Vous êtes sur la page 1sur 25

CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.

XXI - Sensors in Control Systems - David Zook, Ulrich Bonne


and Tariq Samad

SENSORS IN CONTROL SYSTEMS


David Zook
Dresden Associates, Golden Valley, MN, U.S.A

Ulrich Bonne
Honeywell Laboratories, Plymouth, MN, U.S.A

Tariq Samad
Honeywell Laboratories, Minneapolis, MN, U.S.A

Keywords: Transducer, semiconductor, power dissipation, signal processing,


compensation, self-checking, biosensors, miniaturization, reliability, biology, sensing
principles, manufacturing.

S
Contents

S
LS

R
1. Introduction
2. Sensor Fundamentals and Classifications

TE
3. Sensors in Control Systems
O
4. Sensor Technology Developments
5. Biological Systems, Chemical sensors, and Biosensors
-E

AP
6. Sensor-Enabled Visions for the Future
Glossary
O

Bibliography
Biographical Sketches
H
SC

Summary

Sensors and analyzers are a control system’s window to the world. A sensor is defined
E

as a device that converts a physical stimulus into a readable output, and the definition is
N

illustrated with several examples of engineered and biological sensors. The design of
PL

sensors is driven by desired improvements on one or more of surprisingly many


U

performance features and attributes: signal-to-noise ratio, reliability, safety and intrinsic
safety, accuracy, response time, dynamic range, cost, power consumption, size,
M

electromagnetic interference immunity, etc. Recent trends and developments in sensor


technology include the increasing use of signal processing for compensation, typically
SA

used for reducing cross-sensitivity to secondary variables; multivariable inferential


sensing, which allows sensing solutions to be developed for parameters that are
infeasible to directly measure online; and self-checking and self-compensating sensors
that enhance reliability and reduce maintenance costs. The chapter also discusses
biological sensors with specific reference to olfaction and chemical sensing; in these
modalities, our best artificial sensors are no match for biology. In conclusion, visions
for improved safety and efficiency through sensor-enabled automation in automobiles,
commercial aircraft, health care, asset management, and other areas are outlined.

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

An earlier version of this chapter appeared in Automation, Complexity and Control: An


Integrated View, T. Samad and J. Weyrauch, eds., 2000. Portions © John Wiley & Sons
Limited; reproduced with permission.

1. Introduction

Control is more than information processing; it implies direct interaction with the
physical world. Control systems include sensors and actuators, the critical pieces needed
to ensure that our automation systems can help us manage our activities and
environments in desired ways. By extracting information from the physical world,
sensors provide inputs to control and automation systems. We may label our times the
Information Age, but it would be a mistake to believe that advances in automation and
control are solely a matter of more complex software, Web-enabled applications, and
other developments in information technology. In particular, progress in control
depends critically on advances in our capabilities for measuring and determining

S
relevant aspects of the state of physical systems.

S
LS
Technologists tasked with automation and control of systems of ever-increasing levels

R
of complexity, whether as designers, operators, or managers, or in other capacities, thus

TE
need to be familiar with sensor technology. The increasing sophistication of sensors and
O
sensing systems, the considerations driving this sophistication, new sensors and uses of
sensors in control systems, the increasing reliability of sensors, and the like are topics
-E

whose relevance today is not limited to sensor application specialists.


AP
O

Our objective in this chapter is to discuss sensors from these points of view. Our focus
is on the role of sensors in control systems and the trends and outstanding needs therein.
H
SC

Since excellent reviews of recent sensor developments and current applications already
exist, we give some selected examples of new sensor developments, without any claims
C

at comprehensiveness. Rather, our goal is to point out the benefits of increased sensor
sophistication as well as key approaches and areas where more understanding is needed.
E

E
N

In the following sections, we first offer a definition of a sensor along with several
PL

examples. Next, we outline the application of sensors in simple control systems and
U

discuss some of the important attributes of sensors, followed by reviews of recent


developments in sensor systems, including sensor compensation and inferential sensing.
M

We then focus on the sensing capabilities of biological systems and lessons we can hope
to draw from them. We conclude by presenting some visions for how control and
SA

automation can help realize a safer, more productive, and more prosperous future and
the role that sensors will need to play.

2. Sensor Fundamentals and Classifications

We can define ‘sensor’ as a device that converts a physical stimulus or input into a
readable output, which today would preferably be electronic, but which can also be
communicated via other means, such as visual and acoustic. As perhaps the simplest
example, consider the sensor on a keyboard switch actuator – which provides a signal
when the associated key is pressed. The keyboard switch sensor has several desirable
features. It is inexpensive, it has a high signal-to-noise ratio (its on/off impedance ratio),

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

it is compact, and it has low power consumption. Its reliability and ability to operate
over a wide range of environmental conditions are also exemplary.

Unlike most sensors, a keyboard switch sensor lacks an analog input range, and its
output is binary. Temperature, pressure, and flow sensors are more typical examples. In
these cases, the output is not a binary quantity but a value that is sensitive to a range of
those physical conditions. Figure 1 shows an example of a state-of-the-art sensor, in this
case, a mass flow sensor. As evidenced by this example, many advanced sensors today
are microscopic, microstructure devices that leverage the economies of scale and the
fabrication technologies of semiconductor manufacturing.

S
LS

R
TE
O
-E

AP
O

Figure 1: A micromachined mass flow sensor die (not packaged)


H
SC

A semiconductor sensor that provides a wealth of information is the silicon photodiode.


C

A CCD array of such devices typically generates on the order of 109 bits/second. Yet
another example of a near-ideal sensor, it is a very simple device, easily fabricated into
E

arrays with modern solid-state technology, with a very wide dynamic range (the ratio of
N

the maximum to the minimum detectable photon intensity). One characteristic that is
PL

almost uniquely ideal is its stability. It has essentially no baseline drift and excellent
U

scale factor stability. The reasons for these ideal characteristics lie in the physics of the
device. It is basically an energy converter – changing light energy (photons) into
M

electrical energy (electrons in higher energy states that can generate current or voltage).
When there are no incoming photons, there is no photocurrent – hence the baseline
SA

cannot drift. The quantum efficiency is close to unity (meaning one electron per photon)
– hence the scale factor (photocurrent divided by light intensity) is constant and stable
over many orders of magnitude. A well-designed photodiode is linear over eight orders
of magnitude in intensity and provides a primary standard for light intensity
measurement within defined wavelength limits. It is not surprising that photodiodes are
at the heart of CCD cameras and are ubiquitous in video information systems.

Still other types of sensors operate on chemical principles and may consist of single
molecules. For example, a phenolphthalein molecule (the dye in litmus paper) signals a
change in hydrogen ion concentration (i.e., pH value) by changing its color. Similarly, a

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

biosensor may be based on a molecule that interacts with a biological analyte to produce
a signal we can pick up with our senses.

The generic block diagram for a sensor shown in Figure 2 highlights the role of a sensor
as an interface between a control system and the physical world. The detector or
transducer converts a physical or chemical phenomenon into (typically) an electrical
signal. The signal processor performs one or more of various mathematical operations
on the sensed value, such as amplification, rectification, demodulation, digitizing, or
filtering. The measured and processed value is communicated to other subsystems (e.g.,
via a compatible electrical signal) or to a human (e.g., via a display). The sophistication
of these functions and of the calibration process varies widely.

