

Time:
Ampl:
Dtim;
Freq:

1093.8
17.33
.00000
.00000

The Fault Frequency lines line up on the 1x BSF (ball spin frequency). We actually
have peaks in between signifying 2x BSF This signifies 2 defect areas, 180
degrees on one roller.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-7

PeakVue Training
Case Histories and Applications

3-8

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

The roller shows heavy electrical fluting. The same pattern appeared 180 degrees
apart on the other side of the roller. Fluting can occur due to a number of causes.
Basically it is a voltage discharge through a bearing. This can occur due to static
buildup and discharge or when a welder uses the machine for ground when
welding a shaft. These defects were clearly seen in PeakVue spectrum and
especially the time waveform.

MasterTrend Setup
To setup this machine as a route point, the following screens should be followed.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-9

PeakVue Training
Case Histories and Applications

Add a measurement point called CPA, Chipper PeakVue Axial . This spherical
roller bearing shows defects in the axial direction best.

3-10

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

Choose the appropriate Analysis Parameter (AP) set. Using the rules from earlier:
Fmax = Shaft speed * 10 (inner race approx.) * 5 (orders)
Fmax = 5 Hz * 10 * 5
Fmax = 500 Hz

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-11

PeakVue Training
Case Histories and Applications

Case History - Pinion Gear on Generator

Steam Turbine/Generator 220 MW

Reduction
Gears

Generator
Exciter
This equipment is in a large electric generating facility. The turbine turns at 3600
RPM and has a pinion gear to a larger gear. This turns the exciter for the
generator.

4 : 1 R e d u c t io n G e a r

G 8 A /P 8 A

G 7 A /P 7 A
Measurement Points for Gears
3-12

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

The normal data is shown below for G7A, gearbox input shaft, inboard side.
PKY9 - TURBIN GENERATOR SLEEVE
TUR-GEN -G7A EXCITER GBOX INBOARD INPUT AXIAL
Route Spectrum
A
A
28-MAY-97 10:09

0.10

OVRALL= .1068 V-DG

PK = .1064
LOAD = 45.0
RPM = 3586.
RPS = 59.77

0.08
PK Velocity in In/Sec

Input Shaft 1x
Gearmesh
0.06

>PIC PICGEN
A=Grmesh(1>2)
: 1500.

0.04

0.02

0
0

1000

2000
3000
Frequency in Hz

4000

5000

Freq: 1494.2
Ordr: 25.00
Spec: .06725

The gearmesh and sidebands made the analyst decide to use PeakVue to get a
better look.

Setup for Gears

To setup PeakVue for gearboxes here are a few rules:
Setup 2 points at each bearing location.

One high frequency (3.5 times gearmesh, 3200 lines to look at

gearmeshing problems). Set the high pass filter above the Fmax.
*Note: This reading may be extremely high frequency, use the
appropriate accelerometer and mounting.

The other point should be setup for the bearing and lower order shaft
information at 25-50 orders of the shaft speed

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-13

PeakVue Training
Case Histories and Applications

To set the points up in MasterTrend:

Choose the appropriate AP set. The first point would be setup as the lower
frequency. This is a sleeve bearing, so no bearing information is needed. We will
be concerned with shaft turning speed harmonics. Choose the Fmax to be 10-15
times turning speed. The input shaft is 60 Hz * 15 = 900 Hz, so choose the 1,000
Hz set.

The second point is the high frequency point at 3.5 times gearmesh. Gearmesh is at
around 1,600 Hz, 3.5 *1600 is 5,600 Hz. Choose the 5,000 Hz Fmax set with 5,000
Hz high pass filter.

