© 2020 University of South Africa

All rights reserved

Printed and published by the

University of South Africa
Muckleneuk, Pretoria

PRC3601/1/2021

70716226

InDesign
Florida

PR_Tour_Style
 Page
1
1
1.1 Introduction, outcome and assessment criteria 1
1.2 Basic concepts 1
1.2.1 Chemical process 1
1.2.2 Process control systems 1
1.3 The purpose of control systems 1
1.3.1 Minimising process variability 1
1.3.2 Maximising process efficiency 1
1.3.3 Process safety 1
1.3.4 Profit 1
1.4 The process control loop 2
1.5 Instrumentation 4
1.5.1 Instrumentation for pressure measurement 1
1.5.2 Instumentation for temperature measurement 1
1.5.3 Instrumentation for level measurement 1
1.5.4 Instrumentation for chemical composition measurement 1
1.6 Control system documentation 7
1.6.1 Process flow diagram 1
1.6.2 Piping and instrumentation diagram 1
1.7 Piping and connections 9
1.8 Summary 9
1.9 References 1
1.10 Problems
11
2.1 Introduction outcome and assessment criteria 11
2.2 Basic role of a control system 11
2.3 Control system terminology 11
2.4.1 Open-loop control 1
2.4.2 Closed-loop control 1
2.4.3 Combined feedforward and feedback control 1
2.5 Control systems 13
2.5.1 Manual control systems 1
2.5.2 Automated control systems 1
2.6 Controllers 14
2.6.1 ON/OFF controllers (discrete control systems) 1
2.6.2 Programmable logic controllers (PLCs) 1
2.6.3 Distrubted control systems (DCSs) 1
2.6.4 Analogue controllers 1
2.6.5 Digital controllers 1
2.6.6 Emergency shutdown controllers (ESDs) 1
2.7 Communication signals for controllers 1
2.7.1 Analogue signals 1
2.7.2 Discrete signals 1
2.7.3 Digital signals 1
2.8 Advantages and disadvantages of communication signals 15
2.9 Digital communication standards 16
2.9.1 HART communication 1
2.9.2 Fieldbus 1
2.9.3 Ethernet 1
2.9.4 Qireless communication 1
2.10 Transfer function 16
2.10.1 The algebra of a process block diagram 1
2.10.2 The summing point types of a closed control loop 1
2.10.3 Algebraic operations of control loop block diagrams 1
2.11 Transfer function for open-loop control systems 17
2.12 Transfer function for closed-loop control systems 17
2.13 Multi-loop/closed-loop systems 17
2.14 Summary 1
2.15 References 1
2.16 Problems 1
18
3.1 Introduction, outcome and assessment criteria 18
3.2 Process controller algorithms 18
3.2.1 Discrete controllers 1
3.2.2 Multi-step controllers 1
3.2.3 Continuous controllers 1
3.3 Controller modes 18
3.3.1 Proportional control mode 1
3.3.2 Integral control mode 1
3.3.3 Derivative control mode 1
3.3.4 Proportional + integral (PI) control mode 1
3.3.5 Proportional + derivative (PD) control mode 1
3.3.6 Proportional + integral + derivative (PID) control mode 1
3.4 Digital implementation of the PID algorithm 19
3.5 Process controller selection 20
3.6 Control loop performance 20
3.6.1 Definition of control quality 1
3.6.2 Measurement of quality 1
3.7 Control loop performance monitoring 20
3.7.1 Control loop diagnostics 1
3.7.2 Control loop stability 1
3.8 Control system tuning 21
3.8.1 Closed loop tuning: Ziegler-Nichols 1
3.8.2 Open-loop tuning 1
3.8.3 Lambda or IMC tuning 1
3.8.4 Auto-tuning 1
3.8.5 Trial and error tuning 1
3.9 Summary 21
3.10 References 22
3.11 Problems 23
 26
4.1 Introduction, outcome and assessment criteria 26
4.2 Control loop classification 26
4.2.1 Single variable control loops 1
4.2.2 Multi-variable control loops 1
4.3 Feedback and feedforward conrol systems 26
4.3.1 Feedback control systems 1
4.3.2 Feedforward control systems 1
4.3.3 Feedforward + feedback control systems 1
4.4 Advanced process control loops 26
4.4.1 Cascade control loops 27
4.4.2 Batch control systems 27
4.4.3 Ratio control 27
4.4.4 Selective control 28
4.4.5 Fuzzy control 28
4.5 Final control operation 30
4.5.1 Controller signal conversion 1
4.5.2 Actuators 1
4.5.3 The final control element 1
4.5.4 General operation of control valves 1
4.5.5 Types of control valves 1
4.6 Globe valve 32
4.6.1 Fail-safe control design 1
4.6.2 Fail-safe control valves 1
4.6.3 Control characteristics 1
4.7 Valve sizing and selection 32
4.7.1 The valve flow equation 1
4.7.2 Algorithm for valve sizing and selection 1
4.7.3 Case study 1
4.8 Summary 1
4.9 References 1
4.10 Problems 1
Process control engineering is a feature of a number of engineering disciplines,
such as chemical, electrical, and mechanical engineering. It is also applied in a
wide range of physical systems, from electrical circuits to guided missiles and
robots. This engineering science involves basic principles particularly useful to
chemical engineers when applied to physicochemical systems such as chemical
reactors, heat exchangers, and mass transfer equipment [10]. A typical chemical
plant is a complex combination of various processing units (distillation columns,
reactors, pumps, absorbers and heat exchangers, etc.) arranged systematically
in order to operate efficiently and effectively (see figure 1).

During the operation of a chemical plant, steady state is never achieved

because all key measurements change continuously, sometimes displaying

external influences such as variability in demand or supply. Thus, all chemical

processes are dynamic in nature, meaning that the process behaviour changes
constantly with time. In the presence of these ever-changing internal and/or
external influences, the plant must satisfy a variety of requirements [6; 9]. I
will discuss some of these below.

The operation of a plant in general needs to be economically viable in terms of

the utilisation of raw materials, capital, energy and human labour. Therefore,
the required operating conditions should be maintained at all times to ensure
maximum profits and minimum operating costs [6; 9].

A chemical processing plant needs to sustain or surpass its production capacity

and deliver final products of consistent quality [6].
Each of the various types of equipment used in a chemical plant has inherent
operating limitations imposed by its particular design. The operational constraints
of all the equipment have to be taken into account when operating the plant [6].

Operating a chemical plant safely is very important for ensuring the well-being
of the plant workers, the surrounding community and the economic viability of
the overall process. Consequently, maintaining process variables (temperature,
concentrations of chemicals, pressure etc.) within recommended limits is a
primary requirement during plant operation [6].

To preserve the environment and ensure public health, laws governing the
chemical composition and rate of emissions from chemical plants must be
respected at all times.

The key aspects required for a chemical plant to operate successfully are
summarised in figure 2 below [6].

The requirements listed above mean that the process variables of the plant
have to be constantly monitored. This is accomplished through process control
and instrumentation, which I will say a bit more about below [6].

5.1 Process control engineering is the engineering science that ensures

sustained maintenance of the plant at desired conditions by constantly providing
the proper response to counteract ever-occurring disturbances responsible
for the dynamic behaviour of the chemical plant. This engineering practice
contributes significantly to safety, environmental impact minimisation, product
quality and plant profitability [6].

5.2 Instrumentation is the science of using instruments or their applications

for the purpose of observation, measurement and control. Effective monitoring
and control of plant operations relies extensively on the effectiveness of its
instrumentation system [6].

Your next question might well be: What is process control? Also, what does
instrumentation entail?