The term smart sensor implies that some degree of signal conditioning is included in the
same package as the sensor. On the more sophisticated end of the spectrum, the ‘sensor’
unit can include devices or systems with elaborate signal processing, displays, and

S
diagnostic or self-calibration features. Such devices are often referred to as

S
‘instruments’, ‘analyzers’, or ‘transmitters’ (this last term is common in the process
LS
industries), usage that emphasizes that the transducer is but one part of a sensing system

R
in the context of large-scale automation. In this chapter, we will not, however, be

TE
making hard distinctions between sensor categories.
O

There are many ways to list or classify kinds of sensors. Classifications can be found in
-E

the literature based on the form of energy being transduced and on whether the
AP
transduction mechanism is self-generating (like a thermocouple or a piezoelectric
O

material) or a modulating mechanism (like a thermistor or a piezoresistor). Table 1


shows some sensor examples based on a simple classification criterion: human-made
H
SC

versus biological. Either type may be used in an automation system – human operators
can be considered part of the system.
C

Engineered sensors Biological sensors


E

Sensor Sensed quantity Sensor Sensed quantity


N

Photodiode Light intensity Retina Light intensity


PL

Psychrometer Humidity Cochlea Sound


U

Barometer Pressure Ear canal Level, rotation


Thermometer Temperature Taste bud Chemical
M

composition
Phenolphthalein pH Skin Temperature
SA

Timer Time Skin and hair Air flow


Odometer Distance (inferred) Olfactory cells Gas composition

Table 1: Examples of engineered and biological sensors

A classification focusing on the physical effects that sensors can respond to is the basis
of Table 2. One may argue about the classification of some examples, in view of their
principle of operation being based on mixed effects. For example, the Hall effect
depends on an electric current as well as a magnetic field.

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

Figure 2: Sensor block diagram

Sensing Principle Examples


Mechanical motion Pendulum-clock, quartz clock, spring balance, odometer,
(including mechanical piezoresistive pressure sensor, accelerometer, gyro
resonance)

S
Thermal (incl. temperature
LS Thermometer, thermocouple, thermistor, thermal conductivity
differences) detector, transistor built-in voltage, air flow sensors

R
Optical energy (photons) Photodiode, CCD camera, Geiger-Mueller tube, color sensor,
turbidity sensor,

TE
Magnetic field Compass, Hall-effect, magnetoresistance, inductive proximity
O
sensor
Electric field Electrostatic voltmeter, field-effect transistor
-E

AP
Table 2: Classification of sensors based on their sensing principle, with examples
O

3. Sensors in Control Systems


SC

The role of a sensor in a simple automation system is depicted in Figure 3. The


detection and measurement of some physical effect provides information to the control
E

system regarding a related property of the system under control, which we are interested
E

in regulating to within some ‘set point’ range. The controller outputs a command to an
N

PL

actuator (a valve, for example) to correct for measured deviations from the set point,
and the control loop is thereby closed.
U

Because of the simplicity of the control system example of Figure 3, it represents a fair
number of practical control systems. In especially simple systems, a distinct controller
SA

may not be immediately evident. For example, the ‘Honeywell Round’ thermostat
contains a bimetal strip as an analog sensing mechanism that responds to temperature,
and the switch attached to it serves as the actuator. This integration of sensor and
actuator turns a furnace or other space conditioning device on or off, depending on
whether the room temperature is within the set-point differential.

In general, however, the trend is to incorporate more, not less, information processing
with the sensor. The increasing complexity of sensors is in part a consequence of this
trend. In many cases, the information processing is being incorporated within the sensor
device, blurring the distinction between transducer and processor, and between sensor
and instrument.

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

S
Figure 3: Example of a simple control system
LS

R
3.1. Desirable Sensor Attributes

TE
O
All sensors, whether used individually or as part of a larger automatic control system,
need to meet a set of specifications, i.e., performance features, whereas design features
-E

AP
are typically not specified such as "intensive" vs. "extensive" sensors. The most
common and desirable types of specifications include the following:
O


H

High signal-to-noise ratio


SC

• Reliability
C

• Safety (including intrinsic safety)


• Accuracy
E


E

Fast response
• Wide dynamic range
N

PL

• Low cost

U

Low power
• Small size, compactness
M

• Low cross-sensitivity
• Ruggedness
SA

• Compatible output and display


• EMI immunity
• Corrosion resistance
• Cleanliness

We discuss several of these features below.

• High signal-to-noise ratio: The output signal of a sensor is an obvious measure


of the information generated by the sensor. However, the quality of the
information depends on the ‘unwanted’ output, or noise. ‘Noise’ usually refers
to short-term variations in the output signal, which include thermal (or Johnson)

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

noise, shot noise, and sometimes 1/f noise. But the long-term variations may be
the most important feature of a sensor. If such drift is not included in the
specifications and noise effects are estimated based on short laboratory test
periods, the user is likely, in the longer term, to experience surprises, costly
recalibrations, and downtimes or even unsafe conditions in the field. An
accelerated life test period in the factory and the inclusion of long-term stability
in the specifications is therefore recommended.

• Safety: Demonstrating the safety of a new sensor to potential users can be time
consuming. Certain standards have been established so that a manufacturer can
obtain safety certificates (such as from Underwriters’ Laboratory). Safety
considerations relate to any of numerous potential hazards: electric shocks, toxic
materials, radioactivity (several high-performance sensors rely on subatomic
particle emission and detection), acoustic or visual signals, device body
temperature, etc. In some applications, assuring safety of human users under

S
normal conditions is not sufficient. Under more stringent conditions of ‘intrinsic

S
safety’, the sensor is confirmed to not trigger an explosion of a flammable gas
LS
mixture surrounding the sensor, even under the worst failure mode scenarios.

R

TE
Low cross-sensitivity: All transducers are sensitive to properties other than those
O
we are interested in measuring with them. The key point is the extent of this
cross-sensitivity and its inclusion in the specifications. For example, a carbon
-E

AP
monoxide sensor (NDIR- or metal-oxide- based) may also be sensitive to
changes in carbon dioxide or even in humidity. Sophisticated sensors and
O

analyzers are able to compensate or eliminate such cross sensitivity, but also
demand a higher price. Differential pressure sensor outputs can shift when the
H
SC

absolute pressure changes, unless proper compensation is provided.


C

• Ruggedness: An appreciated feature for users is to be able to work with sensors


that are immune to shock and extremes in temperature, humidity and pressure,
E

so that the output signal remains unaffected by changes in environmental


N

PL

conditions. For example: with piezoresistors used as strain or pressure sensors,


their resistance is a function of the strain the device is exposed to. But resistance
U

does not depend only on strain; it is also a function of temperature. Similarly, a


dye whose color changes when exposed to a certain chemical agent will also
M

exhibit color variation with temperature.


SA

• Service life and reliability: Sensors are ultimately physical devices and, as such,
have finite service lives and mean times between failures (MTBF). They can go
out of calibration, elements in them can fail entirely, etc. Obviously, control
actions taken based on faulty sensors can have catastrophic consequences, so
that steps to prevent such events need to be considered: scheduled checks and
maintenance, redundant sensors, sensor self-checks, or even sensor ‘health
monitoring’. Such features add components, cost, software and signal
processing, and, yes, complexity, with all its ramifications discussed in this and
other chapters. The continuing demand for reduction in operating staff in many
industrial and government facilities has led to a recent emphasis on automated
diagnostics and prognostics (predictions of system degradation).

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

• Compact size: Reductions in the spatial dimensions of sensors permit


embedding them within systems that are themselves small-scale. Historically,
new application areas for sensors have frequently been opened up through
advances in miniaturization. A recent example is automotive engine control
systems, which now embed sensors for mass airflow, crank angle, ambient
temperature, humidity, pressure, and incipient knock. Another important area
where compact size is making an impact is biomedical devices. In the past,
implantable devices provided periodic stimulation regardless of the state of the
system. With miniaturization in sensors, devices today are more ‘intelligent’:
stimulation can now be based on demand, i.e., on accurate measurement of
physical parameters. For example, blood sugar of diabetic patients can control
insulin release and pressure waves caused by a patient’s muscle movement, or
body motion can be sensed and used to adjust pacemaker stimulation rate.

