CHAPTER SEVEN

Guidelines for the implementation

of a predictive maintenance
program

Introduction
Conditioning monitoring systems evolved from predictive mainte-
nance programs. The description of these programs will illustrate the
definition of CMS and will provide the basis for their interpretation. This
chapter remarks on the effects of a predictive maintenance program on pro-
ductivity and product quality. It describes the steps to be taken to implement
a predictive maintenance program. Maintenance has evolved through time;
at the beginning, equipment malfunctions were repaired only at the moment
they showed up (corrective maintenance). Afterward, the industry devel-
oped preventive maintenance systems that consist of repair actions based
on statistical studies of the frequency of faults as well as recommendations
from the manufacturer. These maintenance systems became expensive
because only some of the repairs were needed. Then predictive maintenance
systems were developed. They are based on periodic monitoring of the
operating conditions of the machines and the application of techniques to
estimate the expected life remaining in each component. These systems
allow the scheduling of repairs and prevent unexpected failures.
Currently, it is convenient that all production systems based on continuous
improvement techniques include, as a fundamental part of their scheme, pre-
dictive maintenance because it allows the total operation of the plant to be
adjusted dynamically based on the actual operating conditions.
Predictive maintenance is a system that not only benefits the mainte-
nance area, but also influences the improvement of productivity, the quality
of the products, and the profitability of the investment. Therefore, it should
not be taken only as a monitoring system but also as an integral part of the
production strategy in a plant.

Mechanical Vibrations and Condition Monitoring © 2020 Elsevier Inc. 133

https://doi.org/10.1016/B978-0-12-819796-7.00007-X All rights reserved.
134 Mechanical Vibrations and Condition Monitoring

Predictive maintenance consists of periodically monitoring (sample) the

operating conditions of the machinery as well as analyzing the behavioral
tendencies of each component. With this information, it is possible to esti-
mate the time in which failures will occur. In this way, maximization of the
repair intervals is achieved as well as minimizing the operating costs associ-
ated with premature or catastrophic failures and emergency stoppages of the
machinery. To emphasize the concept, in this chapter a machine train is
defined as three or more machine components that are coupled together,
acting as one machine. The predictive maintenance systems are based on
the different nondestructive techniques of analysis, among which the fol-
lowing stand out:
- Vibration monitoring: This technique consists of measuring the amplitude
and vibration frequency of the critical components that form the machine
train. This method was discussed extensively in the previous chapters.
- Monitoring of the parameters of the process: This is the main element of the
predictive maintenance system. It consists of periodically recording the
real values of all the parameters of the machine train (consumed power
in electric motors, air pressure in pneumatic systems, hydraulic pressure,
steam pressure, temperature, etc.). The purpose of this method is to relate
variations in the measured variables with the probability of failure.
The selection of the variables depends on the structure of the machine train
and the type of individual components.
- Thermography: This method is based on measuring the temperature at dif-
ferent points of the machine train. This measuring provides related infor-
mation to the mechanical behavior of each component because when a fail
starts in any element of the machine train, usually the temperature
increases. In this way, the changes in temperature can be associated with
a machinery component with the possibility that it fails.
- Oil analysis: This technique is used as a complementary tool that analyzes
the origin of the substances that contaminate the lubricants. The analysis
helps to infer which of the components of the machine train is suffering
premature wear or a more accelerated deterioration.
Among these techniques, vibration monitoring is the one that covers the most
significant number of failure causes. Each of them occurs at a particular vibra-
tion frequency (as was seen in previous chapters) and their amplitude is a reflec-
tion of the machinery dynamic conditions. Besides, this amplitude in general
increases according to the failure. The estimation of the time-to-failure stage
defines the maintenance program. This estimation permits planning the entire
production, meaning that the supply of spare parts can be done “just in time”
and the maintenance activities can be determined ahead of time.
Guidelines for the implementation of a predictive maintenance program 135

Predictive maintenance does not substitute for traditional maintenance

systems; it is merely a tool that produces updated information allowing
the evaluation of the future behavior of the equipment. The tendencies
of all critical equipment can be graphed and estimated with a perfect approx-
imation of the moment in which failure will occur. With vibration analysis,
it is ensured that the repair intervals of the machinery will be extended and
that the machine train offers the maximum availability of use. The benefits of
predictive maintenance systems can be summarized as:
- Lower maintenance costs
- Lower number of failures
- Lower repair times
- Inventory reduction
- Increase the machinery lifespan
- increase in plant productivity
- Improvement of operational safety
- Verifications of operating conditions of new equipment
- Verification of repairs
- The possibility to implement “just in time” programs
Predictive maintenance programs start with the selection of the equipment
to be monitored, the selection of the monitoring route, and the definition of
the measuring intervals. Both the narrow bands and the alarm levels to each
measuring point within the monitoring route are defined so that the ten-
dency graphs are finally built, allowing the predictive dynamic behavior
of the equipment.