This study guide is divided into four chapters, which deal with the fundamentals
of process control and instrumentation.
To facilitate the learning process and to provide appropriate and simple practical
examples, basic chemical engineering concepts are generally introduced
through steady-state physical systems. However, when we observe a chemical

boiling and viscous material being extruded, with process parameters changing
continuously. This shows that chemical processes are dynamic and that
regardless of their size or complexity, they exhibit transient behaviour – and
this is the key reason why process control is necessary [6].

This chapter presents an introduction to process control and instrumentation

concepts and the elements of process control and instrumentation systems.
After working through this chapter and the example problems and chapter
problems, you should be able to:

• Define the concepts of chemical process and process control systems

• Identify different types of chemical processes
• Explain the purpose of effective process control systems in terms of variability,
efficiency, safety and profit
• Define a basic process control loop
• Describe the main components of a basic control loop and their respective
control functions
• Identify the key factors to consider when selecting appropriate instrumentation
• Explain the techniques used for measuring temperature, pressure, flow,
level and chemical composition
• Distinguish between a process flow diagram (PFD) and a process and
instrumentation diagram (P&ID) and explain the professional uses of these
two types of diagrams

In the chemical industry, the term “process” refers to the methods or means
of refining or converting raw materials into desired end products. During the
process, energy is provided from an external source and raw materials are used
in either a liquid, gaseous, or slurry (a mix of solids and liquids) state and can
be measured, mixed, cooled, heated, filtered, transferred, handled or stored
so as to produce the desired end product (see figures 1.1 and 1.2) [6; 7].

There are basically three types of chemical processes. These are:

• Batch processes,
• Continuous processes and
• Discrete manufacturing processes [7].

(a) Batch process – In a batch process, a fixed amount of raw material is

isolated and subjected to physical and/or chemical modification. In many
instances the modified material is then subjected to another processing
step, or a number of subsequent steps. Many repeats of this process,
 perhaps using different equipment, may be necessary to obtain the end
product. The manufacturing of beer is a good example of a typical batch
process [7].
(b) Continuous process - In a continuous process, the raw material is fed
into the process unit and the resulting product is removed from the unit
at the same time. The petroleum refining process can be classified as a
continuous process [7].
(c) Discrete manufacturing process - In a discrete manufacturing process,
individual (discrete) items (e.g. electronic devices or vehicles) are produced.
During the process separate components, parts or sub-assemblies
are manufactured or assembled to produce a final product. Vehicle
manufacturing is an example of a discrete manufacturing process [7].

The chemical processing industry includes the production, generation,

manufacture, and/or treatment of oil, gas, wood, metals, food, plastics,
petrochemicals, chemicals, steam, electric power, pharmaceuticals and waste
materials. The main characteristics of each type of chemical manufacturing
process are set out in table 1.1 below [6; 7].

We use the word “control” a lot in everyday conversation. We could be talking

about controlling our anger when dealing with a frustrating crisis, or we could
be looking for the remote control for the TV or the air conditioner [6; 7].

Process control in chemical processing plants refers to the techniques or

technologies used to regulate and maintain process variables near the operating
constraints to obtain the desired final product. Therefore, the fluctuation in
manufacturing parameters such as the temperature and pressure required to
handle raw materials, the proportion of one ingredient to another, and the
mixing speed of ingredients can significantly influence the quality of the final
product [7].

The operation of a chemical plant involves combining, refining, handling and

manipulating compounds of various phases to safely and profitably produce
a desired final product. Each chemical process exhibits a specific dynamic
behaviour that governs the transformation of raw material to end product. This
dynamic behaviour is directly related to the changes in the physical properties
of the inputs, the source of energy to the process and the process itself. These
processes are generally precise, demanding and potentially harmful to the
environment and public health. Moreover, even small disturbances can have
a considerable impact on the final product [6].

An example of a control system encountered in everyday life is a household

cooling system, shown in figure 1.3. The system keeps the house warm in
extremely cold weather by producing hot air using a furnace. A thermostat
measures the temperature in the room, which is then compared with the desired
temperature range, referred to as the set point. If the temperature is below the
set point, the heating mechanism (furnace) is turned on. If the temperature is
above the set point, the heating mechanism is turned off. If the temperature
is within the desired range, the heating mechanism status is maintained [6].

The deviations in process parameters (flow, concentration, temperature, etc.)

and the effect of other external factors must be consistently and effectively
controlled so as to maintain the desired specifications of the end product
while minimising raw materials and energy consumption. Process control and
instrumentation technologies are the tools used by chemical engineers to keep
the manufacturing operations running within specified limits and to set more
accurate limits. By instrumentation technologies, I mean devices and their
applications for monitoring and controlling chemical processes. You will find
out more about instrumentation in the next section. Returning to chemical
control systems, the effective control of manufacturing processes is important
for four reasons [7; 11]:

• Minimisation of process variability

• Maximisation of process efficiency
• Process safety
• Profit

Process control aims to reduce variability in the final product, which guarantees
a consistently high-quality end product. Without control, end product quality
will vary. As a result, the plant will manufacture either a final product that is
of a higher grade and more expensive than its market value, or a lower-grade
final product for a higher market price (see figures 1.4 and figure 1.5) [6].

Reducing end product variability can prevent the need for product padding to
meet required product specifications. Padding is the process of upgrading end
product quality to meet market specifications. Variability in the final product
(i.e., poor process control) forces manufacturers to pad the product to ensure
that specifications are met, which adds to the production cost. With accurate,
dependable process control, the set point (desired or optimal point) can be
moved closer to the actual product specification and thus minimise total
production cost. For instance, in blending and batching operations, control
systems are central to maintaining the proper ratio of ingredients to deliver a
product of desired quality consistently. For example, let’s consider the process
of cobalt production. Cobalt is a mineral used in several industries, including
the battery industry and the electroplating industry [6; 7].

The cobalt production process takes place in mines extracting cobalt-containing

ore. The final product, cobalt, generally needs to have a certain percentage
humidity. If there is no effective control system to ensure that the quality of
the process operation is up to standard, it is quite possible at the end of the
production to obtain a batch of cobalt with the wrong moisture content.
Because discarding the production batch would be a huge loss to the company,
the company will have to lower the price of the product obtained, which will
also result in production loss [6].

For maximum efficiency, some chemical processing plants have to be operated

at specific conditions. For example, a chemical process involving a batch reactor
in which 100% conversion of the raw material is achieved could have as control
point precise pressure and temperature values. If there is poor control and the
temperature and pressure variables are not maintained at their desired values,
a less than 100% conversion of the raw material will occur. This means that
the process will not be operating at maximum efficiency. Accurate control of
the process conditions (pressure and temperature) ensures maximum process
efficiency and minimises raw material consumption. For example, let’s consider
the process of brewing beer. Temperature is a very important factor to be
controlled at every step of this process, from mashing to fermentation. The
mashing stage involves mixing grains and malt with hot water for fermentable
sugar extraction. If the mixture is too cool, the enzymes will not successfully
extract the sugar from the starch, whereas very high temperature will result in
the final product (beer) having an unpleasant flavour because the grains in the
mash will have been burnt. Along the subsequent stages of the manufacturing
process, temperature plays a vital role in the foam stability of the finished
product (beer) [6].

Precise process control is necessary mainly to ensure the safety of the staff,
environmental safety and equipment safety. The safety of the workers and the
surrounding community is of primary concern in the operation of any chemical
plant. The consequences of poor control of all of the process variables during
any chemical process operation can be catastrophic. For example, maintaining
proper boiler pressure is very important to prevent the boiler from exploding,
as this could endanger the staff. Examples of safety equipment incorporated
into process control systems include temperature switches, pressure switches
and pressure relief valves [6].