• Compatible output or display: The output of a sensor has to be communicated to

S
other parts of the automation system – for data logging, operator display, control

S
algorithms, etc. Efficient communication means can greatly simplify sensor
LS

R
integration in a control system. In a process plant, the cost of wiring for
communication is often a large component of the cost of sensor installation,

TE
especially in large control systems. A possible solution is the use of wireless
O
communication.
-E


AP
Low power: Power has to be supplied to the sensor. In many industrial and
building applications, the cost of wiring for power can even exceed the cost of
O

the communication channel. Low power consumption opens up other


alternatives, such as scavenging power from the communication bus or RF
H
SC

communication using battery power at the sensor. Batteries can be recharged by


C

environmentally powered sources such as solar cells in some installations.

• Low cost: Most industries that make extensive use of sensors are under perpetual
E

cost-reduction pressures. This issue is not unique to sensors, but it is worth


N

PL

noting here as much as in association with any other aspect of control and
automation systems. Informed decisions on cost-benefit tradeoffs need to be
U

made to achieve the desired performance and safety at minimum cost.


M

Automation solutions for large-scale, complex systems will be sensor-rich. The


developers and operators of such solutions (as distinct from the sensor suppliers)
SA

often tend to view sensors as commodities – selection is based on cost;


performance differentiation is typically not viewed as significant. Leapfrog
technology can, of course, still demand a premium, but in many cases solution
and system providers would only check that minimum performance
specifications are met.

• Cleanliness: does not contaminate the environment it is sensing: In ultra-clean


environments (e.g. in the Semiconductor Industry), any new material introduced
into the processing chamber needs to be evaluated for its potential of
contaminating that process. Similar considerations need to be made to clear new
sensors for the food and pharmaceutical industry. One widely adopted solution

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

is to use very inert materials, at least for the external skin of the sensor such as
chemically polished stainless steel.

The above issues are examples that have motivated research and development on sensor
technology and on sensor applications, as will be discussed below.

3.2. A Tradeoff between Sensor Performance and Power Dissipation

Since a sensor is a provider of information, it would seem natural that the quality of a
sensor can be evaluated by the quantity of information it can provide to a control
system. Thus it seems reasonable that a measure of sensor quality or performance would
come from information theory. One of the benefits gained by adopting an information-
theoretic perspective is a better appreciation of the tradeoff between sensor properties
and sensor power dissipation. We discuss the topic briefly and in general terms in this
section; a full quantitative explication remains to be developed.

S
Information generated by a sensor can be considered to be a series of messages related
LS
to some characteristic of the physical system whose attributes are being measured. The

R
information reduces our uncertainty as to what state the system is in. Information theory

TE
provides a measure of the amount of information in any message called the information-
O
theoretic entropy to distinguish it from the entropy of thermodynamics and statistical
mechanics. It is simply equal to the minimum number of binary digits (bits) necessary
-E

to transmit the message.


AP
O

It takes energy to generate information. Shannon showed that a certain amount of


energy per bit is required to distinguish signals in an electromagnetic communication
H
SC

channel from the unavoidable background of thermal noise. The energy per bit is given
by kTln2, where k is Boltzmann’s constant and T is the absolute temperature. Improving
C

the maximum signal-to-noise ratio (equivalent to dynamic range) or the response time
(related to the bandwidth of the sensor output) of a sensor requires increased power
E

dissipation.
N

PL

The discussion above is in the context of digitally encoded measurements. However, the
U

transducers used for sensing parameters in control and automation systems are not
digital devices that generate coded messages as finite strings of bits. Rather they are
M

analog devices that generate a voltage, current, charge, or change in resistance that must
be amplified. It is worth noting here that the scale-up properties of analog
SA

communication are substantially poorer than for digital communication. Doubling the
dynamic range of an analog sensor output signal requires at least a doubling of the
power required for communicating that signal. On the other hand, if the output is
expressed in digital form, a doubling requires one additional bit.

4. Sensor Technology Developments

The objective of modern sensor developments is to furnish improvements on one or


more of the desirable features listed earlier. For example, if the sensor stability or cross-
sensitivity does not meet the level required for an application, the sensor can drift out of
calibration, its output can lose its meaning, and the system can lose its ability to control.

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

The addition of means and designs (electronic compensation, signal processing, self-
checks, etc.) that prevent such occurrences lead to more sophisticated and complex
sensors. Contributing to this complexity is the desire to offload central data processing
onto local or distributed processing (i.e., to make individual sensors ‘smarter’).
Furthermore, to achieve one desired output, a cluster of sensors, a sensor system,
analyzer, or instrument may be needed.

Similarly, increasing demands on the performance of systems, such as those providing


transportation or chemical processing (greater fuel efficiency, fewer emissions, greater
safety, reduced labor content) cause the complexity of the involved automatic control
systems to also increase, together with their need for better and/or new sensors.

We discuss some salient trends toward increasing sensor complexity in this section.

4.1. Compensation through Signal Processing

S
As alluded to above, one of the key problems in developing sensors that are accurate
LS
over a range of environmental conditions is cross-sensitivity. Most transducer

R
mechanisms exhibit some influence to variations in properties other than the one they

TE
are designed to measure. A flow transducer, for example, may produce a different
O
output as temperature varies, even with the flow remaining invariant. For applications
where precision or accuracy is not required, the errors that result from this coupling may
-E

be tolerable. In most modern automation systems, however, this is not the case.
AP
O

Modern sensors correct for cross-sensitivity by incorporating additional transducers for


measuring the secondary variables and then employing signal processing and/or
H
SC

statistical techniques to ‘cancel out’ the undesired influence. As an example, we discuss


a temperature-compensated flow sensor for which the sensor output, y , is computed as
C

a function of two sensed signals: the temperature, T , and the flow, x . The form of this
function is based partly on our knowledge of the physics responsible for the cross-
E

sensitivity and partly on a formula whose parameters are determined by regression:


N

PL

0.03
⎛T ⎞
y=⎜ ⎟ (a + a1 x p1 + a2 x p2 + a3 x p3 )
U

0
⎝ T0 ⎠
M

This expression tells us that the temperature effect is fairly small – a ±50°C excursion in
SA

T from a nominal temperature T0 of 288 K only changes y by ±0.5%. The ai ’s and pi ’s


are parameters that are fit using data collected from controlled tests. An example of a fit
is shown in Figure 4, where the horizontal axis shows the flow sensor output
(labeled ΔG ) and the vertical axis shows the actual flow rate ( Vs ).

The S L quantity also graphed in the figure is the logarithmic sensitivity of the final
output to sensor noise, defined by SL =| ( x / y )(∂y / ∂x) | . The figure shows that as flow
increases, this sensitivity increases to S L > 1 , a consequence of the output signal
saturation. The signal noise (not shown) also varies over the flow range, from 10% at
low flows to 1% at high flows. Figure 4 shows that the saturation negates the benefits of

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

reduced relative signal noise at high flows. In more complex functions, S L can adopt
values of 10, 50, or greater, especially when there are multiple xi (worst case: an array
of similar sensors) and interactions between them ( S L in the multivariable case is a
summation of the individual sensitivity magnitudes).

