Beruflich Dokumente
Kultur Dokumente
Melbourne, Australia
18 – 20 March 1998
CONTENTS
1.0 Abstract 3
2.0 Introduction 3
3.0 PeakVue™ 4
5.0 Summary 13
6.0 Acknowledgments 13
1.0 Abstract
This application paper describes a relatively new and not widely yet used vibration monitoring
technique called PeakVue™. The technique was developed by CSi and is available as a standard
feature on 2120 analysers. The paper will discuss the theory behind PeakVue™, its
implementation and case histories.
2.0 Introduction
Vibration analysis techniques have successfully been employed as a machine condition diagnostic
and monitoring tool now for over forty years. At first it was the time trace analysis. Swept filters
followed and with progress in electronics, spectrum analysis finally arrived. Roughly thirty years
ago another technique was developed – High Frequency Acceleration Demodulation. However, it
was only recently that this technique became widely available and is now a standard feature on
many analysers/data collectors.
With such a wide array of tools available and with growing experience the diagnosis of a machine
condition was becoming more accurate. Analysts became more confident and more successful.
With that came recognition from industry.
In the 90’s industry has gone through a difficult time. Ever increasing competition, globalisation
of the market and economy in crisis forced companies to look at their finances. It became apparent
that the operation cost had to be reduced. One of the ways to achieve it was to decrease
expenditure on maintenance. However, it had to be done without loss of equipment availability.
Hence, industries ever increasing interest in Predictive/Proactive Maintenance. In such an
environment the vibration analysts’ role has become more important and so has risen the
expectations of good, infallible results delivered by vibration monitoring. Given the above no one
vibration technique can be neglected and every new one should at least be trialled.
Over the past 9 years Siemens Ltd has been providing Condition Monitoring Services to a variety
of industries. Using standard data collection instruments and “off the shelf” software we have been
able to successfully implement vibration programs which included machines with operational
speeds between 50 and 6000 RPM. There was however a class of machines which was difficult to
monitor and the success rate was less than satisfactory. These included all machines running
below 50 RPM.
The use of the acceleration Demodulation did help, although the technique was still not reliable
enough and the fault detection rate was lower than expected. It was especially true with gearboxes
as the filter set up was very critical on this type of equipment making it difficult to use
Demodulation during routine monitoring. Another problem with Demodulation was the
trendability or rather lack of it. Simply put, trending was not reliable enough to be meaningful.
Since the development by CSi of PeakVue™, Siemens Ltd have been using it on a variety of
equipment completely eliminating Demodulation. The technique has proven to be a very reliable
tool for detection of antifriction bearings and gear problems on machinery running as slow as 20
RPM. As it will be shown in this paper the implementation and analysis process is very simple and
does not require any special software or hardware.
3.0 PeakVue™
The peak capture process is well suited for detecting high amplitude, short duration stress
waves. The stress waves are mainly produced by faults where metal-to-metal contact
occurs. Good examples of faults producing stress waves are antifriction bearing faults, gear
defects and surprisingly electrical faults on motors.
One may ask why PeakVue™ not Demodulation and what is the difference between
PeakVue™ and Demodulation. The answer is obvious. PeakVue™ is:
• Trendable as it captures the true amplitude
• Results do not depend much on filter settings
• Can detect bearing defects on very slow machinery
• Can detect defects with gears
For very slow speed machinery the Demodulation is not suitable, as it cannot detect signals
of a very short duration. We found that while monitoring gearboxes the Demodulation
process strongly depends on filter settings and can fail to detect obvious faults. This is not
so with PeakVue™.
Before we start talking about Analysis Parameters set up just a short reminder what
PeakVue™ detects – it detects stress waves associated with antifriction bearing faults and
gear defects. Knowing this helps set up trend parameters and Fmax for spectrum.
For routine data collection Siemens Ltd have designed standard Analysis Parameters Sets
as follows:
• Fmax = 50 orders
• Resolution = 400 lines
• Trendable parameters:
-overall level
-5-50 orders
-waveform peak
The high pass filters are set up at 500, 1000 or 2000 Hz depending on the application. For
trending purposes the overall level and the waveform peak parameters are of the most
value as they are the most meaningful and the fault limit levels are relatively easy to
establish.
Parameter 5-50 orders was set up at the time when we did not have much experience with
PeakVue™. It is of not much value although as we do not want to “upset” the trend we left
it unchanged.
For those accustom to spectrum analysis a PeakVue™ spectrum will look very familiar.