Once product quality and safety requirements have been met, the objectives
of the control system can then be focused on profit. Profit is made when
product quality is not compromised, and that means operating the chemical
plant within the process constraints. The more closely to these constraints the
process is operated, the higher the profit. However, the most difficult element
of process control is maintaining the operation of the chemical plant near the
constraints safely without compromising the quality of the end product [6].

•
•
•
•
•
•

The process control loop is an essential part of an industrial control system.

It consists of all the physical components and control functions necessary to
automatically adjust the value of a measured process variable (PV) to equal
the value of a desired set point (SP) (see figure 1.6) [6; 7].
The characteristic components of a typical control loop are classified and
defined in table 1.2 [6; 7].
Instrumentation refers to a collection of devices, hardware, instruments or
functions or their applications for the purpose of measuring, monitoring or
controlling industrial equipment, an industrial process, or any combination
of the two. It is not possible to separate the concepts of process control and
instrumentation. Process control requires the measurement of key process
parameters in order to ensure that they are maintained at desired values. It is
impossible to regulate or control a process parameter that cannot be measured.
Instruments are the foundation of control systems; they measure the controlled
process variable, and convert the measurement obtained into the corresponding
signal, which is then transmitted to the controller device. The process variables
that are usually regulated during the operation of a chemical processing plant
are: mass, flow, temperature, level and pressure. Consequently, a successful
control system requires appropriate instrumentation. Below I have put together
a list for you of the most commonly used instruments and descriptions of their
inherent functionalities [3; 8; 11].
(a) Sensors – these are primary devices with the inherent properties of
changing in a predictable manner when exposed to the stimulus of the
process variable they were designed to detect. Therefore, the value of the
process variable is obtained by measuring the response of the sensor to the
process. Sensors are the first element in the control loop, and examples of
sensors include: resistance temperature detectors, annubar flow elements,
orifice plates, thermocouples, pressure sensing diaphragms, ultrasonic
emitters and receivers, strain gauges and venture tubes [3].
(b) Transducers – these are devices that convert energy from one form into
another. They convert the output of a sensor into an appropriate signal
that can be used by the controller [3].
The sensor and transducer can be part of a single instrument (see figure 1.7),
or else the transducer can be packaged together with the controller device
(see figure 1.8) [3].
(c) Signal conditioning – this is the processing signal applied to the output of
the transducer. It could sometimes refer to a simple amplification, but it
usually refers to a more complex scenario. Nowadays, modern instruments
have very complicated signal processing built into them to detect the
possible phase shifts in the motion of the sensing elements [3].

The following characteristics of an instrument are important factors to consider

when selecting an instrument, and could also have a significant impact on the
effectiveness of the process control system [3; 8]:

(a) The range – this consists of the lowest and highest values a sensor can
measure within its specification. For example, a temperature sensor can
have a measuring range of –260 °C to +600 °C.

For most instruments, an operating range larger than what is required

reduces the accuracy of the measurement [3].

(b) The span – this is the high end of the range minus the low end of the
range of a sensor. For example, a sensor with a range of –260 °C to
+600 °C would have a span of 860 °C.

An instrument with an excessively wide operating span will lower the

process gain [3].

(c) The resolution – this is the smallest amount of input signal change an
instrument can detect effectively [3].
(d) Accuracy – this describes how close the measurement approaches the
true value of the measured process variable. Accuracy of a measurement
is generally expressed as a percentage error (% error) or an absolute error
over a range [3].

For example, the manufacturer of a pressure instrument specifies that the

instrument has an accuracy of ± 4% of its full scale, and the full scale of
the instrument is 600 Pa.

Accuracy = 0.4% 600 Pa = 0.004 600 Pa = 2.4 Pa

Therefore, the measurement signal from this instrument would be accurate

to 2.4 Pa for all pressures within the device specifications [3].
 The accuracy of the instrument could also be specified as absolute of
± 2.4 over its full operating range. It is preferable to select sensors that
are more accurate than the degree to which it is desired to control the
processing plant [3].

(e) Precision – this is the degree of reproducibility with which repeated

measurements can be obtained under identical process conditions. It is
also referred to as drift or stability of the instrument. Precision is always
an essential inherent property of the instrument needed for good control,
even when accuracy is not required. It is the most important characteristic
of an instrument [8].
(f) Repeatability – this is the same thing as precision for most instruments.
Some sensors, however, exhibit a condition known as hysteresis. In
these sensors, the state of the variable prior to the measurement action
has a significant effect on the measurement value obtained. This usually
happens in systems involving some sort of mechanical element as part
of the sensing system. Bourdon pressure sensors, for instance, are prone
to hysteresis; the measurement these sensors produce will be different
depending on whether the pressure was falling or rising immediately prior
to the measurement action [8].
(g) Drift – this refers to a short- or long-term effect that causes some instruments
to lose their accuracy. It could also possibly lead to the instrument losing
its precision. For example, the corrosion of a thermocouple will deteriorate
its thermoelectric properties, and consequently alter the voltage produced
at a precise temperature [8].
(h) Delay – Sometimes, instruments take time to produce a measurement
after the sampling process. Chemical analysers are good examples of
this, as are automated analytical laboratories that take several minutes to
produce an output value. When it comes to process control, any delay
in the production of measurements can cause problems for the control
system, because the controller would be analysing the process as it was
in the past rather than what it actually is at present. The speed at which
an instrument responds is another very important factor to consider. A
thermometer sensor exhibiting a large resistance will take much longer to
respond to temperature change than a small thermistor, simply because
of the relative thermal capacities [8].
(i) Instrument calibration – The calibration of an instrument of measurement
refers to the process of making two or more standard measurements
in order to match the quantity being measured with the zero and the
span of the instrument to adjust it. For example, various standard buffer
solutions would generally be used to calibrate a pH-meter. Calibration can
sometimes be a difficult and time-consuming process. Nowadays most
packaged instruments such as pressure instruments are already factory
calibrated, and can be directly installed into the process. However, certain
instruments such as pH-meters and orifice plates require calibration in
the plant after installation. pH-meters regularly experience drift, and so
they need to be calibrated on a routine basis [8].
(j) Linearity – Most commercial control systems operate using linear control
algorithms and work best when controlling linear or nearly linear systems.
A linear system is one where doubling a change in the input will result
in double effect on the output. Transducers such as orifice plates and
 thermistors do not generate outputs that are linearly related to the value
being measured. In modern instruments, such signals will be linearized
internally in the instrument. However, high transducer non-linearity will
reduce the instrument sensitivity in some areas [8].
(k) Instrument dynamics – Instruments, like processes, also have dynamic
properties. These properties contribute to the process dynamics that the
controller detects. These include gain, rise time and dead time [8].

The gain of an instrument, also known as the sensitivity of a sensor, is the

ratio of the output signal to the change in process variable [8].

For example, a thermocouple with sensitivity of 3 mV per °C will change

its output by 3 mV for every 0C change in the measure process variable [8].

The rise time of an instrument refers to the time it takes for the instrument
to generate a signal representing 100% of the value of the process variable
it is measuring after a step change in the variable. The instrument dead
time is the time period required for an instrument to start reacting to
process change [8].

(l) Noise – This is a random fluctuation appearing in the measurement

signal. It can either be produced by the process itself, or it can arise in
the communication lines or in the measuring instrument. For example,
the presence of air bubbles in a flow being measured by an orifice plate
can generate noise in the measuring signal [8].

Where instruments generate significant noise, it would be better for

accuracy and precision purposes to choose the instrument that produces
the least noise [8].