Definition of the monitoring route

For the implementation of the predictive maintenance program, all
fundamental machines for the production process of the company must
be selected. In this way, a database with the most relevant information
and the operation efficiency is built.
In each machine, both the measuring points and the monitoring periods
are identified. For each monitoring point, the total vibration spectrum is
recorded (full firms), and the zooms around the characteristic frequencies
are analyzed, in this way, it is possible to establish the starting point of
the predictive maintenance program.
Subsequently, the monitoring intervals must be established to build the
tendency curves and in this way schedule the repairs. When there are no
initial reference points, it is therefore recommended to monitor each
selected point at intervals of between 8 and 15 days. Once the data for
136 Mechanical Vibrations and Condition Monitoring

the first five points are established, the most suitable inspection intervals can
be determined with the objective of “measuring the vibrations at lower time
intervals than the intervals in which variations are detected in the measured
signal in normal operating conditions.” This procedure minimizes the pos-
sibility of a catastrophic failure.
Fig. 7.1 presents the curve of failure frequencies of pieces of equipment
from their installation until removal from production. This graph is known
as the curve of life because it resembles human life. It defines three phases:
youth, maturity, and old age.
- Youth
- Productive life
- Elder
In the youth phase, the failures are greater due to their initial adjustments,
premature wear, settlements, or design changes. The mature stage is consid-
ered to be the productive lifespan of the equipment. In the old age stage, the
failures increase due to the deterioration of all the systems.
The objective of predictive maintenance is to get the machine train to
operate as long as possible in its production life stage at the lowest possible
cost. In this way, it would be ideal that the machine be substituted out due
to a significant change in technology and not because it has reached its
old age.
For the correct identification of the measuring points, it is essential to
know the design of the machine train as well as the functioning details of
the components, such as the rotation speeds of each shaft that form the gear,
the type of supports (ball bearings, bearings), the type of power equipment
(turbine, electrical motor, internal combustion motor, etc.), and their oper-
ational parameters.

%Failure

Youth Elder
Productive life
Life cycle

Fig. 7.1 Life curve of a machine train.

Guidelines for the implementation of a predictive maintenance program 137

Measuring points

2
3

Fig. 7.2 Components of a machine train and the location of monitoring points.

Fig. 7.2 is a schematic representation of a typical machine train. The

numbering of the points where the vibratory firms are measured every sam-
pling time is indicated. It is recommended to carry out the numbering
sequence from the end support of the motor and culminate with the end
support of the output shaft. In this way, there is an ordered sequence of
points and information can be handled with more detail. In each point,
the direction of the signal must be identified. For example, in Fig. 7.2, point
1 would allow identifying the level of vibration of the train in three direc-
tions: horizontal, vertical, and radial. Once the route is established, at each
monitoring point the narrow bands are defined.

Narrow band selection

Once the monitoring points in the gear train are located, the most out-
standing characteristic frequencies of the vibration sources are identified.
Around each of these frequencies, it is required to select the narrow bands
to monitor before setting both the alert and alarm limits. These bands must
contain the characteristic frequency as well as covering variations in the
operating speed that are generally between 10% and 20% of the characteristic
frequency when the speed variations of the equipment are small.
As a general guide, in Table 7.1 the frequencies are indicated where more
information from the vibratory signal can be obtained, depending on the
type of failure under study.
138 Mechanical Vibrations and Condition Monitoring

Table 7.1 Frequencies and the sideband limits for different failure modes.
Type of failure Characteristic frequency (×)
Imbalance Rotation frequency of each of the shafts with a bandwidth of
10% in the radial direction
Misalignment Rotation frequency and its first harmonic with a bandwidth
of 10% in both the radial and axial directions
Bent shaft Rotation frequency and its first harmonic with a bandwidth
of 10% in both the radial and axial directions
Eccentric shaft Rotation frequency and its first harmonic with a bandwidth
of 10% in both the radial and axial directions
Mechanical Rotation frequency of each shaft and to 0.5
, 1.5
, and first
clearance harmonic (2
) in the radial direction
Journal bearings Rotation frequency and its first harmonic with a bandwidth
of 10% in the radial directions
In addition to a narrow band between 0.42
and 0.5
in the
radial direction
Belts Rotation frequency of each of the shafts with a bandwidth of
10% in the radial direction
Pass frequency of the band and its first harmonic with a
bandwidth of 10% in the radial direction
Electrical defects Rotation frequency with a bandwidth of 10% in the radial
direction
Pole pass frequency
Rotor bar pass frequency including at least four sidebands
per side
Line frequency including at least four sidebands of the sliding
frequency and its first harmonic
Hydraulic and Rotation frequency with a bandwidth of 10% in the radial
aerodynamic direction
Blade pass frequency and at least four sidebands
Turbulence (from 50 to 2000 cycles per minute,
randomly)
Cavitation (In the case of the pumps, a signal appears
between 0 and 1
, randomly.)
Roller bearings Inner race pass frequency with a bandwidth of 20%
Outer race pass frequency with a bandwidth of 20%
Ball pass frequency with a bandwidth of 20%
Rolling-element pass frequency with a bandwidth of 20%
Gears Rotation frequency of each of the shafts with a bandwidth of
10% in the radial direction
Gear frequency including at least four sidebands and the first
harmonic of the gear frequency
Guidelines for the implementation of a predictive maintenance program 139