3-14

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

Alarm Limits

Setting Alarm Limits (AL) on PeakVue points may have to be done manually with
a different AL set for each point. Experience with different machines and
historical data may be the best guides. Operational characteristics as well as the
Fmax and High Pass Filter Settings will also have to be taken into consideration.
The first time a PeakVue reading is taken, we may assume that the machine is in
good condition (the majority of machines are not in failure - we hope!).
Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-15

PeakVue Training
Case Histories and Applications

Using this as a baseline, we can set an alarm at 100-300 percent of the baseline as
a guideline .
Considerations for Alarm Limits in PeakVue

Y
Y
Y

Historical data
Operational characteristics of the machine
Fmax and High Pass Filter Settings

RMS Acceleration in G-s

The Peak-Peak waveform alarm and Overall spectrum alarm may be the best
alarms to start out with. The set below was chosen based on the baseline data.
PKY9 - TURBIN GENERATOR SLEEVE
TUR-GEN -P7A EXCITER INBOARD PeakVue

0.008

ROUTE SPECTRUM
03-APR-97 12:49
(PkVue-HP 5000 Hz)
OVRALL= .0188 A-DG
RMS = .0187
LOAD = 100.0
RPM = 894.
RPS = 14.89

0.006
0.004
0.002
0

Acceleration in G-s

1000

2000
3000
Frequency in Hz

4000

5000

0.24

WAVEFORM DISPLAY
03-APR-97 12:49
RMS = .0215
PK(+) = .0962
PK(-) = .0328
CRESTF= 4.48

0.18
0.12
0.06
0
-0.06
-0.12
0

10

20

30

40
50
Time in mSecs

60

70

80

The baseline data shows very low levels in both the spectrum and waveform.
This gearbox shows no major defects, so a generic AL set is made.

3-16

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

Gearbox Alarm Limits

Y
Y
Y

Use reference data (assuming machine is in good shape)

Waveform Pk-Pk Alert: 2 Gs ; Alarm: 4 Gs
Overall Spectrum Alert: .1 G ; Alarm: .3 Gs

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-17

RMS Acceleration in G-s

PeakVue Training
Case Histories and Applications

PKY9 - TURBIN GENERATOR SLEEVE

TUR-GEN -P7A EXCITER INBOARD PeakVue

0.12

ROUTE SPECTRUM
03-APR-97 12:49
(PkVue-HP 5000 Hz)
OVRALL= .0188 A-DG
RMS = .0188
LOAD = 100.0
RPM = 894.
RPS = 14.89
Alert Limit

0.10
0.08
0.06
0.04
0.02
0

Acceleration in G-s

1000

2000
3000
Frequency in Hz

2.0
1.5
1.0
0.5
0
-0.5
-1.0
-1.5
-2.0
-2.5
0

4000

5000

WAVEFORM DISPLAY
03-APR-97 12:49
RMS = .0215
PK(+) = .0962
PK(-) = .0328
CRESTF= 4.48

FAULT
ALERT

ALERT
FAULT
10

20

30

40
50
Time in mSecs

60

70

80

Freq: 3074.2
Ordr: 206.42
Spec: .00017

Conclusions
In this case history, the normal spectrum and waveform information showed
possible gear mesh problems with high GMF with sidebands. This has always been
considered a indicator of wear and/or broken teeth. PeakVue refuted that
conclusion by showing that no major impacting or micro welding of metal was
occurring. These gears were inspected with no abnormal wear conditions found.
The high GMF and sidebands of the pinion and bull gear running speed were
found to load related. At lower loads, the gears were backlashing creating
amplitude and frequency modulation. The modulation was reflected by the
sidebands of run speed around GMF in the spectrum. This caused the spectrum
data to appear in alarm.

3-18

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

Case History - Reciprocating Compressor

Reciprocating Compressor

200 HP
1775 RPM Motor

Piston

Double Row
Spherical Roller Brgs
318448

Crankshaft
345 RPM

Reciprocating compressors present many challenges from a vibration analysis

standpoint. The large forces due to the movement and compression of the piston
(1x, 2x RPM) often drowns out the lower amplitude level bearing faults. The
filtering feature of PeakVue was used to increase the dynamic range and attenuate
the forces of the reciprocating vibration.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-19

PeakVue Training
Case Histories and Applications

Normal Processing
PRAX - NITROGEN COMPRESSOR
NITROGEN -CIA COMPRESSOR INBOARD BEARING AXIAL
Route Spectrum
01-OCT-97 11:46

0.12

OVRALL= .4472 A-AN

PK = .0946
LOAD = 100.0
RPM = 346.
RPS = 5.77

PK Velocity in In/Sec

0.10

0.08

0.06

0.04

0.02

0
0

60

120
180
Frequency in Hz

240

300

Freq: 5.625
Ordr: .975
Spec: .03228

The spectrum is dominated by running speed harmonics. Looseness may be

suspected. This is somewhat normal for reciprocating machines. PeakVue was
utilized to determine the severity of the looseness.