(m) Reliability – It is vital that all instruments installed in a chemical plant be

highly reliable, because the measurements generated by these instruments
are used for control and to ensure safe operation of the process [8].
Generally, reliability specifications are provided by the instrument’s
manufacturer as the mean time between failures (MTBF). These
specifications are provided on the instrument data sheet and can also
be incorporated into a hazard analysis to determine whether or not it is
necessary to perform instrument redundancy [8].
(n) Cost – In the selection of an instrument, its overall lifetime cost is a crucial
element to factor in. Low-cost instruments are not necessarily a better
choice, because a low purchase price might not be the best indication
of a good instrument. A more expensive instrument might last longer,
perform better and require less frequent calibration [8].

New instruments have various other characteristics that could also be considered
when needed. These include averaging or storing measurement values. These
features should only be considered when they are required for efficient control
of the process plant; in other cases, they are generally ignored when selecting
an instrument. In chemical processing plants, the most common measurements
are pressure, temperature, flow and level [8].
In all chemical processing plants pressure measurement, either as a fundamental
measurement or as an implied measurement of level and flowrate, plays
a significant role in the efficient operation of the process plant. Pressure
measurements can be divided into four main categories [4; 5]:

(a) Gauge pressure – This is the pressure measured above the local atmospheric
pressure which generally changes according to weather conditions
and altitude. Therefore, if the pressure measurement instrument is not
connected to a pressure source, the pressure reading will be zero. When
the instrument is configured appropriately, the pressure measurements
obtained below the local atmospheric pressure are read negatively [4].
(b) Absolute pressure – The pressure reading obtained in this case is known
as absolute pressure measurement. Except in a complete vacuum
environment, the measuring instrument will always generate a reading.
In this case, pressure measured below the atmospheric pressure will be
read as positive absolute pressure [4].

Absolute pressure and gauge pressure are related through the following
mathematical expression:

Pabsolute = Pgauge + Patmospheric

For example, an instrument reading a gauge pressure of 92 000 Pa at

a location where the atmospheric pressure is equal to 101 325 Pa is
generating the measurement of an absolute pressure of 92 000 Pa +
101 325 Pa, which is equal to 193 325 Pa [4].

(c) Vacuum – These measurements are usually done in vacuum systems.

Here the pressure below the local atmospheric pressure is measured. For
example: At a location where the atmospheric pressure is equal to 11.8 Pa,
an absolute pressure of 10.8 Pa will be measured as the equivalent of –
1 Pa on a gauge pressure measurement instrument and as the equivalent
of 1 Pa vacuum on a vacuum measuring instrument [4].
(d) Differential pressure – A differential pressure measuring instrument has
two measuring ports and reads the pressure difference between the two
ports. This type of pressure measurement device can also be used as part
of level and flow measurement systems [4].

In the past, Bourdon tubes were the basis of pressure gauges commonly used
in chemical plants (see figure 1.9). In these devices, the sensing element is a
simple coiled metal tube. The measuring process consists of displacing the
pointer on the gauge by an acting force generated by the increase in pressure
inside the tube. Nowadays these instruments are still used in portable devices
such as car tyre inflators. Even though these instruments are cheap, they are
not commonly used in modern chemical plants because the transmission of the
readings obtained to remote locations such as control rooms is quite difficult [4].
Recent pressure sensor devices come in a package including the sensor, signal
conditioning and a transmitter. Of all the types of pressure sensors, the following
two are the most popular [4; 5]:

(a) Capacitance: These devices consist of an electrical capacitor formed by

a diaphragm that is exposed to the pressure to be measured. Basically,
the pressure is measured by the change in capacitance caused by the
deformation of the diaphragm (see figure 1.10) [4; 5].

(b) Piezoresistive: In these instruments, a diaphragm or semi-conductor

material is the sensing element exposed to the pressure to be measured.
The pressure is measured by deforming the diaphragm; this movement
creates a resistance change in the material, which is detected as the reading
of the pressure (see figure 1.11) [4; 5].

Nowadays there are a number of innovative pressure sensor instruments

for all types of pressure measurement applications. For differential pressure
measurement, the sensor devices are known as differential pressure cells [4].
In most chemical plants, where numerous chemical reactions are occurring
simultaneously, temperature is one of the most important variables that need
to be measured regularly for efficient control of the plant process [5].

(a) Thermocouples

A thermocouple is a sensor instrument that consists of two wires of dissimilar

thermoelectric properties (heat generates electrons at different extents) [5].

The two wires are joined at each end. During temperature measurement,
a small voltage generated by the feedback effect and proportional to the
difference between the temperatures at the two ends of the thermocouple
device is detected (see figures 1.12 and 1.13) [5].

In the past, the cold end of the thermocouple was maintained at a fixed
temperature by keeping it in an ice or water bath at all times. In the new
thermocouple devices, the cold end is simulated using an electronic chip made
of a thermistor and complex electronics. This complex electronic system also
serves to intensify the mV output of the thermocouple to generate something
more useful in the rest of the instrument [5].

The most recent industrial thermocouples are available as a package unit

containing a transmitter.
There are numerous types of thermocouples, depending on the temperature
range they cover and their applications [4; 5].

(b) Resistance temperature detectors (RTDs)

Originally, thermocouples used to be the most commonly used temperature

measuring instruments in chemical processing plants. Recently, RTDs have come
to be preferred to thermocouples (see figure 1.14) [5].

An RTD, also known as a resistance thermometer, is a modern temperature

measuring device that has a sensor whose resistance changes with temperature.
RTDs with platinum or platinum alloy sensors are very popular in process plants
because of their stability. These instruments behave like conductors, as their
resistance increases with increasing temperature. The sensor device in platinum
RTDs only produces a change in electrical resistance, which is then measured
by a bridge circuit. RTDs are not easily obtained as a single-piece instrument.
They are often made as a package instrument comprising a sensor, and signal
conditioning and transmitter devices. Because the sensor device in an RTD is
bigger than the sensor element present in a thermocouple, RTDs are a bit slower
to generate a measurement, and are more expensive. Nevertheless, platinum
RTDs have been acknowledged as being more accurate and significantly
more stable than thermocouples. This makes them the preferred option for
temperature measurement between –200 and 500 °C [5].
(c) Thermistors

Thermistors, on the other hand, are semi-conductor-based temperature sensors

whose resistance decreases with increasing temperature. However, the change
in the resistance of these devices is significantly less linear than the change
registered in the platinum-based RTDs. In addition, the overall accuracy of the
temperature measurement is very much lower (see figure 1.15) [5].

Because thermistors can come in very small sizes and are able to measure
temperature very fast, they are nowadays used more often when rapid
temperature measurement is needed. For example, thermistors can be used
effectively to detect vortices in some vortex shedding flowmeters [5].

Controlling the amount of liquids by constantly measuring their levels is a vital

procedure in chemical and manufacturing plants. The main purpose of this
procedure is to monitor the mass balance of the overall processing plant while
keeping liquid at safe recommended levels in order to avoid the presence of
vapours where these not required. The level of a liquid can be measured in
more than one way [5].

(a) Differential pressure measurement

The differential pressure measuring instrument is a very common device used

to measure the level of a liquid (see figures 1.16 and 1.17) [5].

One end of the device is connected to the bottom of the container of the liquid
via a nearby tapping. This end is referred to as the high-pressure connection.
The other end is connected to a tapping at the top and, most important, in the
vapour space above the liquid. Because the differential pressure obtained is
actually the measurement of the hydrostatic head in the container, knowing
the density will enable you to calculate the level of the liquid in the vessel [5].
The disadvantage of the use of this technique is revealed when the liquid either
contains solid particles or is likely to form solid particles. This has to do with the
fact that the high-pressure tapping can become clogged with solid materials.
This can happen with any slurry containing a significant quantity of solids.
If the high-pressure end is partially clogged with solids, the response of the
pressure difference measuring instrument will become slower and slower as it
gets more and more clogged. When the tapping point is completely clogged,
no pressure measurement is possible [5].