0.025

3xo
79.333

0.02

Amplitude [g]
0.015
1xi
BCF
0.01
85.336
2xi
1xo
BBFO
BBFI
0.005

0
0 30 60 90 120 150
Frequency [Hz]
0.0004
78.189
0.00035

0.0003 82.401
81.373
81.038
Amplitude [g]

0.00025

0.0002

0.00015 GMF
Alarm 0.0001

5e-05
safe
0
0 120 240 360 480 600
Frequency [Hz]

Fig. 7.3 Example of a frequency spectrum with the location of the narrow bands.

Once the narrow bands are determined, the predictive maintenance

program starts, analyzing the tendencies of the dynamic behavior of the
selected gear train. In Fig. 7.3, a spectrum with its narrow bands is shown.

Analysis of tendencies
Once the narrow bands in each monitoring point of the gear train
under study are established, the permissible limits of the vibration levels
are defined. The selection of these limits depends on the behavior of each
machine, its importance in the production system, and the experience of
the maintenance staff. As a general rule, it is recommended to establish
the alert limit of measurements to twice the average value of the vibration
amplitude in normal operating conditions, and the alarm limit to five times
this latter value.
In general, the behavior of the vibration follows an exponential law.
Therefore, when the vibration level reaches the alert level, the tendency
140 Mechanical Vibrations and Condition Monitoring

curve can be traced and a date in which the vibration would reach the alarm
limit can be objectively determined. During this period, repair or removing
the machine from production must be done (Fig. 7.4).
The periodic monitoring of the vibration allows the stage of the life
curve in which the gear train is located to be identified. The problem of ten-
dency analysis is that it does not allow establishing the causes that produce
the deterioration of the equipment. For that reason, it is necessary to com-
plement the analysis with other information such as the reports of the oper-
ating conditions, the thermal analysis, the oil analysis and knowledge of the
equipment design. This type of analysis complements the understanding of
the equipment behavior, making the predictive maintenance more efficient.
Because the vibration in a machine does not keep the same levels of
amplitude through its lifespan, it is necessary to provide a statistical followup
to the behavior of the frequency under study to properly establish both the
alert and alarm limits. Among the causes that produce the variation of the
vibration amplitude, the changes in the operation of the machinery, climate
changes, and the deterioration of the equipment itself stand out.
The analysis of the tendencies of the dynamic behavior of the machinery
is the basis of the predictive maintenance system. To best illustrate its appli-
cation, it is necessary to consider the subsequent example.
Vibration amplitude samples
5

4
Acceleration [g]

Histogram
1

0
0 20 40 60 80 100
Time [days]
Fig. 7.4 Analysis of tendencies and histogram.
Guidelines for the implementation of a predictive maintenance program 141

Example
In a chemical plant, there are two turbo-generators identical to the
one shown in Fig. 7.5. The current importance of such equipment requires
including them in the productive process. They were included in the pre-
dictive maintenance program of the entire installation. The turbo-generator
can be defined as a machine train consisting of:
- A two-stage steam turbine with 135 and 150 blades, respectively, with a
centrifugal governor for speed control.
- A gear reducer of a single-stage double-helical type with 21 teeth in the
pinion and 97 in the gear, lubricated with a gear pump of eight teeth con-
nected to the pinion shaft mounted on hydrodynamic ball bearings; the
gear is mounted on bearings of 12 balls with diameters of ball pass
D ¼ 235 mm and d ¼ 40 mm for each ball.
- Electric generator of six poles with a synchronous speed of 1200 rpm. The
rotor is mounted on journal bearings.
Once the design is identified, the characteristics of all the components and
the monitoring route are defined.
In this example, the dominant frequencies that each element generates
are determined next.

Monitoring route
For the example, four monitoring points were selected. The first was at the
free end of the turbine and the latter at the free end of the generator.

2
3 4
1
6

Turbine 8
Gearbox

7 Generator

Fig. 7.5 Schematic representation of a turbo-generator.