3-20

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

PeakVue Processing
PRAX - NITROGEN COMPRESSOR
NITROGEN -CIA COMPRESSOR INBOARD BEARING AXIAL
Analyze Spectrum
01-OCT-97 11:46
(PkVue- HP 500 Hz)

0.10

RMS Acceleration in G-s

0.08

RMS = .2045
LOAD = 100.0
RPM = 347.
RPS = 5.78

0.06

0.04

0.02

0
0

60

120
180
Frequency in Hz

240

300

Freq: 5.781
Ordr: 1.001
Spec: .06494

PeakVue shows harmonics of running speed signifying there is impacting metal as

the shaft rotates, not just reciprocating forces. In addition, some non-synchronous
peaks begin are also now visible.
PRAX - NITROGEN COMPRESSOR
NITROGEN -CIA COMPRESSOR INBOARD BEARING AXIAL
Analyze Spectrum
C
C
C
C
01-OCT-97 11:46
(PkVue- HP 500 Hz)

0.10

RMS Acceleration in G-s

0.08

RMS = .2045
LOAD = 100.0
RPM = 347.
RPS = 5.78

0.06

>KOY JHM318448/10
C=BPFO : 46.83

0.04

0.02

0
0

60

120
180
Frequency in Hz

240

300

Freq: 47.34
Ordr: 8.195
Spec: .04590

The outer race and inner race (BPFO, BPFI) fault frequencies are visible,
signifying faults in the crankshaft bearings. They were lost in the noise in normal
processing.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-21

PeakVue Training
Case Histories and Applications
PRAX - NITROGEN COMPRESSOR
NITROGEN -CIA COMPRESSOR INBOARD BEARING AXIAL
Analyze Spectrum
D
D
D
01-OCT-97 11:49
(PkVue-HP 1000 Hz)

0.10

RMS Acceleration in G-s

0.08

RMS = .1572
LOAD = 100.0
RPM = 347.
RPS = 5.79

0.06

>KOY JHM318448/10
D=BPFI : 63.01

0.04

0.02

0
0

60

120
180
Frequency in Hz

240

300

Freq: 62.50
Ordr: 10.80
Spec: .01953

The waveform shows 1 G impacting, a moderate fault. The bearing was reported
as being in stage 3 of bearing failure due to the high number of harmonics in the
spectrum and levels in the waveform.
PRAX - NITROGEN COMPRESSOR
NITROGEN -CIA COMPRESSOR INBOARD BEARING AXIAL
Waveform Display
01-OCT-97 11:49

1.2
1.0
0.8
0.6
Acceleration in G-s

RMS = .1947
LOAD = 100.0
RPM = 347.
RPS = 5.79

FAULT
ALERT

PK(+) = 1.01
PK(-) = .4237
CRESTF= 5.20

0.4
0.2
-0.0
-0.2
-0.4
-0.6

ALERT

-0.8

FAULT

-1.0
0

3-22

0.3

0.6

0.9

1.2

1.5 1.8 2.1

Time in Seconds

2.4

2.7

3.0

3.3

Time: 1.599
Ampl: -.190

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

Conclusions
The filtering feature of PeakVue was used to eliminate a large amplitude vibration,
increasing the dynamic range so lower amplitudes could be seen. The metal
impacting caused stress waves identifying severe looseness in the bearings. The
bearings were changed and found to be loose and have significant eye visible
faults.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-23

PeakVue Training
Case Histories and Applications

Case History - High Speed Grinding Spindle

High Speed Grinding Spindle

3600 RPM
Wheel Motor

15-18,000 RPM
Wheel Grinder

This spindle grinder grinds the race of a wheel bearing. The high speed bearings
are oil mist lubricated. High speed bearings need to be monitored on a more
frequent schedule because minor defects can progress into major defects quickly.