All the usual fault frequencies will be the same. Fig. 6 shows a typical PeakVue™
spectrum with a bearing fault present.
Things are a little different with time waveform. As it can be seen in Fig.7 the waveform is
truncated or one sided and is similar to Demodulated waveform.
0.6
0.4
0.2
0
Freq: 18.50
0 40 80 120 160 200
Ordr: 6.183
Frequency in Hz Spec: 1.289
18
RMS = 3.56
A cceleration in G-s
15 LOAD = 100.0
RPM = 180.
12 RPS = 2.99
9
PK(+) = 22.44
6 PK(-) = 2.65
CRESTF= 6.58
3
-3
-6
0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3.0 3.3
Revolution Number
R M S A c c i n G -s
R M S A c c i n G -s
0 20 40 60 80 100 0 5 10 15 20 25 30 35 40 45
A cc e l e ra ti o n i n G -s
A cc e l e ra ti o n i n G -s
12
2 WAVEFORM DISPLAY WAVEFORM DISPLAY
8
23-SEP-97 11:26 09-DEC-97 17:21
1
RMS = .5769 4 RMS = 1.78
0
PK(+) = 2.43 PK(+) = 9.74
-1
0 PK(-) = 3.30
PK(-) = 2.39
-2
CRESTF= 4.21 CRESTF= 5.48
-3 -4
Ordr: 4.907 Ordr: 4.902
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0 2 4 6 8 10
Freq: 121.46 Freq: 121.32
Revolution Number Revolution Number Spec: .469
Spec: .00554
As the motor is critical for the operation of the plant during our next vibration survey we
have collected also PeakVue™ data. It enabled us to correctly identify the fault. Fig. 9
contains PeakVue™ data. An inner race fault is clearly visible. The time waveform
provided additional information on severity of the fault.
Following our recommendation the customer replaced the motor during the Christmas
break. The motor was sent for overhaul and the bearing was inspected. A crack in the inner
race was found.
One could argue that the fault could have been detected using “conventional” methods,
however by applying PeakVue™, analysis was easier and more accurate.
R M S A c c i n G -s
10 1.6
Analyze Spectrum ROUTE SPECTRUM
27-FEB-97 08:42 1.2 C C C C C C C C C C 27-FEB-97 08:43
(PkVue-HP 1000 Hz)
RM S V e l oci ty i n m m /S e c
8 0.8
RMS = 11.47 OVRALL= 2.95 A-DG
LOAD = 100.0 0.4 RMS = 2.94
RPM = 100. LOAD = 100.0
6 0
RPS = 1.66 RPM = 180.
0 10 20 30 40 50 60 70 RPS = 2.99
Frequency in Order
A cce l e ra ti o n i n G -s
18 WAVEFORM DISPLAY
27-FEB-97 08:43
12
2 RMS = 3.41
6 PK(+) = 22.44
0 PK(-) = 2.65
CRESTF= 6.58
0 -6
Ordr: 3.014 Ordr: 6.183
0 10 20 30 40 50 60 70 0 1 2 3 4 5 6 7
Freq: 5.012 Freq: 18.50
Frequency in Order Spec: .298 Revolution Number Spec: 1.273
RM S A cce l e ra ti o n i n G -s
2.4 Plot
Scale
2.1 -- Baseline --
Value: .258 0.7
1.8
Date: 25-JUL-95
1.5 09-APR-97 11:21
1.2 0
FAULTY BEARING
0.9
0.6
0.3
NEW BEARING
0 27-FEB-97 08:45
Date: 23-OCT-97 Ordr: 6.230
0 200 400 600 800 1000 0 10 20 30 40 50 60 70
Time: 08:10:33 Freq: 18.65
Days: 25-JUL-95 To 23-OCT-97 Ampl: .139 Frequency in Order Sp 1: 1.281
“Normal” vibration data collected on this particular bearing did not detect any faults. As it
can be seen in Fig.14 there are no peaks in the spectrum, which would indicate a bearing
fault. Time waveform is considered to be normal for this particular application.
Things look different when it comes to PeakVue™ data (refer Fig.15). Spectrum contains
BPFO harmonics only so the diagnosis was simple. Waveform data provides further
information on fault severity.