(b) Capacitance measurement

This level measurement technique is suitable for vessels made of conducting

materials (metal). In this case a long probe is inserted into the vessel containing
the liquid as one side of an electrical capacitor. The metal wall of the vessel
is usually used as the other electrical conductor [5].

The increase or decrease in the level of the liquid will induce a change in
the capacitance measurement. This change can be calibrated to provide the
measurement of the level of liquid in the vessel (see figures 1.18 and 1.19) [5].

It is very common for this type of level measurement technique to produce

errors. This is attributed to the fact that the electrical properties of the liquid, the
probe and the wall of the vessel could change. For this reason, this technology
is not commonly used in certain process industries. A similar and innovative
alternative method, known as radio-frequency admittance (RFA), is being
increasingly frequently implemented in a number of process and manufacturing
industries [5].
(c) Ultrasonic and radar measurement
Ultrasonic and microwave measurement instruments are expensive devices that
can be utilised to estimate the level of a liquid in a container (see figure 1.20).
These devices use a transmitter element to send a radar or ultrasound signal
into the container from the top. The surface of the liquid is used to reflect the
incoming signal, which bounces off the surface of the fluid and is then detected
by a receiver. The level of the liquid in the vessel can be estimated from the
distance between the surface of the liquid and the ultrasonic or radar unit.
This distance can be derived from the time lag between the signal transmission
from the unit and the reception of the return signal [5].

The accuracy of the readings produced by an ultrasonic level measurement

instrument can be negatively influenced by dense foam layers that dampen
the sound signal. However, the readings of both the radar and ultrasonic level
measurement instruments are independent of the liquid density [5].
Chemical composition is a very important factor that should be critically
monitored in a number of elements of chemical processing plants such as flash
operation units, distillation columns, membrane units and chemical reactors.
Consequently, the measurement of chemical composition is one of the most
valuable measurement actions in chemical plants. [5].

(a) pH meters

pH measurement is very common in water treatment plants because it is

crucial to ensure that the pH of the waste effluents is close to neutral (pH =
7) before their discharge into the environment. Moreover, pH control is very
important inside chemical process reactors because they support the occurrence
of certain reactions and either prevent or promote the precipitation of certain
chemical substances [5].

The pH of a solution is mathematically expressed as the negative logarithm

to the base ten (Log) of the concentration of hydronium species in a solution
according to the equation below [5]:

A pH electrode is the standard measuring instrument for pH (see figure 1.21).

The pH meters used for simple laboratory applications are way less bulky than
those used for industrial applications. These devices are usually obtained as
self-contained transmitters, and the signal from the instrument will often match
any standard instrumentation system (e.g. digital, 4–20 mA). When selecting
a pH meter for industrial application, it is important that the electrode of the
pH meter should suit the particular process conditions of the plant so as to
avoid the progressive deterioration of the instrument [5].
(b) Chemical analyser

Online chemical analysers are not popular in industry because of the numerous
disadvantages associated with their use. The first step of the analysis process
consists of collecting the sample of the material to be analysed. This step on
its own involves piping, pumps, valves and control gear. The sampling session
also induces time delay or dead time in the overall measuring process. The
presence of significant dead time in a control system might cause various
problems for the system controllers [5].

Moreover, the fact that chemical analysers are made specifically for the
measurement of particular chemical substances makes them quite expensive;
because they cannot be used for a general purpose and for different materials, the
corresponding analyser would have to be calibrated for that particular material.
Some of the commonly used modern analysis techniques are Raman scattering,
microwave acoustic, UV-visible, mass spectrometry and chromatography (see
figure 1.22) [4; 5].

(c) Laboratory analysis and inferential control

The use of online chemical analysers is very expensive and associated with
various technical difficulties, which is why they are only rarely used across
process industries. Effective and more affordable chemical analysers are
being developed. Meanwhile, most process industries monitor and control
chemical composition in processing plants through well scheduled sampling
and laboratory analysis combined with inferential control. Inferential control
is a control technique that uses the measurements of other process parameters
to infer the measurement of a variable of interest. In this case measurement
of other process parameters is used to infer chemical compositions. As an
example of a simple inferential control system, the temperature of the tray
near the top of a distillation column can be utilised to infer the compositions
of the compounds present at the top of the distillation column. To improve the
accuracy of the control system, numerous measurements of related process
variables can be coupled with an appropriate process mathematical model
to generate the desired composition. Considering the distillation column
mentioned previously, a more complex inferential control system might involve
the measurement of the tray temperature, the knowledge of equilibrium data
and the measurement of the column pressure combined with the application
of an appropriate process modelling to infer the compound compositions at
the top of the distillation column. Regardless of the types of inferential systems,
process results are based on assumptions. Therefore, the systems should be
calibrated on a routine basis using periodic laboratory analysis [5].

The documents that describe modern industrial processes use a schematic,

symbol-based “language” that is maintained in a form that is both simple and
easily read, without all of the detailed information needed by a specialist. These
documents present general information that defines processes, and they serve as
the key to the more detailed documents. Information presentation and storage
therefore become more efficient. The symbols presented in the document may
be inaccessible to those unfamiliar with the process nomenclature, but provide
a wealth of information to those trained to understand them [1; 2].

Like any living language, the documentation (symbols and their applications)
used to describe modern control systems has been improved constantly over
many years to meet modern challenges while maintaining the primary objective
of efficiently and clearly communicating important points about a specific
process to the trained reader. These documents include [1; 2]:

• process flow diagrams (PFDs)

• piping and instrumentation diagrams (P&IDs)
• instrument lists or indexes
• specification forms
• binary logic systems
• installation details
• location plans
• loop diagrams

For the purposes of this module, we will focus mainly on PFDs and P&IDs.
A PFD is a type of diagram usually used in chemical engineering and process
engineering to illustrate the relationships between major components of
chemical plants (see figure 1.23).

PFDs provide a macroscopic, schematic view of the major features of a chemical

plant process, but do not show minor details such as piping details and
designations. They show what and how much of each product the plant
will make, the quantities and types of raw materials necessary to make the
products, what by-products are produced, and the critical process conditions
such as pressures, temperatures and flows necessary to make the product. They
serve as a communication tool for managers, planners and the specialists of
a process design team. Because of its macroscopic nature, the control system
design engineers have little involvement in developing the PFD; however,
they may find the document quite useful at a later stage when developing
instrumentation for the control system. Manufacturers use PFDs to document
a process, improve a process, or model a new one. A PFD may also be called
a flowsheet, process flow chart, schematic flow diagram, top-down flowchart,
block flow diagram, macro flowchart, system flow diagram or system diagram,
depending on its use and content [1; 2].

P&IDs are the master design documents that use symbols and words to describe
the equipment, piping, instrumentation and control system for a process.
They are also the key to finding additional information about any specific
device or equipment on many other documents relating to the processing
plant. Developing P&IDs for manufacturing plants is a very interactive process
requiring the input of many specialists. These include experts in the fields of
electrical and mechanical equipment design, control systems, piping, and
even civil and structural design. Information is added progressively by each
group of engineers in a standardised way. When properly done, P&IDs serve
as records of the history of the plant design of any manufacturing facility. They
are also a great training tool for process operations. Because P&IDs are primary
coordination documents for design, it is very important to use the appropriate
equipment symbols and instrument tag numbers. An example of a typical
P&ID of a chemical processing plant is illustrated in figure 1.24 below [1; 2].

The standard method of indicating an instrument on a P&ID document is to

show the symbol (see figure 1.25) defining the type of instrument inside a
bubble. Inside the bubble the instrument tag number identifies the device [2].

