2020

edition
E-BOOK

The condition monitoring

comparison guide

Should you use vibration, oil, heat, noise, or current and voltage to
monitor the health of your industrial assets? Find out which technologies
work best for your situation.
Contents

A brief history of condition monitoring techniques 03

Vibration analysis 04
Oil analysis 04
Infrared thermography 04
Acoustic emission analysis 04
Motor current signature analysis 04

Condition monitoring, turbocharged 05

Who's good—and bad—at what 07

Sensor technology 07
Sensor location 09
Fault detection 10
Energy and performance insights 12
Summary table 13

Conclusion 14

Contact 14

E-book | The condition monitoring comparison guide

A brief history of condition monitoring techniques

In a sense, condition monitoring systems have That quest kicked into high gear in the early
been around as long as there have been 20th century, when enterprising engineers
machines—if you count “shut it off! It’s developed ever more sophisticated ways to
smoking!” among them. But we wouldn’t be measure equipment health. Here, we
human if we didn’t try to perfect the process catalogue the most widely used techniques
of catching machine degradation before it that have been developed to monitor the
turns into full-blown failure. health of industrial assets.

Figure 1. Timeline of technologies to monitor the condition of industrial equipment.

E-book | The condition monitoring comparison guide

03
 A brief history of condition
monitoring techniques

Vibration analysis (VA) is generally considered the first technology that could
VA monitor machine health to provide advance warning of failure. The first research
article to explore vibration as a way to monitor industrial assets dates to 1938.
Sensors are installed directly on the component to be monitored, and can warn of
upcoming failure if they detect vibration patterns that fall outside the range
exhibited by a normally functioning piece of equipment.

Oil analysis (OA) was first used in 1946, when the US railroad industry analyzed
OA diesel engine lubricant to detect component wear and tear. Spent oil was shipped
to researchers who used a spectrograph to detect individual chemical elements
such as iron and copper. The technique began expanding to other industries in the
late 1950s, as handheld spectrometers were developed that could analyze
samples on the spot.

Infrared thermography (IR) was initially developed for military use in the late
IR 1950s, and today’s thermal imaging cameras are based on that technology, which
was opened to civilian use in the 1990s. Infrared imagers map the heat emitted
by an object; changes from the usual pattern can signal a developing problem.

Acoustic emission analysis (AE) emerged in 1970. AE makes use of sensors

placed on the asset (called structure-borne analysis) or very near it (airborne
AE analysis). These sensors detect transient elastic waves generated by the
processes associated with wear and tear, such as friction, crushing and cracking.
Though these processes may produce audible sounds, the AE method usually
measures frequencies above the range of normal human hearing (a.k.a.
ultrasonic).

Motor current signature analysis (MCSA) also dates to the 1970s, and was first
MCSA suggested as a way to monitor nuclear power plant equipment that couldn’t be
reached by other methods. MCSA is based on the observation that the current
drawn by a motor contains information about the state of the machine the motor
is driving. Sensors are placed at a point between the power supply and the motor,
rather than on the asset being monitored. MCSA + voltage adds voltage
measurements to the basic technique, which enhances the system’s sensitivity
(by enabling it to correctly identify power-supply fluctuations) and makes it
possible to analyze energy consumption and process efficiency in addition to
asset health.

E-book | The condition monitoring comparison guide

04
Condition monitoring, turbocharged

Before the widespread adoption of computers 21st century has provided arguably the best
in the 1980s, all these techniques required reason to step up those efforts: the marriage
not only manual data collection, but also between artificial intelligence (AI) and the
skilled human experts to analyze and interpret industrial internet of things (IIoT). We now
that data. Manual condition monitoring have the ability to apply automated, expert
systems still dominate the field; many real-time analysis to vast amounts of
industries have yet to complete the complex continuous data collected by wireless
move from offline to online systems. But the sensors.

Figure 2. The rise of machine learning and wireless connected sensors makes it possible to expertly analyze vast
amounts of asset health data in real time.

E-book | The condition monitoring comparison guide

05
 Condition monitoring,
turbocharged

That means tried-and-true condition amounts of data (far more than a human can,
monitoring techniques can now be used to and far faster), they’re constantly refining
collect asset health datapoints not just once a their knowledge based on new data, and their
month or every six weeks, but thousands of conclusions are always based on fact, never
times per second, all day every day. On the on intuition. All this means that AI-based
back end, machine learning software can now condition monitoring systems also get even
process that data in real time to automatically better at what they do over time.
detect developing faults and even pinpoint  
where they’re arising: bearings, drive shaft, It goes without saying that 21st century
coupling, impeller, pulley, belt, stator, rotor— companies need digital condition monitoring
the list goes on. based on AI and the IIoT. But that still leaves
  the underlying technology—vibration, oil, heat,
AI-based analysis has significant advantages noise, or current and voltage—to choose. So
over human expertise: machine-learning let’s look next at where each type of system
systems can rapidly process immense stumbles and soars.