4.2. Inferential and Higher-value Sensors

The example above illustrated one of the simpler uses of signal processing or statistical
techniques in sensors. Over the last few years, the notion of producing a sensor or
instrument output that is derived from multiple sensing mechanisms has been extended
to a higher level. This concept has reached the point where ‘sensors’ for parameters that
have not been feasible to measure cheaply and reliably can now be developed for broad-
based use. The terms ‘inferential sensors’ and ‘higher value sensors’ are often used for
such systems, but their connection with traditional definitions of sensors is tenuous. In
most cases, inferential sensors are not hardware products, unlike most sensors. They are

S
more likely to be software implementations on control system platforms that can access
LS
sensor inputs from distributed locations.

R
TE
O
-E

AP
O

H
SC

C
E

E
N

PL

Figure 4: Characterization of 200 L/min flow sensor BF #123 with pure nitrogen.
Vs = 2.8836ΔG 0.93 + 2.9743 × 10−7 ΔG 3 + 2.8886 × 10−13 ΔG 4 ± 2.02%
U

One example is a ‘comfort sensor’. Devices (other than biological ones!) that directly
sense comfort are difficult to even conceive of. However, it seems reasonable to
SA

postulate that comfort is correlated with parameters that we can indeed measure. The
problem then is a matter of determining the function (or mapping) from measurable
parameters to comfort. Formulations of comfort indices have been proposed, with
Fanger’s Predicted Mean Vote (PMV) measure having gained some degree of
acceptance. PMV is based on six variables: air temperature, mean radiant temperature,
relative air velocity, humidity, activity level, and clothing thermal resistance. The
International Standards Organization has adopted PMV as its thermal comfort index
standard and ‘comfort sensor’ products have since been developed.

In most cases, however, the relationship between the measured properties and the one
that is desired to be inferred is unknown. In these cases, an empirical model can be

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

developed. A popular methodology for such modeling is artificial neural networks. A


detailed explanation of neural nets is beyond the scope of this chapter; suffice it to note
that they provide a nonlinear statistical regression technique that has some attractive
properties for high-dimensional problems relative to classical models such as the (low-
dimensional) temperature-compensated-flow example shown above. Developing an
inferential model using neural networks (or any other empirical method) requires first
the collection of data from the system. The data must also include measurement of the
inferred property value as determined through a temporary sensor or through an off-line
laboratory analysis – these devices are not needed once the inferential sensor is
developed. The development and use of an inferential sensor is outlined in Figure 5 for
a process industry scenario. Inferential sensors are now widely used in the process
industries, such as for estimating octane number in oil refining, smokestack emissions
in power plants, and lignin concentrations in wood pulping.

Both this subsection and the previous one highlight the increasingly multidisciplinary

S
aspect of sensor technology today: statistics, signal processing, and software are now

S
critical aspects of sensor development.
LS

R
TE
O
-E

AP
O

H
SC

C
E

E
N

PL
U

Figure 5: Inferential sensor development (top) and operation (bottom)


SA

4.3. Self-checking and Self-compensating Sensors

From the above discussion on the key need for sensor reliability and stability, it is clear
that demand for self-checking and self-compensating sensors is high. Electronic scales
today automatically reset their zero settings before objects are placed on them, based on
the assumption that most of the time the scales are empty. Even the span or actual
weight reading can be easily checked by calibration with a standard weight. Precision
pressure transmitters cannot depend on such methods since they are under load most of
the time. They must depend on the stability of the piezoresistive elements and of the
sensor package to maintain stable baseline and span stability. However, in some cases,
auxiliary actuators or valves may be used to enable performing a check on baseline and

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

span for pressure sensors and flow sensors. The following examples illustrate such
methods.

4.3.1. Chopper as Auto-zero Device

Infrared detectors that respond to very small temperature differences can be viewed as
the classic case where a mechanical actuator is used to provide a reference signal. A
lens focuses an element of the scene onto the detector and a rotating chopper wheel is
placed in the optical path. The detector alternately sees the scene and the chopper blade,
which is always at the same (reference) temperature. A timing signal from the chopper
is sent to a switch that alternates the polarity of the output from the infrared detector.
Thus the reference signal is subtracted from the signal from the scene. The synchronous
signal corresponds to the difference in temperature between the scene and the reference,
and the nonsynchronous baseline drift is ignored. It is not unusual for temperature
differences of 0.01ºC (corresponding to 0.00001ºC at the detector) to be easily seen in a

S
thermal imager.

S
LS
Chopping a signal and synchronous detection are widely used in optical systems, for

R
example in infrared spectrometers and gas analyzers. The motivation is frequently (but

TE
not always) the IR detector drift. In the optoacoustic method for gas concentration, the
O
gas is enclosed in a chamber that acts as a resonator. The chamber is illuminated with
IR radiation from a filtered chopped IR source. A microphone is used to pick up
-E

pressure changes that occur only if analyte molecules absorbing at that wavelength are
AP
present. Chopper-stabilized amplifiers using switches at the input of an amplifier are
O

similarly used to convert a dc voltage to an ac voltage to eliminate the effect of baseline


drifts.
H
SC

4.3.2. Self-checking Flame Sensors


C

Highly reliable optical sensors are needed to monitor the presence of flames in furnaces
E

or boilers so that fuel flow can be interrupted if the flame should go out for any reason.
N

Despite the proven stability of the ultraviolet sensors employed, a further layer of safety
PL

can be added to positively check that the sensor’s response still meets the ‘fail-safe’
U

requirement. A mechanical shutter (chopper) that interrupts the optical path between the
sensor and the flame during known intervals is used to perform such checks on flame
M

sensors that are certified to be ‘fail-safe’. Another proven strategy is to use a unique
signature or response of the object to be sensed (e.g., the flame-rectified current signal
SA

of a flame ionization sensor). In this case, ac interference from faulty wiring appears as
a nonrectified ac signal that is readily distinguished from the rectified flame signal.

4.3.3. Self-checking Oxygen Sensor

The classical Nernstian, ZrO2-based oxygen sensor generates a dc signal that depends
on the logarithm of the ratio of unknown/reference oxygen concentrations. One check
on the health of such a sensor consists of measuring its impedance: if the value is above
a set limit, its health is suspect and it should be replaced. A more sophisticated approach
is to force oxygen into and out of a closed ZrO2 cell (by application of a fixed voltage
between reversals) so that set outputs are reached. The ‘pumping time’ can thus become

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

a linear measure for the absolute oxygen concentration. In addition to checking the
cell’s impedance, any asymmetry in the pumping time or current profile reveals a faulty
sensor.

4.3.4. Self-compensated Flow Sensor

A different calibration problem arises in the case of a mass air flow sensor based on
thin-film thermal anemometry. In practice, the output signal depends (not on pressure,
but) on the thermal properties of the gas being sensed, in spite of its mass flow
dependence being touted in many textbooks. A theoretically correct compensation can
be made if at least the thermal conductivity and specific heat of the fluid are known.
Means to measure these on-line have been developed employing one additional thermal
microsensor mounted in a recess, away from the flowing fluid. This approach now
enables the on-line determination of fully compensated (composition, temperature, and
pressure) flow of any thermal flow sensor, although it requires additional calibration of

S
thermal conductivity and specific heat sensors.

S
LS
A much simpler approach for composition correction, which can be applied to thermal,

R
pressure-difference-based and other flow sensing approaches, is based on mechanical

TE
actuation: While one sensor measures the main flow, a second one of equal design is
O
exposed to the small dither-flow generated by a small membrane, in a side cavity. Self-
compensation for changes in fluid composition, temperature and/or pressure can be
-E

achieved by simply forming the ratio of the two sensor signals. Fig. 6 shows a sketch of
AP
this sensor’s operating principle, as realized with thermal flow microsensors.
O

H
SC

C
E

E
N

PL
U

M
SA

Figure 6: Sketch of self-compensated volumetric flow sensor, using two thermal flow
microsensors and an earphone-speaker membrane actuator.