The instrument tag number consists of a few letters that briefly convey the
function of the device, plus a combination of a number and letters that uniquely
identify the device. Usually these numbers are associated with a particular
control loop (see figure 1.26) [1].
Identification letters on the instrument symbols indicate [1; 2]:

• the variable being measured (e.g. temperature, flow, pressure)

• the function of the instrument (e.g. valve, switch, sensor, transmitter, indicator)
• some modifiers (e.g. low, high, multifunction)
Piping and connections are represented on P&IDs by means of a number of
symbols (see figure 1.27):

• Piping is represented by a heavy solid line

• Process connections to instruments (e.g. impulse piping) are shown by a
thin solid line
• A dashed line represents electrical signals (e.g. 4–20 mA connections)
• A slashed line denotes pneumatic signal tubes
• Data links are represented by a line incorporating circles

Other connection symbols are hydraulic signal lines, capillary tubing for filled
systems (e.g. remote diaphragm seals) and guided electromagnetic or sonic
signals [1; 2].
Good control design makes it possible to achieve a hierarchy of control

operating objectives of the plant. Process control reduces variation and results
in consistently high product quality that closely approaches the theoretical

manufacturing plant are the devices used to control the essential variables
that enable the conditions of a process to be determined at any time t [9; 10].

In this chapter, I set out the basic principles of process control and instrumentation.
I also presented an overview of the control loop and its elements. We will
talk about these topics in detail in later chapters, and you will gain a more
quantitative understanding of the basic concepts of control systems and their
application [10].

The key points of the chapter are [10]:

• Process control can be applied in any situation in which a process variable

is regulated and maintained to some desired value or range of values.
• The rational arrangement in which the elements of measurement, error
detector, controller, and control element are connected to provide the
required regulation of the process variable is illustrated by the block diagram
in figure 1.6.
• P&ID drawings and symbols are the standard graphical representations used
to present process control systems of manufacturing processes.
 (1)
Society of America, Research Triangle Park, NC, 1984.
(2) ISA, ISA-S5.3. Graphic Symbols for Distributed Control/Shared Display
Instrumentation, Logic and Computer Systems, Instrument Society of
America, Research Triangle Park, NC, 1983.
(3) ISA, ISA-S5.4-I989. Instrument Loop Diagrams, Instrument Society of
America, Research Triangle Park, NC, July, 1989.
(4) ISA, ISA-S5.5. Graphic Symbols for Process Displays, Instrument Society
of America, Research Triangle Park, NC, 1985.
(5) Kane, L. (Ed.). Handbook of Advanced Process Control Systems and
Instrumentation, Gulf Publishing, Houston, 1987.
(6) Luyben, M.L. and Luyben, W.L. Essentials of Process Control, McGraw-
Hill, New York, NY, 1997.
(7) Marlin, T.E. Process Control: Designing Processes and Control Systems
for Dynamic Performance, 2nd ed., McGraw-Hill, New York, NY,
2000.
(8) Matley, J. (Ed.). Practical Instrumentation and Control II, McGraw-Hill,
New York, 1986.
(9) Mayer, Otto. Origins of Feedback Control, MIT Press, 1970.
(10) Ogunnaike, B.A. and Ray, W.H. Process Dynamics, Modelling and Control,
Oxford University Press, Oxford, 1994.
(11) Seborg, D.E., Edgar, T.F. and Mellichamp, D.A. Process Dynamics and
Control, 2nd ed., Wiley, Hoboken, NJ, 2004.

1.1 Study the process of a traditional electrical oven in a household kitchen.

Explain its inherent control system. Is it a manual or automated control
system? Identify the controlled variable, the manipulated variable and the
disturbances.
1.2 Consider an automated gas-fired geyser used in a typical modern
house. Identify the controlled variable, the manipulated variable and the
disturbances.
1.3 Explain the difference between the final control element and the
manipulated variable.
1.4 The efficient treatment of industrial waste effluent before its disposal into
the environment is very important to preserve the ecosystem and ensure
public health. Consider the acid neutralisation process depicted in figure
1 below. The acidic waste process stream enters the neutralisation unit
at a constant flow rate. The base influent to the unit is a standard base
stream of fixed concentration and the flow rates of the base influent and
neutralised effluent can be adjusted.

(a) What is/are the process objective(s)?

(b) What are the possible controlled variables, manipulated variables
and disturbances?
1.5 Review the storage process depicted in figure 2 below. The objective
is to maintain a desired level of the liquid in the storage tank by either
increasing or decreasing the outlet flow rate from the storage tank, since
the flow rate of the inlet stream is constant. Draw a basic labelled control
loop for this process.

1.6 At the bottom of a tank containing liquid methyl ethyl ketone, a gauge
pressure of 2500 Pa is measured. At that specific location the atmospheric
pressure is equal to 101 325 Pa. Determine the absolute pressure measure
at the bottom of the tank.
1.7 State the functions that the following symbols represent in a chemical
processing plant:
Productivity and the improvement of the quality of end products in most industries
is achieved through the use of advanced control systems adopting complex
control strategies for both non-critical and critical applications. Generally, these
contemporary industrial control systems are made to produce one or more
control actions based on the application implemented. Imagine that you are
sitting in the seating area of your home on a cold winter evening, with a small
fire burning in the fireplace. Because you are feeling very cold, you put another
log on the fire. This is a typical example of a basic control strategy applied
in everyday life. The controlled or process variable (temperature) in the loop
of your control system fell below the set point (the temperature at which you
are comfortable), and you corrected the fluctuation to bring the process back to
the desired condition by taking the action of adding fuel to the fire. Afterwards
the control system remains inert until the temperature again deviates from the
set point (the temperature at which you are comfortable) [1; 2].

This chapter describes the basic types of control strategies commonly used
in industrial control systems. Some basic concepts of control strategies are
presented, and the fundamental principles required to understand more complex
control processes and algorithms that we will be discussing later are provided.
Additional terms and concepts relating to process control are also defined.
After studying all the sections of this chapter and completing the activities and
exercises you should be able to:

• Differentiate between the three tasks necessary for process control to occur:

– Measurement
– Comparison
– Adjustment

• Define the concepts:

– Control system, control system lag, dead time and capacity

– Control algorithm
– Process upset, load, demand, steady and transient state
– Load disturbance and response
– Set point and disturbance change
• Differentiate between open-loop control and closed-loop control strategies
• Distinguish between manual control systems and automated control systems
• Distinguish between analogue and digital control systems.
• Explain supervisory control and data acquisition systems, discrete control
systems and programmable logic controllers
• Derive the transfer function of open-loop and closed-loop control systems
• Derive the algebra of process block diagrams

Regardless of the type of industrial process, process control loops work in the
same way, requiring that three tasks be carried out. These are:

• measurement
• comparison
• adjustment

Although various instruments and devices (sensors, transmitters, valves,

controllers and pumps) may be used in control loops, the three tasks of
measurement, comparison, and adjustment are always necessary in order for
the control mechanism to occur. Let’s examine the control loop illustrated in
figure 2.1. The level in the storage tank is measured and transmitted by the
level transmitter (LT) as a corresponding signal to the controller (LC). Afterwards
the controller compares the transmitted value with the set point value, which
in this case is the maximum tank level established by the plant operator. If the
maximum tank level is reached, the controller generates an adjustment signal
to open the valve at the bottom of the tank in order to let some liquid out of
the tank, thus lowering the level of the liquid in the storage tank [2].

A control system is a combination of devices or components acting together

to achieve a specific objective. It generally exchanges with the surroundings
through signals, which are normally functions of time (see figure 2.2). There
are two types of signals [3]:
(a) Input signal [u (t)]: This comes from the surroundings and affects the
control system behaviour in a particular way.
(b) Output signal [y (t)]: This is generated by the control system with the
purpose of influencing the surroundings in a particular way.