3-24

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

RMS Acceleration in G-s

Progression of Fault
GRND - D18 WHEEL
D18 WHEEL -VOH SPINDLE OH D18 PEAKVIEW

0.10

ROUTE SPECTRUM
05-MAR-97 13:46
(Filtr- HP 600 Hz)
OVRALL= .0789 A-DG
RMS = .0788
LOAD = 20.0
RPM = 15152.
RPS = 252.54

0.08
0.06
0.04
0.02
0

Acceleration in G-s

1000

2000
3000
Frequency in Hz

2.0
1.5
1.0
0.5
0
-0.5
-1.0
-1.5
-2.0
-2.5

4000

5000

WAVEFORM DISPLAY
05-MAR-97 13:46
RMS = .0853
PK(+) = .3515
PK(-) = .1906
CRESTF= 4.12

PK ALARM

CF ALARM
CF ALARM

PK ALARM

10

20

Label: Good condition

30

40
50
Time in mSecs

60

70

80

Freq: 252.54
Ordr: 1.000
Spec: .00706

When the spindle is in good condition, the levels are low in both the spectrum and
waveform.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-25

RMS Acceleration in G-s

PeakVue Training
Case Histories and Applications
GRND - D18 WHEEL
D18 WHEEL -VOH SPINDLE OH D18 PEAKVIEW

0.40
0.35

ROUTE SPECTRUM
23-APR-97 18:03
(Filtr- HP 600 Hz)
OVRALL= .6731 A-DG
RMS = .6693
LOAD = 20.0
RPM = 15209.
RPS = 253.49

0.30
0.25
0.20
0.15
0.10
0.05
0

Acceleration in G-s

1000

2000
3000
Frequency in Hz

4000

5000

WAVEFORM DISPLAY
23-APR-97 18:03
RMS = .8030
PK(+) = 4.41
PK(-) = 1.04
CRESTF= 5.49

CF ALARM

4
2

PK ALARM

0
-2

PK ALARM

-4
CF ALARM

-6
0

10

20

30

40
50
Time in mSecs

60

70

80

Freq: 253.13
Ordr: .999
Spec: .09424

One and a half months later, the fault has progressed to a major failure. The
monitoring schedule should be adjusted because of the speed. The spectrum shows
impacting at half harmonics indicating severe looseness. The non-synchronous
peaks are severe bearing defects.

Conclusion
PeakVue worked well at detecting this high speed bearing failure. The monitoring
schedule may need to be adjusted as speed increases. Even small faults can
progress to catastrophic at high speeds.

3-26

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

Case History - Slow Speed Gearbox

Gearbox
(half)

280 HP
1190 RPM
AC Motor

This gearbox is used in a mining operation to remove material underground. It

powers two large boring heads at a very slow speed. Normal vibration monitoring
was not very effective at monitoring the gears inside. The complexity of the
spectrum made it difficult to determine when and where the problems occurred.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-27

PeakVue Training
Case Histories and Applications

63 T

17 T

40 T

1 = 20 Hz
2 = 5.39
3 = 1.89
4 = .66
5 = .27
6 = .36

43 T

14 T

15 T

41 T

31 T

17 T

PK Velocity in In/Sec

OUTPUT
SHAFT 16.4 RPM
K1UG - # 20 Miner Normal
#20 NORMAL-R9 Right Side Point 9

0.05

ROUTE SPECTRUM
30-JAN-97 11:25
OVRALL= .0477 V-DG
PK = .0475
LOAD = 100.0
RPM = 1190.
RPS = 19.83

0.04
0.03
0.02
0.01
0

Acceleration in G-s

2.0
1.5
1.0
0.5
0
-0.5
-1.0
-1.5
-2.0
-2.5
0

12
15
18
Frequency in Hz

21

24

27

WAVEFORM DISPLAY
30-JAN-97 11:25
RMS = .0417
PK(+) = .1191
PK(-) = .1304
CRESTF= 3.13

PK ALARM

CF ALARM
CF ALARM

PK ALARM
0.2

0.4

0.6

0.8
1.0
1.2
1.4
Time in Seconds

1.6

1.8

2.0

Freq: 9.853
Ordr: .497
Spec: .02015

Normal processing shows two peaks related to a chain drive.

