Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Chapter 16

SIGNATURE ANALYSIS
Most failures of rotating and reciprocating machinery exhibit characteristic vibration
profiles that are associated with specific failure modes. This phenomenon is due to the
forcing function, caused by a developing defect, having a unique characteristic signature. None of the filtered bandwidth monitoring methods provides the means to detect
and evaluate these unique profiles. Signature analysis provides this capability and its
use is required in a comprehensive predictive maintenance program.

CHARACTERISTIC
VIBRATIONSIGNATURES
A vibration signature provides a clear, accurate snapshot of the unique frequency
components generated by, or acting on, a machine-train. Such a signature is obtained
by converting time-domain data into its unique frequency components using a fast
Fourier transform (FFT). Such a vibration signature, referred to as frequency-domain
data, is used in signature analysis to evaluate the dynamics of the machine.
Frequency-domain vibration signatures form the basis for any predictive maintenance
program designed to detect, isolate, and verify incipient problems within a machinetrain. These signatures are the basic tools used for in-depth analysis methods such as failure-mode, root-cause, and operating dynamics analyses. Operating dynamics analysism,
which is beyond the scope of this module, uses vibration data and other process
parameters, such as flow rate, pressure, and temperature, to determine the actual operating condition of critical plant systems.

TYPESOF SIGNATURE ANALYSIS


In general, new or immature predictive maintenance programs are limited to comparative analysis or waterfall trending. Although these comparative techniques provide

181

182

Vibration Fundamentals

the ability to detect severe problems, they cannot be used to isolate and identify the
forcing functions or failure modes. These methods also are limited in their ability to
provide early detection of incipient problems.
As the predictive maintenance program matures, root-cause analysis and operating
dynamics analysisTMmethods can be used. With the addition of these more advanced
diagnostic tools, vibration signatures become an even more valuable process performance improvement tool.
Automatic Trending Analysis

A predictive maintenance program utilizing a microprocessor-based vibration analyzer and a properly configured database automatically trends vibration data on each
machine-train. In addition, it compares the data to established baselines and generates
trend, time-to-failure, and alert/alm status reports.
The use of just these standard capabilities greatly reduces unscheduled failures. However, these automated functions do not identify the root causes behind premature
machine-train component failures. In most cases, more in-depth analysis allows the
predictive analyst to identify the reason for pending failure and to recommend corrective actions to prevent a recurrence of the problem. Again, the specific microprocessor-based system used determines how much manual effort is required for more indepth analysis.
More In-Depth Trending Analysis

More in-depth analysis is called for when the automatic trending analysis described in
the previous section indicates that a machine-train is exhibiting excessive vibration.
Obviously, machine-trains that are operating within acceptable boundaries do not
require further investigation. Care should be taken, however, to ensure that the automated functions of the predictive maintenance system report abnormal growth trends
as well as machine-trains that are actually in alarm.
Comparative Analysis (Waterfall Trending)

FFT signatures that are collected on a regular schedule provide a means of trending
that can help the analyst identify changes in machine condition. Changes in the operating parameters, such as load, will directly affect the signatures generated by a
machine.
Unlike trending analysis, which is based on broadband and narrowband data, comparative analysis is a visual comparison of the relative change of the machine-trains full
vibration signature and its discrete frequency components over a period of time.
Because vibration signatures are acquired at regular intervals in a predictive maintenance program, this form of trending is very effective in identifying changes in
machine condition.

Signature Analysis
STATION:

Process Plant

MACHINE:

MPT:

Centrifugal Pump

183
V03

.13

.12
y

.ll

w
6

.10

.09

2
z
W

2
k

.08

.07

.05
.04

.03

3
Q

.02

05
05
05
05
05

.06

os
05
05
05
05
05
05
05
05
05

.01

10

20

40

80

160

320

Mor
Feb
Jon
Dac
NO
oct
sap

Aug
Jul

Jun

Yoy
Qpr

Yo,

Feb
Jon

640

FREQUENCY IN CPM
Figure 16.1 Illustration of a water$allplot.

Displaying the signatures in a waterfall or multiple-spectra display (sequentially by


data-acquisition time) allows the analyst to easily see the relationship of each frequency component generated by the machine (see Figure 16.1). Any significant
change in the amplitude of any discrete frequency is clearly evident in this type of display, which is used in many of the figures in subsequent sections.
Although comparative analysis can be used to help the analyst identify specific
changes that are generated by process changes, each signature must be normalized for
process variations. Therefore, as part of the acquired data set, the analyst must record
the specific process conditions for each data set. With this information and the waterfall display of vibration signatures, the analyst can quantify the changes that result
from variations of these parameters.
Developing problems within a machine-train can be identified by comparing the FFT
signature to the following: (1) a baseline or reference signature, (2) previous signatures, or (3) industrial standards. This method determines if a potential problem exists
and can be used to isolate within the machine-train the probable source of developing
problems.

184

Vibration Fundamentals
MULTIPLE SPECTRA DISPLAY
STATION:

Process Plant

MACHINE:

A-C

Pump

MEASUREMENT POINT:

V03

0.50

v,

June 15, 1997


11:3301 P.M.

z
z

h
May 1. 1997

12:OO:Ol P.M.

50

100

150

200

250

300

350

400

450

500

FREQUENCY IN HERTZ

Figure 16.2 Comparison to baseline reference.