The state of the process being controlled is key in order to select the appropriate
control system. There are two main states that can be used to describe a
process [3; 4]:

(a) Steady state of the process: This is a constant state of the process. It does
not vary with time – in other words, all the process variables remain
constant all the time.
(b) Transient state of the process: In this case, the state of the process changes
with time – in other words, the process variables change every time.

There are certain terms that are used consistently in relation to a typical control
system to describe the control mechanism taking place during the operation
of the system. These are summarised in table 2.1 [4; 5]:
Real industrial processing plants always operate in the presence of disturbances.
Consequently, they require the implementation of effective control systems.
Fundamentally, there are two types of control loop: open-loop (feedforward)
control and closed-loop (feedback) control. In rare cases a strategy involving
the combination of both open-loop (feedforward) control and closed-loop
(feedback) control is utilised [4].

By definition, an open-control loop exists where the process variable is

not compared with a set point value and the control action is independent of
the condition of the process variable, but is instead taken without regard to
process variable conditions [4; 5].

Feedforward control is a typical example of an open-control system (see figures

2.3 and 2.4). The feedforward controller converts one or more conditions that
could disturb the controlled variable into corrective action with the objective
of minimising the fluctuation of the controlled variable if [4]:

• the disturbance affecting the process output can be identified

• the disturbance can be measured
• the manner in which the control signal affects the process output is known

We encounter all sorts of open-loop control systems in everyday life. These

include traffic lights, washing machine cycles, lights and heaters, where the
control result is known to be nearly sufficient under normal conditions without
feedback being required. For example, a washing machine is programmed to
maintain a series of operations needed to wash a load of clothes. Therefore,
it runs through its cycle without any feedback information regarding the
condition of the washed clothes. In the case of disturbance, the machine will
shut down and only a human operator verifying the status of the wash and
finding it unsatisfactory can implement the corrective action [5; 6].

(a) Advantages

(1) It is simple, good for slow systems, and reduces control system complexity.
(2) It acts before the effect of the disturbance has been felt by the process.
(3) It does not involve control loop instability.

(b) Disadvantages

(1) This type of control system is ineffective if the disturbance cannot

be measured.
(2) The control system requires the identification and measurement of
all possible disturbances.
(3) Good knowledge of the process model is required.
(4) The control system is sensitive to process parameter fluctuation.

Consider as an example a continuously stirred tank where two process streams

are mixed. The objective is to control the composition of compound A in the
stream exiting the tank. The variation of the composition of compound A in
the aqueous stream entering the mixing tank represents a disturbance to the
mixing process. A feedforward control loop can be designed (see figure 2.5)
to maintain the exiting composition of A.
A closed-loop control system is essentially a control system that has a feedback
monitoring mechanism. The deviation signal formed as a result of this feedback
is used to control the action of a final control element in a specific manner
in order to reduce the deviation of the controlled/process variable from the
set-point to zero [6; 7].

A closed-control loop exists where a process/controlled variable is measured

and compared with a desired set point, and corrective action is taken to oppose
the effect of any deviation from the set point. These control systems basically
operate at fixed frequency, and the corrective action from the controller is
dependent on the process output. The frequency of the corrective changes made
by the controller to the driving signal is generally the same as the frequency
of the controlled variable sampling rate. In other words, after receiving and
reading each new sample from the sensor, the controller software recalculates
the changed state of the process and adjusts the corrective drive signal sent to
the final control element. The process responds to the implemented correction
with another sample taken, and the cycle is repeated. The process will eventually
reach the desired state and the controller software will then stop implementing
changes [6; 7].

Feedback control loop systems are quite effective, as they are based on
the measurement of the process output. The measured variable is often the
process variable that needs to be controlled. Controllers that utilise feedback
mechanisms are classified as closed-loop control systems. The feedback
information is used to decide on the corrective changes that need to be sent
to the control signal driving the plant [7; 8].

There are a large number of closed-loop controls, such as the cruise control
in a car, elevator drives and thermostats. For instance, if your car is going too
fast, the cruise control system temporarily reduces the amount of fuel fed
to the engine. Similarly, a feedback mechanism (thermostat) indicates that
the temperature in a room is above the set point, and this switches the air-
conditioning device on until the room is at the desired temperature [8]. The
block diagram of a basic closed-control loop system is given in figure 2.6.

The basic components of a typical feedback control system are summarised

in table 2.2 below.
The error, conditioned by the controller type and the tuning software, initiates
the corresponding changes in the final control element in two ways [8]:

Negative feedback – this is the desirable situation, during which corrective

action taken by the controller drives the process variable towards the set point.
In this case we have:

Error = set point – feedback signal

Positive feedback – this is the situation in which the controller drives the
controlled variable farther away from the set point. The error in this case is
calculated as follows:

Error = set point + feedback signal

(a) Advantages

(1) The identification of all possible disturbances and their precise

measurement is not necessary.
(2) This type of control system is insensitive to parameter changes.
(3) This type of control system is not affected by process modelling errors.

(b) Disadvantages

(1) It only starts operating once the effect of the disturbance has been
felt by the system.
(2) It can create instability in the control loop.
(3) This kind of system is not suitable for slow systems.

Consider again the continuously stirred tank depicted in figure 2.5. A

feedback control loop can also be designed (see figure 2.7) to maintain the
exiting composition of A.
The strategy of combined feedforward and feedback control is often used when
either feedback or feedforward control strategies are not as effective as one
would like them to be. Basically, whenever possible all major disturbances are
measured to calculate the feedforward control action. Meanwhile, feedback
control serves as a long-term correction in the sense that it acts against the
slow deviation of the control variable from the set point which the feedforward
cannot be expected to correct [9]. An example of a combined feedforward
and feedback control can be clearly illustrated using the scenario of the mixing
tank (see figure 2.8).
Before process automation, process control systems were operated manually.
In this case an operator implements the controller output and manipulates the
final control element accordingly to maintain the recommended set point. A
control system that requires human action to make adjustments is referred to
as a manual control system. For example, a control system that regulates the
temperature of the water coming out of a showering device in a household
bathroom can be operated manually. A human operator senses the temperature
of the water on their skin and compares it with the desired temperature [8].
Correction action is taken by simply adjusting the flows of hot and cold water
(see figure 2.9).

Control operations in which no human intervention is involved are classified

as automated control systems. In these instances machines, electronics
or computers replace the operations of the human operator. The control
system depicted in figure 2.10 is an example of a typical automated control
system that uses an instrument called a sensor to measure the value of the level,
and then send this information in the form of a signal as input to a machine,
electronic circuit, or computer called the controller. The controller performs
the function of the human in evaluating the measurement and providing an
output signal that automatically adjusts the valve actuator according to the
controller output (see figure 2.10) [8].

Controllers are programmed to perform either simple or complex mathematical

operations to compare a measured value of the process variable with the set
point. Effective controller devices should always have the ability to receive
input from a sensor or transducer to perform a mathematical function in
order to compare the input with the programmed set point, and to produce
the corresponding output signal to be sent to the final control element (see
figure 2.11).
Many kinds of automated controllers are commonly used in chemical plants.
These include [4; 5]:

These are among the most elementary types of controllers, and have been in
use for a number of years. They are called ON/OFF control systems because
the final control element has only two states: on and off. As a result, the
controller output is also required to have only two equivalent states (see figure
2.12). They are the simplest and least expensive types of controllers, and they
are often used where cycling can be reduced to an acceptable level. ON/OFF
controllers are quite effective when the overall system has a relatively long
response time. However, this will result in instability if the process under control
has a rapid response time. These types of controllers are sometimes termed
two-step controllers, and they are essentially a switch which is activated by the
error signal entering the process to provide an ON/OFF corrective signal [6; 7].