4.3.5. Rebalanced Sensors

Another way an actuator is used to eliminate the effects of zero-drift and nonlinearity is
to balance the input variable with an opposite effect to keep the input to the sensor at
zero. In a force-rebalanced accelerometer, an electromagnetic force balances the force
on the proof mass due to acceleration. The rebalancing force is linearly proportional to

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

the actuator current and provides the measure of the acceleration. The proof mass is
maintained at its rest (or null) position so that a highly sensitive method such as a
scanning tunneling microscope (STM) tip can be used as the sensing mechanism. Such
a sensor can easily detect sub-Angstrom motion and thus is an ideal null detector, even
though its output is highly nonlinear.

4.4. Sensor Manufacturing

Unlike many other critical automation technologies, sensors are fundamentally


‘hardware’ devices. Manufacturing, in the traditional sense of the term, thus remains an
important consideration for complexity management. Manufacturing for sensors covers
both the ‘base’ detector (increasingly based on microelectronic fabrication processes)
and the assembly of the packaged sensor. The software-intensive nature of most of the
technologies discussed in this book may lead to the conclusion that manufacturing has
been reduced to irrelevance. This would be a mistaken extrapolation. In fact,

S
manufacturing innovations have been at the center of the development and successful

S
commercialization of new sensors – e.g., pressure sensors with automated testing to
LS
customize individual compensation algorithms, mass airflow sensors, laser-based

R
gyroscopes, and proximity sensors.

TE
O
Usually a business plan for a new sensor product is developed in response to either
demand from an identified set of customers or a new capability, technology, or sensor
-E

design resulting from research and innovation. Potential customers are asked to help
AP
define a ‘minimum common denominator’ specification for the new sensor and to help
O

identify the potential market size. Based on laboratory and field test results, there must
be confidence that the specifications can be met. Before productization can proceed,
H
SC

alternative technologies or manufacturing processes must be examined, the intellectual


property position (proprietary information and patents) must be assessed, and possible
C

manufacturing scenarios dependent on volume, labor cost, and location must be


considered. But we have left out one key parameter of the productization and
E

manufacturing process: the human factor. It is the one that significantly increases the
N

complexity of the process, but its discussion falls outside the scope of this text.
PL
U

4.5. Other Drivers for R&D


M

Several additional factors drive research and development in sensor technology, which
we briefly note here:
SA

• Ease of application: For sensor data to be meaningful, not only must the sensors
be working as assumed, but they must also be positioned properly and installed
correctly (i.e., consistent with the control system model). Because field
installations are labor intensive and costly, ease of installation can be a key
competitive edge for a sensor.
• Sensor-configured system: The performance of the control system is limited by
what we can measure and the feasibility of obtaining measurements. To achieve
optimal system performance, we need to have a thorough understanding of the
behavior of the system, its controls, and its sensors. We can no longer afford to
design a system, add controls at a later stage in its development, and add its

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

sensors at a still later time. A smart designer will therefore keep in mind the
capabilities of available sensors as he or she designs and configures the system.
• Rangeability and adaptability: In line with pressures to reduce the cost of
inventory, parts, and installation, sensors have become more flexible and
remotely programmable. For example, a more limited range of a lower cost
microprocessor may be used if the sensor range can be reprogrammed after
installation. In addition, this added flexibility enables a reduction in inventory.
Ultimately, one envisions such flexibility and reprogramming to be done
automatically, by the ‘smart’ sensor itself, in response to the status of the
system.
• Integration: Further cost savings are achieved by merging the data and signal
processing of the sensor and control. Integration of the sensor functions of
Figure 2 into one monolithic block of a silicon chip is also being practiced,
especially where fabrication volumes are high (perhaps more than 1
million/year) and the fabrication processes are compatible. Examples include

S
Honeywell’s integrated pressure sensor and Analog Devices’ integrated

S
accelerometers, which incorporate both sensing and data processing on a single
LS

R
silicon chip. Further integration will include means for I/O, such as is needed for
wireless communication. Additional benefits from such integration are

TE
compactness, ruggedness, and low power consumption.
O
• Sensor arrays and massive data processing: It is now affordable to manufacture
thousands of silicon-based sensors as compact arrays. This capability has
-E

AP
enabled the development of such sensors as the CCD camera noted earlier and
many other ‘imaging’ sensors. These other examples include scanners and
O

processors for imagery and tomography of signals from infrared, ultrasonic, X-


ray, and electron- or nuclear-spin resonance signals.
H
SC

• Extreme sensitivity: The need to detect parts-per-billion and even parts-per-


C

trillion concentrations of trace pollutants, explosives, nerve gases, and bacteria


has given impetus to the development of sensors based on solid-state surface
effects, biological sensors, and fluorescent chemosensor techniques. The latter
E

have zero baseline and thus (in principle, at least) allow the detection of single
N

PL

molecules. New developments achieving multi-stage pre-concentration gains of


analytes in gas and liquid analyzers (~10,000×) have been reported.
U

• Extreme accuracy: Some applications require sensor accuracy in the ≤±0.01%


M

range. Differential pressure sensors with such capabilities are now available and
can maintain this accuracy despite changes in temperature and absolute pressure.
SA

5. Biological Systems, Chemical Sensors, and Biosensors

Biological systems provide an existence proof that low-performance sensors can


provide the meaningful data needed to perform sophisticated control functions.
Consider the flight control system of a fly – taking evasive action from the fly swatter
and landing safely on the ceiling or a male moth finding a female several miles upwind.
The sensing and control system performance in low-power, compact packages that is
seen in the biological world is still a futuristic vision for engineers and technologists;
the realization of this vision is not imminent.

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

In Table 1, we presented a list of biological sensors and sensing mechanisms. In the area
where we have nearly ideal sensing characteristics (electronics and photonics, with
CCD imagers providing a particularly good example), we can outperform biological
sensors, even the best eyes that have evolved for night vision. For other sensing
modalities, chemical and olfactory sensing in particular, we are far behind the
performance of biological systems.

5.1. Chemical and Olfactory Sensing in Biological Systems

A long-standing need exists for chemical sensors – for environmental control, for
process control, and for testing almost everything from air and water quality to food and
medicine. Microminiaturization of NMR, Raman, MS and GC analyzers is progressing
at a fast pace, driven by customer's demand for low-power, compactness, speed and
low-cost. We will only mention one example for such ongoing developments: that of
MEMS micro gas chromatograph. Gas chromatography stands out among the others for
its simplicity. By reducing the diameter (~100 μm) and thickness of the adsorber-film

S
phase (≤ 0.2 μm), analysis times below one second have been demonstrated. A related
LS

R
development is the phased heater array structure for enhanced detection (PHASED).
This is a Si-chip-based micro gas chromatograph featuring multi-stage pre-

TE
concentration, separation and thermal conductivity detection. Figure 7 shows a sketch
O
of such a system: the sample gas (serving also as carrier gas) inlet is at left and drawn
through ~100 x 100 μm channels etched into silicon.
-E

AP
O

H
SC

C
E

E
N

PL
U

Figure 7: Functional block diagram of a PHASED sensor system. The analytes are
SA

concentrated, "injected" and separated, before being sensed by the differential thermal
conductivity detector.