Baseline or Reference Signatures


A series of baseline or reference data sets should be taken for each machine-train

included in a predictive maintenance program (Figure 16.2). These data sets are necessary to compare with trends, time traces, and FFT signatures that are collected over
time. Therefore, baseline data sets must be representative of the normal operating
condition of each machine-train in order to have value as a reference.
In integrated process plants where most machines are subject to variable operating
conditions, this exercise requires more than one reference data set for each machinetrain. To be of benefit, a series of baselines must be acquired from each machine-train,
each of which should accurately represent a specific operating variable (Le., product,
machine setup, load, etc.). It is important that all data sets (whether baseline data or
current operating data) be clearly identified in order to be useful. Current operations
data must be compared to a reference data set having the same operating conditions
(Figure 16.2).
Note that baseline references must be updated each time a machine-train is overhauled, replaced, or when a new process setup is established. A current set of valid
reference data is essential when performing comparative analysis.

SignatureAnalysis

185

Table 16.1 Yibration Severity Standards*


Machine Classes (IPS-PK)

Condition

Good operating condition


Alert limit
Alarm limit
Absolute-fault limit

I
0.028
0.010

0.156
0.260

I1
0.042
0.156
0.396
0.400

I11

0.100
0.255
0.396
0.620

IV
0.156
0.396

0.622
1.ooo

* Applicable to a machine with running speed between 600 to 12,000 rpm. Narrowband setting: 0.3X to
3.0X running speed.
Machine Class Descriptions:
Class I
Small machine-trains or individual components integrally connected with the

Class I1
Class I11
Class IV

complete machine in its normal operating condition (i.e., drivers up to 20 hp).


Medium-sized machines (i.e., 20- to 100-hp drivers and 400-hp drivers on specia1 foundations).
Large prime movers (i.e., drivers greater than 100 hp) mounted on heavy, rigid
foundations.
Large prime movers (Le., drivers greater than 100 hp) mounted on relatively
soft, lightweight structures.

Source: Derived by Integrated Systems, Inc., from I S 0 Standard 2372.

Nonbaseline Signatures
Visual comparison of two signatures can enable the analyst to determine if a problem
is developing. As with the case of filtered energy data, all signatures must be normalized for process variables such as speed, load, etc., in order for comparisons to be
valid. Direct comparison is useful only when both data sets reflect the same operating
conditions or parameters.
Common-shaft analysis is used to identify the strongest vibration by visually comparing the signatures of all measurement points on a common shaft. It is a useful technique for isolating the source of abnormal vibrations. Although this method does not
absolutely identify the problem, it reduces the number of machine components that
must be inspected or evaluated to correct the problem.

Industriul Standards
One form of comparative analysis is direct comparison of the acquired data to industrial standards or reference values. The vibration severity standards presented in Table
16.1 were established by the International Standards Organization (ISO). These data
are applicable for comparison with filtered narrowband data taken from machinetrains with true running speeds between 600 and 12,000 rpm. The values from the
table include all vibration energy between a lower limit of 0 . 3 ~true running speed

186

Vibration Fundamentals

and an upper limit of 3 . 0 ~For


. example, an 1800-rpm machine would have a filtered
narrowband between 540 (1800 x 0.3) and 5400 rpm (1800 x 3.0). A 3600-rpm
machine would have a filtered narrowband between 1080 (3600 x 0.3) and 10,800
rpm (3600 x 3.0).

Microprocessor Comparisons
Many of the microprocessor-based predictive maintenance systems also allow direct
comparisons of the relative strengths of each frequency component. Such microprocessor comparisons do not require knowledge of the machine-train or vibration analysis techniques, but both data sets must be acquired under the same operating
conditions. Increases in relative strength indicate more vibration and a developing
problem in the machine-train.
Cross-machine comparison is an extremely beneficial tool to the novice analyst. Most
vibration monitoring systems permit direct comparison of vibration data, both filtered
window energy and complete signatures, acquired from two machines. This capability
permits the analyst to directly compare a machine that is known to be in good operating condition with one that is perceived to have a problem. There are several ways that
cross-machine comparisons can be made using microprocessor-based systems: multiple plots, ratio, and difference.

Multiple Plots
Two or more signatures can be shown on a single display. This method permits the
analyst to directly compare the actual signatures generated at each measurement
point on both the suspect and a reference machine-train. This multiple-signature display permits direct comparison of each frequency component within the signatures
(Figure 16.3).

Ratio Analysis
With this technique, the signature from the suspect machine is divided by the signature of the reference machine, frequency by frequency. The resultant display shows
the relative amplitude, both positive and negative, of each frequency component in the
suspect machine-train (Figure 16.4). As an example, the display may indicate that the
gear-mesh energy in the suspect machine is 40% higher than that in the reference
machine (i.e., ratio = 1.4). With this information, the analyst can isolate specific
machine components that are potential problems.
DifferenceAnalysis
With the difference analysis technique, the signature of the reference machine is subtracted from that of the suspect machine, frequency by frequency. The resultant plot
displays the difference value, positive and negative, of each frequency component
within the two (see Figure 16.5).

187

Signature Analysis
STATION:

Process Plant

MACHINE:

A-C

Pump

MEASUREMENT POINT

V03

04-05-97
15:30:43

03-05-97
1043:59

02-05-97
13:40:03

01 -05-97
1 1 :03:30
0

50

100

150

200

250

300

350

400

450

500

FREQUENCY IN HERTZ

Figure 16.3 Multiple-signature displuy.

COMPARISON SPECTRA
STATION:

Process Plant

MACHINE:

A-C

Pump

MEASUREMENT POINT

V03

lo

Vl

z
3

0
Z
I

June 15, 1997


11:33:01 P.M.

0
I<
-

DIVIDE0 BY:

LT

May 1 , 1997

12:OO:Ol P.M.

50

100

150

200

250

300

350

FREQUENCY IN HERTZ

Figure 16.4 Ratio of two signatures.

400

450

500

188

Vibration Fundamentals

DIFFERENCE
OF
SPECTRUM

Figure 16.5 Difference of two signatures.

You might also like