E-book | The condition monitoring comparison guide

06
Who's good—and bad—at what

Sensor technology
Condition monitoring requires data, which means whatever system you choose will install some kind
of sensor. These tiny data-capture gizmos are a subcategory of transducers: devices that convert one
form of energy to another.

Vibration sensors come in different forms, but the most widely used are
VA piezoelectric accelerometers: devices that convert the mechanical force caused by
a change in motion into a proportional electrical charge. (Piezo comes from the
Greek word meaning “to squeeze.”) Commonly used acoustic structure-
borne sensors are also piezoelectric accelerometers; after all, sound is a type of
vibration. The difference is in the vibrational frequency and strength (or
AE amplitude) the sensor can measure. In general, AE sensors will measure smaller
vibrations at higher frequencies than ordinary VA sensors.   
 
Piezoelectric devices are most sensitive in a frequency range determined by the
properties of the material used (commonly quartz or a synthetic ceramic). Their
location and orientation on the machine also affect what they can detect, and how
well. There’s also a tradeoff between the amplitude an accelerometer can
measure and its sensitivity. Accelerometers can be thrown off by noise or
vibrations from the surrounding environment. These four points make it important
to carefully assess the characteristics of both the asset and its surrounding
environment in selecting and installing VA or AE sensors.

Acoustic airborne sensors are essentially microphones, where sound waves in

AE the air physically move a thin diaphragm (often made of plastic), which in turn
moves a metal coil back and forth across a magnet, producing an electric current.
They are notoriously sensitive to background noise and to anything standing in the
path between the sensor and the object being monitored.

E-book | The condition monitoring comparison guide

07
 Who's good—and bad—at what:
sensor technology

Oil sensors come in many different types. Some measure the oil’s dielectric
OA constant, which changes as the oil degrades or becomes contaminated. (A
substance’s dielectric constant reflects its ability to keep an electric field from
forming in it.) Other oil sensors measure optical characteristics and compare them
to model conditions to assess the oil’s quality (a technique called Fourier
transform infrared spectroscopy). Still others use magnetic fields to detect and
classify metallic particles in the oil (a sign of wear). And still others again use x-ray
emissions to detect the presence of foreign elements.
 

Thermal cameras form an image using infrared radiation, which is the same
IR phenomenon as ordinary light—just at wavelengths outside our visible range.
(Radio waves and x-rays are other sections of this same electromagnetic
spectrum.) Thermal imagers have the advantage that they can capture data for an
area, not just a single object. Like acoustic airborne sensors, they require a direct
line of sight to the objects of interest. Their main drawback is their sensitivity to
ambient temperature and to the thermo-optical properties of the objects being
monitored. As with VA and AE sensors, it’s vital to know the precise
characteristics of the asset and its surrounding environment.
 

AC current and voltage sensors are typically small clamps that attach to a wire
MCSA and convert the electric current (or voltage) into a proportional output signal.
There are non-contact versions, where the clamp surrounds but does not pierce
the wire, and contact or "vampire" versions that do pierce the wire. The contact
versions directly measure the current (or voltage) in the wire. The non-contact
versions take a slightly more complicated route: the current flowing in the wire
induces a magnetic field in the sensor, which in turn induces a current in the
sensor’s internal windings. Current and voltage sensors tend to have very high
sensitivity and bandwidth and are not location-sensitive (since the current flowing
in the wire is the same at every point along its length). MCSA systems typically
have one sensor for each wire in a three-phase AC power supply; MCSA + voltage
systems add another three sensors for voltage.

E-book | The condition monitoring comparison guide

08
 Who's good—and bad—at what:
sensor location

Sensor location
All condition monitoring sensors are sensitive pieces of electronic equipment, and in general they
must be shielded from environmental extremes: very high or low temperatures, high humidity, strong
magnetic fields, shocks, corrosive substances, dirt and grit, and so on.