Composition or chemical micro-analysis is an area in which biological systems may


have sensitivity advantages over today’s manufactured sensors, but are still struggling
with meeting stability goals. Biological chemical sensors depend on protein sites on the
cell surface, which have a selective affinity for certain molecules. During the last
decade, specific olfactory receptors have been identified. Coupled with advances in
genomics and cellular biochemistry, such identification provides the beginnings of a
molecular understanding of olfaction. Adsorption of the odor molecule causes a change
in the firing rate of the olfactory neuron. It is believed that moths can detect sex
pheromones at sensitivities approaching the level of individual molecules. Humans,

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

although much less sensitive, can discriminate between more than 10,000 odor
molecules (odorants) and have about 50 million olfactory neurons that bear about 1000
different types of odorant receptors in four so-called expression zones. Within each
zone the odorant receptors are distributed randomly, presumably to provide redundancy
in case of damage in a small area.

The evolution of odor receptor proteins has not solved the cross-sensitivity problem
discussed above, but in fact makes use of it. Individual types of odor molecules activate
multiple receptors, with each receptor type having a different affinity (and therefore a
different proportionality between concentration and firing rate) for different odorants.
Such a sensor array approach is likely to be a necessary feature of manufactured
(nonbiological) chemical sensors for organic molecules, since there is little utility to a
sensor that will respond to only one type of molecule. The size of the sensor array is an
important issue in chemical sensors. Mammals (and even worms) typically have 1000
types of odorant receptors, whereas fruit flies make do with as few as 100. Since

S
odorants interact with multiple receptors rather than individual ones, the possible

S
combinations exceed by orders of magnitude the number of odors an animal can
LS
actually distinguish. Presumably, animals only discriminate between odors that are

R
biologically important to their survival.

TE
O
Recently, several research groups have developed ‘artificial noses’ inspired – however
loosely – by mammalian olfactory prowess. These devices are based on different types
-E

of sensor technologies, including metal oxide semiconductors, conducting polymers,


AP
surface acoustic wave, and gas chromatography. This is an exciting topic with
O

considerable application potential: food and pharmaceutical manufacturing,


environmental monitoring, security systems, and military applications. Although the
H
SC

best of these systems compare poorly with biological noses – they have at most a few
dozen sensors and cost up to $100,000 – this research promises to bring a sensory
C

modality that has hitherto been the exclusive preserve of biology within the domain of
engineered systems.
E

E
N

An important and highly desirable feature of the olfactory mechanism is that it is


PL

apparently unresponsive to odorless molecules, whereas manufactured sensors tend to


U

be sensitive to oxygen and water vapor, in particular. They also respond to odorless
combustible gases such as hydrogen and methane. An exception to this is the use of
M

chemical sensors based on conjugated polymers. These have demonstrated high


specificity together with the high sensitivity characteristic of biological sensors,
SA

although they are based on quite different principles. The transduction strategy is to
synthesize polymers that are poly-receptor assemblies that act collectively. The
electrical conductivity of conjugated polymers is a well-known example of a property
determined by low concentrations of defects. Fluorescence is another example, where a
single defect can significantly quench the fluorescence of the entire chain.
Chemosensors for TNT have been demonstrated that can detect concentrations in the
parts per billion, without significant interference from environmental constituents such
as water vapor and CO2. Further developments in this area could be very exciting.

The biological process has some undesirable features as well. It can have a nonlinear
response, which can be misleading. For example, a low concentration of indole has a

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

floral scent, while a high concentration has a putrid odor. These features are not
unexpected for a system developed by evolution since there is no apparent survival
advantage to being able to identify strong odors. The signal processing method (as
realized in the olfactory cortex) that evolved along with the chemical receptors did not
get trained on high concentrations.

5.2. Audition

An example of a nonolfactory biological sensing system that we can learn from is the
biological sonar systems used by bats. The fact that bats use the ‘chirp radar’ technique
was well known long before it was rediscovered during World War II. Bats emit
complex ultrasonic signals and listen to the returning echoes for orientation information
and for hunting flying insects. There are nearly 800 species of bat, and probably all of
them use echolocation. Range information is conveyed by echo delay, whereas velocity
information is carried by Doppler shift. Frequency is detected by the stimulation of hair

S
cells and is coded by the anatomical location of a spiral array of ganglion cells. The rate

S
and pattern of nerve impulses express the amplitude, duration of signals, and time
LS
between signals. The signal processing systems use delay lines, multipliers (or AND

R
gates), and frequency tuners that are tolerant of wide variations in signal level. Signal

TE
processing in the auditory system is parallel-hierarchical. All of this sophisticated
O
sensing and signal processing takes place in a very small volume with low energy
usage. If similar sensing systems could be produced by micro-machining and integrated
-E

circuit techniques, they would undoubtedly find application in security and surveillance
AP
systems.
O

5.3. The Role of Sensors in Biological Control Systems


H
SC

The role of sensors in biological systems is illustrated in Figure 6 and is similar in many
C

respects to the role of sensors in engineered control systems. The biological senses
provide the interface between the organism and its environment. Many biological
E

organisms constantly probe their environment and compare the information from
N

different senses, providing a means of calibrating the sensory mechanisms.


PL
U

The building blocks of living systems are cells, and each cell has evolved its structure
and its control systems to improve its survivability. Even the simplest single-cell
M

organism has amazing information processing, sensing, and control system mechanisms
with complex feedback networks, all tuned to ensure survivability in its environment.
SA

Natural selection provides the mechanism for producing systems that optimally use
resources and avoid threats. The billions of generations of evolution have not only
produced optimum response to the environment, but because of the history of changing
environments, they have produced systems that can adapt and operate effectively (i.e.,
survive) in the face of changing environments – changing resources, changing threats,
etc. The incredible success of living systems suggests that we have a lot to learn from
them, despite their complexity.

Our knowledge about the information pathways and feedback mechanisms in biological
systems is meager. We know that information is stored in the sequences of nucleic acids
in RNA and DNA and communication takes place by electrochemical pulses and

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

complex biochemical reactions. Biological sensors use the same mechanisms. Since
sensing and information processing are both critical to the control system, it is a safe bet
that they co-evolved and are closely tied together.

An evident aspect of biological evolution is the increase in complexity of the system.


Cells became more complex and evolved into multicell organisms. Reproduction
mechanisms have evolved, as well as sophisticated mechanisms of communication
between organisms. In fact, evolution can be viewed as the accumulation of useful
information. As such, we may expect that control systems will continue to evolve as
well. Here we are also learning from biological systems. The concepts of neural
networks and genetic algorithms were inspired by biological systems.

S
LS

R
TE
O
-E

AP
Figure 8: Schematic depiction of biological sensing
O

We can make several inferences from a comparison between biological systems and
SC

today’s control systems:


C

• Increased reliability and survivability (robustness) are achieved by increased


E

redundancy, which means increased numbers of sensors and interconnections


and therefore increased complexity.
N

PL

• Adaptability and flexibility go hand-in-hand with increased complexity.



U

In terms of complexity, today’s control and automation systems are in their


infancy compared with even the simplest biological systems.
M

• As we understand biological systems better, we learn from them. Examples


presently in use are neural networks and genetic algorithms.
SA

6. Sensor-Enabled Visions for the Future

We have noted a number of trends in the use of sensors for automation and control.
These are driven by the desire to have machines replace humans in performing tasks
wherever it is economical and safe to do so. A familiar example is the increased use of
sensors and computer controls in automobiles. Computers and sensors not only control
the engine, but also the brake system, the airbags, the windows and door locks, and even
the radio antenna. Some of these uses are simply for convenience, some are for safety or
efficiency, and some are required for environmental protection. In all cases, they add to
the cost and complexity of the vehicle, but clearly the benefit is perceived to be worth it

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

– the automotive industry operates in a highly competitive, consumer-driven, and


regulation-driven market. In the future, we can envision automobiles that warn us of
other cars or of traffic problems ahead, that suggest an alternative route; sensor systems
already exist that let us know when we are approaching the next turnoff or when
maintenance tasks need to executed. This requires the development of low-cost
navigational and collision avoidance sensors. These are all devices that can be expected
to become affordable during the next decade, as sensors and microanalyzers continue to
improve in smartness, reliability, ruggedness, compactness, low-power consumption,
and wireless communication interfaces and networking, .