HOIST#1 -G2 SHAFT #1 DRUM END HOIST#1 -G2D SHAFT #1 DRUM END DEMOD
R M S A c c i n G -s
1.2 ROUTE SPECTRUM 2.4 ROUTE SPECTRUM
1.0 E 28-OCT-97 12:28 1.8 C C C C C C C C 28-OCT-97 12:29
0.8 OVRALL= .9581 A-DG (PkVue-HP 2000 Hz)
0.6 1.2
RMS = 4.22 OVRALL= 3.50 A-DG
0.4
LOAD = 100.0 0.6 RMS = 3.46
0.2
RPM = 1004. LOAD = 100.0
0 0
RPS = 16.74 RPM = 738.
0 10 20 30 40 50 60 0 20 40 60 80 100 RPS = 12.30
A cc e l e ra ti o n i n G -s
3
WAVEFORM DISPLAY 18 WAVEFORM DISPLAY
1
28-OCT-97 12:28 28-OCT-97 12:29
12
-1 RMS = .9547 RMS = 3.37
PK(+) = 2.41 6 PK(+) = 21.03
-3 PK(-) = 3.72 PK(-) = 2.32
0
CRESTF= 3.90 CRESTF= 6.24
-5 -6
Ordr: 33.12 Ordr: 9.752
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Freq: 554.28 Freq: 120.00
Revolution Number Spec: .681 Revolution Number Spec: 2.016
In this particular case the customer was very surprised to hear that there was a severe outer
race fault, not a cage defect. The bearing was replaced during the next scheduled
maintenance window. As predicted severe spalling and scoring was present in the load
zone of the outer race.
This case study proves that even if there is no historical data it is still possible, using
PeakVue™ technology, to detect bearing faults and assess its severity.
This particular pulley was causing some problems in the past. Prediction of bearing faults
was difficult. As this is a conveyor drive drum, vibrations from the gearbox were
transmitted through the shaft to the bearing housing effectively masking any bearing faults,
which could have been present. The Demodulation technique employed did help, however
the results were still far from satisfactory.
As a result of such experience with this particular pulley Siemens Ltd has decided to try
PeakVue™. First we noticed a considerable improvement in data quality. Trends became
“trendable” and reliable. Fig.16 shows a vibration overall trend recorded on this particular
bearing. Note the change after a new bearing was installed. Fig.17 is a comparison spectra
of a new and a faulty bearing.
0.012
Plot
Scale
0.003
0.008
28-NOV-97 12:58
0
FAULTY BEARING
0.004
0
07-OCT-97 11:10
Date: 28-NOV-97 Ordr: 9.758
0 50 100 150 200 250 300 350 400 0 20 40 60 80 100
Time: 12:58:22 Freq: 5.399
Days: 02-DEC-96 To 28-NOV-97 Ampl: .00340 Frequency in Order Sp 1: .00394
020 - REJECT CONVEYOR RR1M ROM#1 020 - REJECT CONVEYOR RR1M ROM#1
RR1M -6RD RETURN (TAIL) PULLEY RHS BRG766 RR1M -6RD RETURN (TAIL) PULLEY RHS BRG766
0.20 0.0030
Trend Display ROUTE SPECTRUM
of J J J J J J 29-AUG-97 10:24
PEAK WAVEFORM (PkVue-HP 1000 Hz)
0.16 0.0024
R M S A cc e l e ra ti o n i n G -s
R M S A cc e l e ra ti o n i n G -s
OVRALL= .0068 A-DG
RMS = .0075
0.12 0.0018 LOAD = 100.0
RPM = 23.
RPS = .38
0.08 0.0012
>SKF 22213C
J=BPFO : 7.73
0.04 0.0006
0 0
Date: 15-JAN-98 Ordr: 7.730
0 100 200 300 400 500 0 10 20 30 40 50 60
Time: 17:08:26 Freq: 2.929
Days: 02-DEC-96 To 15-JAN-98 Ampl: .02307 Frequency in Order Spec: .00155
5.0 Summary
Introduced 2 years ago a new methodology, PeakVue™, of vibration signal processing has proven
to be an invaluable tool for prediction of antifriction bearing faults and gear problems. Its
application is simple and does not require special software or hardware.
A standard CSi 2120 machinery analyser and Master Trend™ software is all that is needed.
For medium and high speed machinery PeakVue™ is similar to Demodulation. For low speed it
gains an advantage as it can detect stress waves usually missed by Demodulation.
PeakVue™ data is trendable and as such is more suitable for routine condition monitoring.
The time waveform provides true amplitude making assessment of fault severity possible.
6.0 Acknowledgments
I like to express my gratitude to all Rockhampton CM Team Members whose sweat paved the
road during collection of the vibration data used in this paper.