PeakVue Processing
3-28

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

K1UG - # 20 Miner Peak Vue

#20 PV -L10 Left Side Point 10

0.12

Analyze Spectrum
30-JAN-97 12:56
(Filtr-HP 1000 Hz)
RMS = .3106
LOAD = 100.0
RPM = 1200.
RPS = 19.99

1.32

0.08

>MIN MINER #20

V=Grmesh(4>6)
: 11.20
W=Tth Rept(4>6)
: .0212
X=Speed #6 : .36
Y=Gm(4>6)-Sp4
: 11.86
A=Gm(4>6)-Sp6
: 11.56

2.64

1.98

.66
0.06

4.61

0.04

5.27

3.95

3.29

RMS Acceleration in G-s

0.10

0.02

0
0.5

1.0

1.5

2.0

2.5

3.0 3.5 4.0

Frequency in Hz

4.5

5.0

5.5

6.0

Label: OUTPUT, IDLER,

Freq: .656
Ordr: .03282
Spec: .06006

Harmonics of the #4 shaft are apparent. The #6 shaft harmonics is also visible.
K1UG - # 20 Miner Peak Vue
#20 PV -L10 Left Side Point 10

0.12

Analyze Spectrum
30-JAN-97 12:56
(Filtr-HP 1000 Hz)
.72

W X

RMS = .3082
LOAD = 100.0
RPM = 1200.
RPS = 19.99

0.08
.36

>MIN MINER #20

V=Grmesh(4>6)
: 11.20
W=Tth Rept(4>6)
: .0212
X=Speed #6 : .36
Y=Gm(4>6)-Sp4
: 11.86
A=Gm(4>6)-Sp6
: 11.56
4.69

4.33

3.61

3.97

0.02

2.89

2.17
2.53

1.81
1.44

0.04

3.25

0.06

1.08

RMS Acceleration in G-s

0.10

0
0

0.5

1.0

1.5

2.0

Label: OUTPUT, IDLER,

2.5 3.0 3.5

Frequency in Hz

4.0

4.5

5.0

5.5

Freq: .362
Ordr: .01813
Spec: .06396

The gearbox continued to run, for a while. The picture below shows the results
when the failure occurred.
Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-29

PeakVue Training
Case Histories and Applications

Conclusion
3-30

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

PeakVue Training
Case Histories and Applications

PeakVue is outstanding for slow speed and complex gearboxes. The filtering can
get rid of the high amplitude GMF, and allow us to see the impacting of shafts,
gear teeth and bearings.

Copyright 1997, Computational Systems, Incorporated. All rights reserved.

3-31

 Copyright 1999, Computational Systems Incorporated. All rights reserved.

Doctor Know Application Paper

Title:
Source/Author:
Product:
Technology:

An Improved Methodology for Defect Detection in Gear Boxes

J.C. Robinson
MasterTrend
PDM

ABSTRACT
A microprocessor based instrument has been developed to enhance the analysis of complex
mechanical devices such as gear boxes for defect detection employing vibration analysis
techniques. The physical phenomena relied on for analysis are the stress waves which are emitted
as a result of metal-to-metal impacting, microwelding, etc. For rotating equipment, these stress
waves are generally emitted at a periodic rate; therefore, synchronous averaging can be employed
for signal enhancement. The pulses used for timing pulses from which synchronous averaging
occurs are generated digitally within the microprocessor based instrument. The digitally
generated output pulse string is analytically generated from a reference input pulse string
acquired from the mechanical device under analysis.
The methodology implemented in the new microprocessor based instrument described herein was
applied to a small double reduction gear box with a known defective tooth. The results obtained
from synchronous averaging of the stress waves are compared to synchronous averaging of the
vibration signals (standard methodology). The results obtained using the stress wave analysis
were more informative and simpler to interpret