142 Mechanical Vibrations and Condition Monitoring

In each measuring point, the frequency ranges were determined in

which spectra were recorded as well as the frequencies of interest. For this,
the frequency values that characterize each component of the turbo-
generator were calculated. In this case, the rotation frequency of the turbine
shaft (point 1) was calculated, depending on the gear reduction relation, in
such a way that:
 
nc
ωT ¼ ωG
np
97
¼ ð1200Þ ¼ 5542:8 cpm
21
The blade pass frequency was calculated as:
FP A1 ¼ NA1 ωT
¼ ð135Þ5542:8
¼ 748,285:7 cpm
And for the second pass:
FP A2 ¼ NA2 ωT
¼ ð150Þ5542:8
¼ 831,428:5 cpm
For point 2, the rotation frequency of the pinion is the same as that of the
turbine, so in these points, the frequencies of interest were those of the gear,
the one of the teeth pass of the lubrication pump, and those associated with
the problems of the hydrodynamic ball bearing.
The gear frequency is:
FE ¼ np ωT
¼ ð21Þ5542:8
¼ 116,398:8 cpm
For the lubrication pump:
FEB ¼ nB ωT
¼ ð8Þ5542:8
¼ 44, 342:4 cpm
And for the ball bearings, a range between 0.35 and 0.5 was estimated, and
Fch was:
Fch ¼ 1940 to 2770 cpm
Guidelines for the implementation of a predictive maintenance program 143

In point 4, the vibration was measured at the rotation frequency of the gear
(8w ¼ 1200 rpm) and to the frequencies that characterize the bearings:



FRB ¼ 1=2 1  r 2 x


FRP ¼ 1=2 ð1  r Þx
OBPE ¼ ðn=2 Þð1  r Þx
OBPI ¼ ðn=2 Þð1 + r Þx
With the data of the ball bearings, the fundamental frequency of the bearing
(fundamental train frequency) was determined, having:
FRP ¼ 497:87 cpm
The ball spin frequency is:
FRB ¼ 3422:8 cpm
The frequency due to a fault in the outer race is:
OBPE ¼ 59744:4 cpm
And the frequency due to a failure in the inner race is:
OBPI ¼ 8425:53 cpm
In the measuring points that correspond to the electrical generator, both the
power line frequency and its harmonics (in this case, there is no slip fre-
quency) were specified, in addition to the rotation frequency.
FL ¼ 3600 cpm
2FL ¼ 7200 cpm
Once the characteristic frequencies and the narrow bands for each of the
selected measuring points was identified, the narrow bands were selected
based on the previous information in this chapter, rounding their values
according to the resolution of the used vibration analyzer equipment
(Table 7.2).

Analysis of tendencies
Once both the monitoring route and narrow bands were established, it will
proceed to the capture of the frequency spectra. At the beginning of the
program, it is required to monitor the system under study with higher fre-
quency. This frequency depends on the type and operating condition of the
gear train. However, as an average value, it is estimated that by carrying out
144 Mechanical Vibrations and Condition Monitoring

Table 7.2 Characteristic frequencies and narrow bands of the measuring points
selected.
Reference frequency
Measuring point Direction (cpm) Side bands
R01 Radial 5500 5000–5750
11,100 10,000–12,200
Z01 Zoom (radial) 750,000 725,000–770,000
R02 Radial 5500 5000–5750
11,100 10,000–12,200
Z02 Zoom (axial) 830,000 810,000–855,000
R03 Radial 2300 1940–2780
5550 5000–5750
11,000 10,000–12,000
A03 Axial 5500 5000–5750
11,100 10,000–12,200
Z03 Zoom 116,000 93,800–138,000
R04 Radial 2300 1940–2780
5500 5000–5750
11,000 10,000–12,000
A04 Axial 5500 5000–5750
11,100 10,000–12,200
R05 Radial 500 400–600
1200 1100–1320
2400 2200–2600
3400 3000–3500
6000 5750–6600
A05 Axial 1200 1100–1320
2400 2200–2600
R06 Radial 500 400–600
1200 1100–1320
2400 2200–2600
3400 3000–3500
6000 5750–6600
A06 Axial 1200 1100–1320
2400 2200–2600
R07 Radial 1200 1100–1320
2400 2200–2600
3600 3550–3650
7200 7150–7250
A07 Axial 1200 1100–1320
2400 2200–2600
R08 Radial 1200 1100–1320
2400 2200–2600
3600 3550–3650
7200 7150–7250
Guidelines for the implementation of a predictive maintenance program 145

measurements every 15 days, a pattern of behavior of the mechanical rotat-

ing equipment can be established. With the information collected, the ana-
lyst graphs the tendency curves for each point and adjusts the reading
intervals according to the equipment performance.
Even though a large amount of data is handled, modern computer sys-
tems allow its analysis in real time, which facilitates its handling, allowing the
maintenance engineer to define both the inspection intervals and the repair
of the machinery with objective criteria.