In the aviation enterprise, we can also expect increased use of sophisticated navigation
and collision-avoidance sensors. We can envision the total elimination of airplane
collisions, and the ability to choose routes that avoid turbulence or severe weather. In
this case, hardware developments are needed in the area of radar and laser scattering to
detect clear air turbulence and nearby aircraft. The reliability of the aircraft would be

S
improved by using sensors to detect bearing wear and structural fatigue and to predict

S
when parts are about to fail so that they can be scheduled for maintenance before they
LS
do so.

R
TE
Sensors can also be used to improve the safety and reliability of our basic infrastructure
O
– the essentials that many of us take for granted, such as electricity, clean water, and
clean air. With widespread environmental monitoring, pollution sources can be
-E

identified sooner and breakdowns in the distribution systems (for water and electricity)
AP
can be detected and bypassed. As resources become scarcer, sensors and controls will
O

be increasingly used to improve efficiency and regulate consumption.


H
SC

Health care is an area where the use of diagnostic tests and instruments is growing
rapidly. Improvements in capabilities are becoming possible through the use of DNA
C

sequencing to identify genetic defects and diseases at the very early stages, when they
can be cured or prevented. At present, diagnostics are offline, but we can imagine that
E

sensors on the wrist could tell us when we are overstressed or overindulgent. With
N

instant feedback we would be reminded to close the loop and change our actions to
PL

optimize our health. Although the cost of diagnosis and testing is increased, the overall
U

cost of health care will be decreased.


M

An area that is not so obvious, but that also has great potential for productivity
improvement, is asset management. Significant time and resources are spent in
SA

industries, hospitals, and offices looking for equipment or other assets that are not
where they are presumed to be. Inexpensive tags and sensors that can remotely detect
the tags can locate and track assets. At the extreme end of the spectrum, the homeowner
could use his home computer to keep track of items that may get misplaced in the home,
such as keys or even eyeglasses. This may seem like overkill, but the technology would
not have to be much different from that used to keep track of clothing in a department
store.

The above examples suggest that the possibilities for increased use of sensors in
automation and control systems are endless. Improvements in the performance of
existing sensors, improvements in reliability and robustness, and the development of

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

new sensors, further integration of sensors within control systems are some of the trends
in evidence today. As always, research and innovation is driven by application
demands. In the case of performance, critical process control and the control of new
processes such as biologically based manufacturing of pharmaceuticals are some of the
drivers. However, many of the applications being envisioned do not require high-
performance sensors and transmitters that generate high-quality data. The long
development times and high costs of sensor development and maintenance lead us to
conclude that cost reduction of existing sensors and finding new ways of using them
will continue to dominate the field of sensor development.

Glossary

Artificial nose: An engineered sensor for detecting or measuring odors


Charge An array of capacitors able to transmit their charge profile (e.g.
Coupled reflecting the light inputs to an associated array of photo-detectors)
Device (CCD): to a receiving entity at the end of the array

S
Cross- The sensitivity of a sensor for a particular property to variations in
sensitivity: other properties
LS

R
Detector or A device that converts a physical phenomenon into an electrical
transducer: signal

TE
Instrument, A complete sensor unit, often with elaborate signal processing,
O
analyzer, or displays, and diagnostic features
transmitter:
-E

AP
Inferential or Sensors that produce an estimate of a property without measuring it
higher-value directly, instead performing statistical calculations on outputs from
O

sensors: multiple sensing systems


H

Intensive (vs. A category of sensors whose output signal is independent on the


SC

extensive) amount of material involved in the measurement, such as a


C

sensor: thermocouple or U-tube manometer. Examples of extensive sensors


are piezoresistive pressure and resistance temperature sensors
E

Self-checking: The ability of a sensor to test its correct functioning automatically


E

Self- The ability of a sensor to automatically compensate for


N

PL

compensation: inaccuracies resulting from drift or other device changes


Sensor: A device that converts a physical stimulus into a readable output
U

Smart sensor: A sensors with signal conditioning and processing included in the
M

sensor package
SA

Bibliography

Allen, P. (1984) Sensors in silicon. High Technology, September, 43–50. [Description of an integrated
pressure sensor which incorporates both sensing and data processing on one integrated circuit chip.]
Ayres, R.U. (1994) Information, Entropy, and Progress – a New Evolutionary Paradigm. American
Institute of Physics, Woodbury, NY. [This book presents an information-theoretic perspective on
evolution.]
Barron, A.R. (1993) Universal approximation bounds for superposition of a sigmoid function. IEEE
Transactions on Information Theory, 39(3), 930–945. [This article identifies the inherent advantage in
predictive accuracy of statistical models with adjustable basis functions—such as feedforward neural
networks—compared to standard multiple linear regression.]

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

Bonne, U. (1990) Flowmeter fluid composition correction. U.S. Patent No. 4,961,348. [An approach to
resolving the problem of changes in output due to changes in fluid composition, in a mass flow sensor
based on any type of thermal anemometry.]
Bonne, U. (1992) Fully compensated flow microsensor for electronic gas metering. Proceedings of the
International Gas Research Conference, III, 859, Orlando, FL. [A fluid-property-based approach for
compensating for composition, temperature, and pressure sensitivities for thermal flow sensors.]
Bonne, U and Kubisiak, D. (2001) "Actuation-Based Microsensors," J. Smart Materials & Structures,
vol. 10, p. 1185. [A fluid motion-based approach for compensating for composition, temperature, and
pressure sensitivities for any type of flow sensors. See Fig. 6]
Bonne, U, Detry, J., Higashi, R.E., Rezacheck, T. and Swanson, S. (2002) New gas composition and
trace contaminant sensors. Proceedings of the GTI Gas Technology Conference, Orlando, Forida, 30 Sep.
- 2 Oct.. [Describes application of phased heater array structure for enhanced detection (PHASED) sensor
technology with application to gas composition as well as other actuation-based sensors.]
Bonne, U. (1996) Sensing fuel properties with thermal microsensors. Proc. SPIE Conference on Smart
Electronics and MEMS, paper no. 2722-24, San Diego, CA. [Description of a particular temperature-
compensated flow sensor including the nonlinear function used for the compensation.]

S
Desvergne, J.P. and Czarnik, A.W. (eds.) (1997) Chemosensors of Ion and Molecule Recognition. Kluwer
Academic Publishers, Dordrecht, The Netherlands. [This edited volume contains contributions that
LS

R
include a description of fluorescent chemosensor techniques for single-molecule detection and the use of
chemical sensors based on conjugated polymers.]