1.0 BACKGROUND
An accelerometer measures the absolute motion (acceleration) of the surface it is attached to.
Surface motion can occur because of 1) motion of the entire body in a global sense, 2) structural
resonances of the body, and 3) stress waves traveling through the body. Entire body motion is
generally in the lower frequency range and directly proportional to the driving force. Structural
resonances are dependent on the makeup of the body and can significantly amplify motion at
specific (resonant) frequencies. Stress waves are generally set up in a body due to localized
effects such as impacting, microwelding (tearing), etc. For metallic objects impacting on a
metallic surface, the stress waves emitted consist of bending and shear waves. As bending waves
travel away from the initiation point, they cause a small surface motion much like the pressure
waves seen in a pool of water. Shear waves, which travel faster than bending waves do not
introduce surface distortions, but do get converted to bending waves when traveling through
interfaces. For metal-to-metal impacting, the dominant frequency in bending waves will be in the
1KHz to 15 kHz range; however, the impact signature will excite a broad band of frequencies.
The bending waves will attenuate as they travel away from the point of impact, but the higher
frequencies will attenuate more rapidly than the lower frequencies (approximately inversely
proportional to frequency). Furthermore, the frequency content within the impact is dependent on
the impact itself, e.g., small objects or sharp surface impacting will have higher frequencies than
larger objects or smooth surface impacting.
The large variability in frequency content due to the shape and size of the impacting object and
Copyright 1997, Computational Systems Incorporated. All rights reserved.

B-1

PeakVue Training
Appendix B

attenuation due to dispersion within an impacting signature makes a specific frequency difficult
to impossible to monitor. The frequencies within impact signatures are generally greater than
those within bulk vibration or larger structural resonances and hence can be separated and
monitored from a timing and energy perspective. Impacts occurring at the moving shaft (gear
teeth, inner race on bearings, etc) will generate stress waves which must travel through interfaces
where significant attenuation could occur and hence may require signal enhancement (such as
time synchronization) to provide a reliable monitoring technique.
The methodology developed by CSI Patent pending. for stress wave monitoring with time
synchronous averaging capability has been implemented in a microprocessor based instrument
and is described below.

2.0 ANALYSIS PROCEDURE

2.1

Introduction

The methodology employed achieves signal enhancement through synchronous averaging. The
signal to be averaged is transported to the sensor as stress waves. The signal that is averaged is
specially processed to filter out low frequency mechanical contributions and then peak-hold to
insure that short duration impacts do not go undetected. The method used to generate the
synchronous pulse as well as the signal to be synchronously averaged is discussed briefly below.

2.2

Generation of Synchronous Pulses

The input for the digitally generated pulse is a once per rev pulse from an accessible (reference)
shaft. The speed of the output object (shaft, bearing, etc.) must be deterministic relative to the
reference shaft. A microprocessor is used to generate the output tach signal. The output pulse
string can be a higher or lower rate than the input pulse rate. The user can enter a decimated
number or two integers where the ratio determines the output pulse string rate. A level of
intelligence has been built into the processor wherein it predicts the time of an incoming pulse. If
the pulse is not within a reasonable time of the expected time, it assumes that pulse is missing (or
false) and continues the output pulse string based on historic data. This permits an occasional
dropout or an occasional false trigger from the incoming pulse string. Many false inputs (or
misses) will cause the output pulse string to be terminated. These features are very important to
insure that the subsequent synchronous averaging process is not compromised.

2.3

Signal Processing

The signals to be processed are the stress waves setup in metal due to impacting, micro welding,
etc. The primary characteristics of these signals are that they will be of low amplitude, short
duration (5 to 100 msec), and have carrier frequencies in excess of a kilohertz. The signal from
the sensor (an accelerometer) will be routed through a high pass filter, amplified, and then
processed. The information is obtained from the presence of the stress wave and its energy
content; the frequency of the carrier is of less importance. The stress waves typically travel
through a tortuous route prior to reaching the sensor and hence will generally be small.
Therefore, signal enhancement such as synchronous averaging is frequently necessary to make
the defect events detectable.
The basic signal from stress waves (impacts, etc.) general1y cannot be synchronous averaged
B-2

Copyright 1997, Computational Systems Incorporated. All rights reserved.