Harsh environments
VA Because their sensors must be placed on or near the asset to be monitored,
vibration analysis, oil analysis, infrared thermography and acoustic emission
analysis can’t be used on inaccessible assets such as underground drainage
pumps. Assets that are remote or widely spaced (such as offshore wind turbines),
in hard-to-reach places (such as encased within larger machines), or in hazardous
environments (such as ATEX zones) make on-asset sensor installation difficult,
OA time-consuming and expensive. These proximity-based systems are also less
suitable for equipment that operates under harsh conditions, such as hot-strip steel
mill conveyors, where the sensors would be subject to damage or destruction. The
benefits of "near" vs. "on" the asset are limited; IR sensors and AE airborne sensors
have the additional disadvantage that they must be in the direct line of sight.

IR Systems based on motor current signature analysis don’t have these limitations;
their sensors install in the motor control cabinet, regardless of where the asset itself
is located. By design, motor control cabinets are generally easy and safe to access. A
single motor control cabinet contains the power lines for multiple motors, further
reducing the cost and time to install MCSA-based sensors for a complete production
AE line.

Multi-machine systems driven by a single motor

MCSA-based systems have the benefit that one set of sensors can capture data for
the entire load behind the motor; on-asset systems like VA and AE require
individual sensors for each item of interest, which can get expensive fast. On the
MCSA other hand, MCSA can't localize developing faults in a group of similar assets
driven by a single motor, because there's only one set of current (and voltage)
inputs to measure. Consider an escalator where it's crucial to know exactly which
step is malfunctioning. MCSA-based systems will correctly register a problem using
the system's electrical signals, but they can't tell you which step is causing it. If you
installed acoustic or vibration sensors on each step, however, you would know
exactly which one was failing. The visual heat map produced by infrared imaging
could also localize the problem.

E-book | The condition monitoring comparison guide

09
 Who's good—and bad—at what:
fault detection

Fault detection
Each industrial process is prone to specific will suffer, the less energy it will waste, and
kinds of equipment failure. For example, a the less it will cost to service it.
centrifugal pump may experience cavitation,
where bubbles form in the liquid being Figure 3 shows a sample P-F curve for bearing
pumped; when these bubbles collapse, the failure in a specific production system. The P-
resulting shock waves can damage the impeller F curve will be different for every kind of
(a rotating component similar to a propeller). equipment and failure mode. The curve also
Common conveyor system problems include a depends on the specifics of the condition
misaligned belt or a failure in the drive monitoring system you've chosen, as well as
coupling, pulley or gears. the environment in which the machine
operates.
Each of these failure modes has a unique
fingerprint in terms of vibration, noise, heat, oil For example, in figure 3 structure-borne
changes and so forth. Some of these acoustic emission analysis is able to detect
fingerprints are so pronounced that a sensor bearing degradation the soonest. That will
can pick them up as soon as they start to only be true if you match the characteristics of
develop; others don’t reach a measurable level the acoustic sensor to the specific bearings to
until system failure is imminent. (Think smoke.) be monitored. If you choose a sensor that's
That means it’s important to choose a sensitive in the wrong frequency range, for
condition monitoring technology that excels at example, then AE won't detect developing
seeing the problems you want to detect at an problems in this machine until much later.
early stage.
Based on figure 3, we can also conclude that
The P-F curve displays this information in this machine's environment is free of
visual form. P stands for potential failure; F extraneous noise and vibrations that would
stands for functional failure. The P-F interval is confuse piezoelectric accelerometers. The P-F
the time between when we notice that a curve for this same asset would be different in
machine is starting to deteriorate (point P) and a noisier environment.
when the machine actually fails (point F).
There's a third point, A, which is when the
machine actually begins to degrade. The closer
P is to A for a given condition monitoring
system, the better. The sooner you detect a
developing fault, the less damage the machine

E-book | The condition monitoring comparison guide

10
 Who's good—and bad—at what:
fault detection

PVA
vibration
analysis POA
oil
A analysis

PMCSA
Asset condition

PAE PIR
motor current Pnoise
structure- signature analysis infrared
audible noise
borne thermography
ultrasound

Pheat
hot to
the touch Psmoke
smoke
PREDICTIVE CORRECTIVE REACTIVE

F
Time
Figure 3. A sample P-F curve for bearing failure in a specific production system. The locations of the various technologies
on the curve will be different for each piece of equipment, production environment and failure mode, so be sure to calculate
it for the specific assets and types of degradation you want to monitor.

In figure 3, we've divided the P-F curve into is describing, last-ditch remedies. Generally,
three sections: predictive, corrective, and so much damage will have been done by the
reactive. Technologies in the predictive time the problem is detected that multiple
section will detect a developing fault very soon components will need replacement. In our
after it begins, before it does lasting collateral example, one of the two shafts connected by
damage. (Consider a loose coupling: if you the coupling may crack from the added wear.
catch it right away, you can simply tighten it.) At the extreme, you run the system to failure.