TE
Fanger P.O. (1970) Thermal Comfort. McGraw-Hill Book Co., New York. [A comprehensive discussion
O
of thermal comfort, including the Predictive Mean Vote (PMV) formula.]
Frank, R. (1996) Understanding Smart Sensors. Artech House, Boston, MA. [Reviews the principles
-E

AP
behind and developments in the embedding of “intelligence” in sensor systems. Also presents a
classification of different sensor types.]
O

Honeywell Inc. (1998) Shutter-check flame detectors, models C7012E, C7012D, C7024E, C7024F,
C7061A, C7076A and C7076D. Tradeline Catalog, Publ. No. 70-6910, Honeywell Home and Building
H
SC

Control, Golden Valley, MN, 753–763. [ “Fail-safe” systems with optomechanical sensors for flame
detection.]
C

ISO 7730 (1994) Moderate Thermal Environments – Determination of the PMV and PPD Indices and
Specification of the Conditions for Thermal Comfort (2nd ed.). International Standards Organization,
E

Geneva, ref. no. ISO 7730:1994(E). [Adoption of Predicted Mean Vote as an international standard for
thermal comfort of human occupants.]
N

PL

Kroot, P. (1996) A new oxygen sensor for cleaner, more efficient combustion. Scientific Honeyweller, 15,
29–33. [A sophisticated self-checking oxygen sensor.]
U

Liu, C.H., Barzilai, A.M., Reynolds, J.K., Partridge, A., Kenny, T.W., Grade, J.D. and Rockstad, H.K.
M

(1998) Characterization of a high-sensitivity micromachined tunneling accelerometer with micro-g


resolution, Journal of Microelectromechanical Systems, 7, 235. [A force-rebalanced accelerometer used
SA

to eliminate the effects of zero drift and nonlinearity in a sensor.]


Margulis, L. and Sagan, D. (1997) Microcosmos. University of California Press. [A view of evolution
that emphasizes the interconnectedness of all life on our planet, drawing on insights from microbiology.]
Medtronic (1997) Medtronic rapid technological growth. http://www.medtronic.com/corporate/ early.htm.
[A brief history of Medtronic, Inc., a company specializing in biomedical devices. Modern devices may
be based on sensing of properties such as blood sugar and body motion.]
Middelhoek, S. and Audet, S.A. (1989) Silicon Sensors. Academic Press, London. [An excellent review
of modern silicon-based sensor technology. A classification scheme for sensors is also detailed.]
Nagle, H.T., Schiffman, S.S. and Gutierrez-Osuna, R. (1998) The how and why of electronic noses. IEEE
Spectrum, 35, 9, 22–31. [This article reviews different technological approaches to the development of
artificial nose sensors and describes several companies who are developing products and solutions in this
area.]

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

Pierce, J.R. (1980) An Introduction to Information Theory – Symbols, Signals and Noise. (2nd rev ed.).
Dover Publications, Inc., New York. [An accessible treatment of information theory.]
Shannon, C. (1948) A mathematical theory of communication. Bell System Technical Journal, 27, 379–
423, 623–656. [The seminal papers that established information theory. Shannon’s contributions
included relating energy and information content.]
Soloman, S. (1994) Sensors and Control Systems in Manufacturing. McGraw-Hill, New York. [A review
of sensors for manufacturing applications along with their integration into automation and control
systems.]
Suga, N. (1990) Cortical computational maps for auditory imaging. Neural Networks, 3, 3–21. [The
processing in the bat cortex of returned sonar signals.]
Swager, T.M. (1997) New approaches to sensory materials: molecular recognition in conjugated
polymers, in Chemosensors of Ion and Molecule Recognition (eds. J.P. Desvergne and A.W. Czarnik).
Kluwer Academic Publishers, Dordrecht, The Netherlands. [New chemosensors for TNT with parts-per-
billion sensitivity.]
Travis, J. (1999) Making sense of scents: scientists begin to decipher the alphabet of odors. Science News,
155, 236–238. [An article that highlights advances in our understanding of olfaction.]

S
Van der Ziel, A. (1976) Noise in Measurements. John Wiley & Sons, Inc., Chichester, U.K. [Discusses
LS
sources of and types of noise in measurement and instrumentation systems.]

R
Wolffenbuttel, R.F. (ed.) (1996) Silicon Sensors and Circuits. Chapman and Hall, London. [An edited
volume with contributions describing a variety of silicon-based sensor technology.]

TE
O
Yamatake (1999) Integrated environmental comfort control. http://www.yamatake.co.jp/innovat/
iecc.htm. [Products and technology from one manufacturer in the area of comfort control for indoor
-E

environments, based on the PMV standard.]


AP
Biographical Sketches
O

David Zook is currently Chief Technology Officer with Dresden Associates, a small consulting firm in
H
SC

the Minneapolis–St. Paul metropolitan area (U.S.A.). He received his Ph.D. in Solid State Physics from
the University of Wisconsin–Madison in 1962. He spent 38 years with Honeywell developing
C

mechanical, optical and chemical sensors and systems for commercial and military applications and
recently retired from Honeywell Laboratories as a Principal Research Fellow. He has authored over 20
patents and 60 refereed papers and twice won the Sweatt Award (Honeywell’s highest technical
E

achievement award). He has experience with working over a wide range of technology areas, including
N

electronic and photonic device physics, optical system design, plasma–surface interactions and micro-
PL

electro-mechanical systems (MEMS). His most recent work involved MEMS-based fiber optic pressure
sensors that are self-resonant such that unmodulated light incident on them is reflected as modulated light
U

with the modulation frequency dependent on pressure. He has also been involved with sensors and
M

actuators for chemical and biological sensing.

Ulrich Bonne is a Senior Research Fellow at Honeywell Laboratories in Plymouth, Minnesota, U.S.A.
SA

He received a B.S. degree in Physics at the University of Freiburg i. Br., Germany, in 1957; and his M.S.
(Physics, 1960) and Ph.D. (Chemical Physics, 1964) degrees from the University of Göttingen, where he
was a Research Associate in the Physical Chemistry Department at the University until joining
Honeywell Laboratories in 1965. His current R&D interests and activities include micro sensors and
actuators for monitoring air quality and water quality for DoD, DoE, EPA, aerospace, process industry,
combustion engines, and appliance applications; non-intrusive approaches for sensing and controlling
micro-flows; low-cost sensing of fuel oxygen demand; and sensor self-calibration and actuation-based
sensors. His earlier work was in the areas of flame ionization, PZT transformers, liquid crystal displays,
and IR/visible/UV spectroscopy in low-pressure flat flames. Dr. Bonne is the author or co-author of some
130 technical publications and 60 U.S. patents. He is the recipient of several technical Honeywell awards:
an H.W. Sweatt Award (1971) for development of combustion instrumentation; a Honeywell “Pioneer”
Award (1990) for electronic gas meter development; Product Development Team Awards for a Gasoline
Vapor Sensor (1997) and a Breath Flow Sensor (2001); and a Minnesota Society of Professional

©Encyclopedia of Life Support Systems (EOLSS)


CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION – Vol.XXI - Sensors in Control Systems - David Zook, Ulrich Bonne
and Tariq Samad

Engineers “One of the Seven Wonders of Engineering” Team Award for the development of an advanced
heat pump control in 1979.

Tariq Samad is a Corporate Fellow with Honeywell Automation and Control Solutions in Minneapolis,
U.S.A.. He has been with various R&D organizations in Honeywell for 17 years, and he has been
involved in a number of projects that have explored applications of intelligent systems and intelligent
control technologies to domains such as autonomous aircraft, building and facility management, power
system security, and process monitoring and control. He was the editor-in-chief of IEEE Control Systems
Magazine. He is the author or coauthor of about 100 publications and holds 11 U.S. patents. His recent
publications include the edited volumes Automation, Control, and Complexity: An Integrated View
(Wiley), Perspectives in Control: Technologies, Applications, New Directions (IEEE Press), and
Software-Enabled Control: Information Technology for Dynamical Systems (Wiley, in press). Dr.
Samad received his B.S. in Engineering and Applied Science from Yale University and his M.S. and
Ph.D. in Electrical and Computer Engineering from Carnegie Mellon University. He is the recipient of an
IEEE Third Millennium Medal and corecipient of an “Excellence” award for Control Systems Magazine
from the Society for Technical Communications, New York Metro Chapter.

S
LS

R
TE
O
-E

AP
O

H
SC

C
E

E
N

PL
U

M
SA

©Encyclopedia of Life Support Systems (EOLSS)