PeakVue Training
Appendix B

because a) the frequency within the packet varies somewhat due to the nature of the stress
producing source and b) the reasonably high frequency of a few kilohertz would require
unrealistic stability in the synchronous timing pulse. The envelope of the stress packet could be
synchronously averaged but would require a sampling rate in the kilohertz range to assure
detection of the peak (detect the energy within the packet). The methodology adopted herein is to
employ a peak detector which can be reset at a predetermined rate. This ensures the detection of
the peak at reasonable slow sampling rates. The basic technique is to hold the signal from input
to output such that the output in the interval t to t + _t is the maximum of the input signal from
the previous time (t-_t) to t
To illustrate the functionality gained from this approach, consider monitoring a bearing on a shaft
turning at 1 RPM. The fault frequency would be in the 5 to 25 RPM range. Early bearing defects
will generate stress waves (impacts) with time durations in the 5-50 msec range whose peak
value could be very short lived. To look for bearing faults, one would like to capture a time
record consisting of two or more shaft revolutions. Most portable meters have a minimum
(analysis bandwidth) frequency range for the spectral analysis of (FMAX) 10 Hz which means the
minimum sampling frequency is 25.6 samples/sec. For an FMAX of 10 Hz (sampling frequency of
25.6), and a specified block size of 400 frequency lines, the sampled block of 1024 data points
would take 40 sec to fill (specified 1600 frequency lines would take 160 sec to fill). A sampling
rate of 25.6 sec means that one sample occurs approximately every 39 msec. If one were using
envelope detection, the event would be entirely missed many times and synchronous averaging
would be of very little value. Using the peak hold methodology, a hold time of greater than 40
msec would guarantee the peak would always be detected. This ability to correlate the peak hold
time with sampling rate permits the use of synchronous averaging for stress packet analysis on
very slow to fast machinery. Accordingly, one can carry out synchronous averaging for stress
wave analysis on a wide spectrum of machinery using time synchronous averaging.

3.0 Application to a Gear Box

3.1

Introduction

The gear box chosen for demonstration is a reduction gear box having one internal shaft. The
gear box is driven by a 1 horsepower DC motor. The load is relatively light. The input shaft has a
16 tooth pinion driving a 88 tooth gearwheel on the intermediate shaft. The intermediate shaft
has a 17 tooth pinion driving a 54 tooth gear wheel on the output shaft. The total speed reduction
from input to output is:
(16/88)(17/54) = 17/297
It was known that there was a cracked tooth on one of the gears.
The sensor used for analysis was an accelerometer mounted over the bearing on the input shaft.
The tach signal used for reference was from the slow output shaft. The results presented below
are:
1.
2.

No synchronous averaging.
Synchronous for each shaft of the straight vibration signal.
Copyright 1997, Computational Systems Incorporated. All rights reserved.

B-3

PeakVue Training
Appendix B

3.

Synchronous for each shaft for the stress wave using the peak follower methodology.

The last subsection will be a summary of the findings from the results.

3.2 Results
3.2.1 No Synchronous Averaging
The spectra for a four block average. Four blocks were chosen because this is typical for
nonsynchronous averaging. With the time trace for one block of data for an analysis bandwidth
of 800 Hz with 800 lines frequency analysis are presented in Figure 1. In Figure 1, there are no
clear indicators that at least one gear has a defective tooth.

3.2.2 Synchronous Averaging of Vibration Signal

The synchronization on each shaft was accomplished by digitally generating a tach pulse from
the output (slow) shaft reference tach. The multiplication factors used were as follows:
1.

1.0 for synch to output shaft.

2.

54/17 for synch to intermediate shaft.

3.
17.471 for synch to input shaft.
The results obtained for a 25 block average (time and spectra) are presented in Figure 2 through
Figure 4. Examination of the time trace suggests there may be a problem on the input pinion (it is
not a clear conclusion).