Technologies in the corrective section won't (There's a fourth section we could add to this
detect a developing fault until some collateral graph on the far left: proactive. This is where
damage has been done: in our coupling you choose the right equipment for the job: in
example, this might be excessive wear on the our example, a coupling that's rated for the
gearbox bearings. correct power, torque, starting load and so on.
But that's outside the realm of condition
Technologies in the reactive section are, in the monitoring.)
context of the specific situation the P-F curve

E-book | The condition monitoring comparison guide

11
 Who's good—and bad—at what:
general rules of thumb |
energy and performance insights

Fault detection: general rules of thumb

Every production system is different, meaning can be analyzed. Motor current signature
there's no one-size-fits-all condition analysis works well for rotating machinery at
monitoring technology. Nonetheless, we can any speed, but it only works on equipment
state some general rules of thumb. driven by an AC induction motor. That covers
the majority of industrial applications, but if
Infrared thermography and acoustic emission yours requires the use of DC motors (because
analysis are your best bet if you need to full torque at low speed or speed-independent
monitor assets that might leak air or steam. constant torque is critical, for example),
Infrared thermography is great for problems MCSA-based systems are not an option.
related to cooling and air flow. It's less Neither oil analysis nor vibration analysis can
suitable for rotating machinery; IR often won’t spot electrical problems. MCSA-based
detect a problem before damage has already systems excel there. They're also great at
occurred. Vibration analysis works well on detecting and localizing mechanical problems,
rotating equipment, but not when it rotates as long as the system isn't too complex. If it's
very slowly (less than 5 rpm). Oil analysis does critical to precisely locate faults in a multi-
work well there (as does acoustic), but it has machine system driven by a single motor,
other limitations; not all assets have oil that choose VA, AE, or IR instead.

Energy and performance insights

Tracking down developing problems before motor is drawing, you can calculate its
they cause disaster is the prime reason operational efficiency. You can use that same
companies adopt a condition monitoring data to calculate the ideal motor for that
technology. It’s what they’re made to do, after process, and to calculate whether the process
all. But MCSA + voltage has an additional is inefficient by design. (See our sustainable
advantage: it can report on the energy industry white paper [link] for more
efficiency of your monitored assets and information.)
processes.  
  If your company is also looking for metrics to
It takes both current and voltage help you shrink your environmental footprint,
measurements to make that possible. In an MCSA + voltage system can help you feed
short, if you know the current and voltage a two birds with one scone.

E-book | The condition monitoring comparison guide

12
Who's good—and bad—at what: Summary table

VA OA IR AE MCSA MCSA
+v

Remote or inaccessible assets

ATEX or other harsh conditions

Noisy or vibrating environments

One motor driving many assets L L L M M

Leaks

Mechanical faults

Electrical faults

Direct current (DC)

Rotating machinery

Very slowly rotating machinery

Energy insights

Table 1. Summarizing what we've discussed: what each condition monitoring technology does well (blue check mark) and
not so well (no check mark).

Legend

VA – vibration analysis
OA – oil analysis
IR – infrared thermography
AE – acoustic analysis
MCSA – motor current signature analysis
MCSA + v – motor current signature analysis plus voltage

L – downside: requires individual sensors for each component of interest to capture data for the entire load
behind the motor. Upside: can pinpoint failure to that specific component.

M – upside: one set of sensors can capture data for the entire load behind the motor. Downside: if these assets are
similar, only faults arising in the motor can be accurately localized to the failing component. Faults in the driven
equipment will be detected but not pinpointed to a specific asset.

E-book | The condition monitoring comparison guide

13
Conclusion

So that’s the scoop on the major condition If MCSA + voltage is among your selection,
monitoring technologies in use today. We we’d be happy to tell you more about our
hope this guide helps you choose the best solution, SAM4. Please contact us to book a
system for your situation. Or systems—there’s demo at your convenience.
no reason you can’t use two or more for
equipment that should never, ever fail.

Passionate about solving the problem of

unplanned downtime, Semiotic Labs uses
AI-driven electrical waveform analysis to
create smart condition monitoring
solutions that increase productivity while
saving customers' time and money.

www.semioticlabs.com

Contact
Lars Ligtenberg
Sales development representative
[email protected]
+31 653 325 256

E-book | The condition monitoring comparison guide

14