3.2.3 Synchronous Averaging of Stress Wave

For this analysis, the output signal from the accelerometer was preprocessed through the system
developed for peak following of the stress wave signal and then routed through an analyzer (the
CSI Model 2120 was used but any similar device would be appropriate).
Results obtained by synchronous averaging 25 blocks in the data collector are presented in Figure
5 through Figure 7 (the input to the data collector was the output of the microprocessor based
stress wave signal conditioner). The stress wave signal conditioner was set up as:
1.

High pass set point: 500 Hz and a delay time of 5 msec.

2.

Multiplication factors for input tach pulse (from gear box output shaft) were:

1.0
forFigures
54/17 forFigure6
17.471 forFigure7

B-4

Copyright 1997, Computational Systems Incorporated. All rights reserved.

PeakVue Training
Appendix B

Accordingly, the synchronous results for Figure 5 are relative to the outlet or slow shaft, the
synchronous results for Figure 6 are relative to the intermediate shaft, and the synchronous
results for Figure 7 are relative to input (or fast) shaft.
From the results presented in Figures 5 through 7 it is obvious that a stress wave (probable
impact) is occurring once per revolution of the input shaft (Figure 7). There also are stress wave
activity occurring 11 times per revolutions of the intermediate shaft (Figure 6). There is no
discernable stress wave activity associated with the output (or slow) shaft (Figure 5).
The probable causes of the observed stress wave activities are:
1.

One tooth on the input shaft pinion is damaged (probably cracked).

2.
The one defective tooth on the drive pinion has damaged 11 teeth on the 88 gear wheel
and some impacting is occurring when the mating pinion engages any of these 11 teeth. The
combination of 16 teeth driving 88 teeth (prime number of 2 and 11) ensures that any specific
tooth on the pinion will engage only 11 of the 88 teeth on the gear wheel.

3.3

Summary of Results from Test Gear Box

From the results presented from an accelerometer and a tachometer used on the test gear box, it
can be concluded:
1.

The microprocessor generated tachometer presents a reliable pulse string from which
signal enhancement employing synchronous averaging can be successfully employed for
both the vibration component and the stress wave component of the signal obtained from
an accelerometer.

2.

The peak following circuitry employed for synchronous averaging of the stress wave
component functioned very well.

4.0 Conclusions
The use of a microprocessor to generate an output pulse string at a deterministic rate relative to
an input reference pulse string proved very useful to permit signal enhancement through
synchronization. Synchronous averaging was made possible for both the normal vibration
component as well as the stress wave component of signals acquired by using an accelerometer
as the sensor. The methodology employed here for processing the signal representing stress
waves was:
1

High pass the signal to remove the normal vibration component.

2.

Holding and output the peak value over a specified time range.

3.

Synchronously averaging this specially conditioned signal.

Copyright 1997, Computational Systems Incorporated. All rights reserved.

B-5

PeakVue Training
Appendix B

This monitoring of the peak value is capable of providing the similar information at a relatively
slow sampling rate that envelope detection would at a relatively high sampling rate. The use of
the relatively slow sampling rate makes possible the use of the typical data collector to collect
long duration data records with the assurance of capturing short lived impulse (stress wave)
events.
In addition to the ability to execute synchronous averaging of the vibration signal and the stress
wave signal to inaccessible shafts in a gearbox which was demonstrated in this paper; the
methodology presented here would be equally applicable to hard to detect bearing failures such
as inner race defects in the presence of high background vibration. The ability to synchronize on
specific fault event rates and the peak follower capability should prove very useful in monitoring
bearings in slow machinery, gear boxes, etc.
Figure 1.
Spectra and time trace from a 4 block average for an Fmax of 800 Hz and 800 line spectra.
Figure 2.
Twenty five block synchronous average of vibration signal relative to output (slow) shaft.
Figure 3.
Twenty five block synchronous average of vibration signal relative to intermediate shaft.
Figure 4.
Twenty five block synchronous average of vibration signal relative to input shaft.
Figure 5.
Twehty five block synchronous average of stress wave signal relative to output shaft.
Figure 6.
Twenty five block synchronous average of stress wave signal relative to intermediate shaft.
Figure 7.
Twenty five block synchronous average of stress wave signal relative input shaft.

B-6

Copyright 1997, Computational Systems Incorporated. All rights reserved.