These controllers are primarily computers connected to a set of input/output

(I/O) devices. These computers are programmed to respond to inputs by sending
corresponding outputs to maintain all processes at desired set point. Figure
2.13 below depicts the multiple connections involved in the installation and
implementation of PLCs. For larger processes or systems, more sophisticated
or complex central units of PLCs are required [6; 7].
In addition to performing control functions, DCSs also provide readings of the
status of the process, and maintain databases and advanced human–machine
interfaces (see figure 2.14). DCSs were designed to control non-discrete process
operations [7].

These controllers exist when all variables in the system are analogue
representations of another variable. You can see the layout of the interior of
a typical analogue controller in figure 2.15 below.
Consider the process illustrated in figure 2.16 below, in which a furnace is used
to control temperature in a house, and assume in this case that the furnace is
operated by means of electricity.

For this scenario, the furnace output, Q, is an analogue of the excitation

voltage, and thus heat can be varied continuously. Therefore, every output
is an analogue of the output temperature; the error E is an analogue of the
difference between the reference voltage and the temperature voltage. The
reference voltage is simply the voltage that would result from measurement
of the specified reference temperature (set point) [7].
These control devices use computers in two different contexts:

(a) Supervisory control and data acquisition (SCADA) – at the time

when computers were first considered for application in control systems,
they tended to suffer frequent failures and breakdowns. The need for
continuous operation of control systems therefore meant that the use
of computers to perform the actual control operations was excluded,
and supervisory control was implemented as an intermediate step. The
intermediate control system uses the computer to monitor the operation
of analogue control loops and to determine appropriate set points (see
figure 2.17). The advantage of this control system is that a single computer
is able to monitor numerous control loops and use appropriate software to
optimise the set points for optimum overall plant operation. Also, a potential
failure of the computer will not affect the plant operation; instead, the
analogue loops will keep the process running using the last programmed
set points until the computer comes back online. This technology was
developed to monitor and control very large process facilities. Moreover,
it makes it possible to collect data from one or more distant facilities while
sending limited control instructions to those facilities [7].

(b) Direct digital control (DDC) – Nowadays computers have become more
reliable and much smaller, and as a result they are taking over the controller
function, which means that the analogue processing loop is no longer
necessary. Figure 2.18 below illustrates the layout of typical micro-devices
used in DDC systems [6].
These are designed to minimise the consequences of emergency situations
typically associated with the escape of hydrocarbons, uncontrolled flooding or
outbreak of fire in hydrocarbon carrying areas or areas which may otherwise
be hazardous. Emergency shutdown equipment for a process control system
consists of an emergency shutdown (ESD) valve and an associated valve
actuator. The controller provides output signals to the ESD valve in the event of
a failure in the process control system. This type of control system is required
in processing plants where a high safety integrity level is required (see figure
2.19) [7].
Process control systems are implemented in industrial plants as part of an
information processing network – they are the centre of communication of the
processing plant. To put it simply, from the various instruments, they gather
information regarding the actual state of the plant process. The information
gathered is then compared with a set of standards incorporated into the system.
This comparison enables the system to produce the appropriate action required
to shift the process towards the desired state [8].

Finally, the information is sent to the corresponding final control elements

in order to generate the required change in the process. To successfully
convey the information from the various process instruments to the
controllers and from the controllers to the numerous final control elements,
an efficient control communication network is needed. A process
control communication network can consist of several different types
of communication signals. Many processing plants usually have only one
kind of process control communication system. However, the system will
be a combination of various types of communication signals throughout the
process [8].

These types of signals are continuous. They are referred to as analogue signals
because the intensity of the communication signal is directly equivalent to the
value that is being conveyed or communicated. Analogue signals can handle
any value that is within the range between their minimum and maximum
acceptable values. Various types of analogue signals are commonly used
as communication signals within control systems in chemical processing plants
[6]. Some of the most popular of these are:

(a) Pneumatic signals – A pneumatic signal is an air-pressure communication

signal. The pressure range is between 3 and 5 psig, with the lowest value,
3 psig, as the offset zero. This minimum value of the pressure range
makes it possible to detect instrument malfunction or tube breakage. The
pneumatic signal is one of the oldest forms of control communication signal
utilised for automatic and remote control in industrial plants. The signal
is generally transmitted through metal tubes of small diameter – about
10 mm on average. The practical length of the communication tubes is
limited by the fact that an increase in the tube length is directly responsible
for a substantial increase in dynamic delay due to the longer time it will
take for the control instrument producing the signal to add the amount of
air necessary to increase the tube pressure to the required value. For this
reason, pneumatic communication signals are now outdated. However, in
some old processing plants where they are still used, control rooms have
to be situated close to the site of the control instruments [6; 7].
(b) Current loops – Controllers using the 4–20 mA current loop as communication
signal are the most commonly used instrument in chemical processing
plants. The instrument operates by sending an electrical current around
a loop between the signal transmitter and the signal receiver. This type of
analogue communication signal is preferred to the voltage signal because
the effectiveness of the signal transmission is not affected by the length
 of the communication line. The voltage signal is dependent on the length
of the communication line, as an increase in the communication line will
result in an increase in voltage signal loss. In a 4–20 mA current loop, the
transmission device increases the communicating voltage until the current
circulating in the loop corresponds to the desired signal. Current loop
analogue signals are usually carried in electrically twisted or not earthed
shielding wires. These communicating cables usually include additional
wires for power supply [6; 7].
(c) Calculations for analogue signals
Consider a temperature measuring device calibrated to a given span of
35 °C – 145 °C. It can be connected to any analogue signal transmitter
device, as the conversion is linear. If the temperature measured by this
instrument is 90 °C, calculate the output of a pneumatic and a current-
loop transmitter.

Solution

(a) Pneumatic transmitter: pneumatic signals operate in the pressure range of

3 to 15 psig; therefore, the output of this transmitter can be calculated as:

(b) Current-loop signal: current-loop signals operate in the current range of

4–20 mA; therefore, the output of this signal transmission device can be
calculated as:

In chemical manufacturing and processing plants there are a large variety of

instruments such as switch devices for pumps, motors, agitators, and instrument
switches (e.g. temperature and level trips). These operate in one of two states
only (usually either the ON or the OFF state), and use discrete signals as means
of communication. Discrete signals do not have a universal standard range.
However, the two states are represented by 0 and 24 V [7].

In this communication mode, the measured value is converted into a number

before it is transmitted down the communication line. Only one number
can be sent down the communication line at a time; in consequence,
digital communication signals are not continuous, but are sampled instead.
This means that information collected from a digital communication line
is basically a series of numerical data representing the state of a quantity
measured at various times. Digital communication signals are conveyed through
transmission lines as two binary numbers: 0 and 1. These binary values are
represented by voltages: 0 V and 5 V are usually the most popular voltages
used for 0 and 1 respectively. Depending on the electronic device selected,
the digital signal will be transmitted in various combinations and resolutions
of the binary numbers [7].

Digital signals can be transmitted in two ways: either in series or in parallel.

The series mode of transmission consists of sending the binary digits (also
known as bits) one after the other. In this transmission approach additional
control bits are always present before and after the digit of the numerical value
being measured by the instrument. Therefore, a set of two communication
wires can effectively convey a serial signal communication. For the parallel
transmission approach, a communication line is attributed for each bit of the
binary signal. Extra signal lines will be incorporated in order to monitor the
signal communication process. For example, a 16-bit parallel signal line will
require more than 16 communication wires [6; 7].

For the past decade, serial communication has by default increasingly become
the transmitting mode of digital signals in most industrial applications. However,
the parallel mode of digital signal communication is generally much faster than
the serial mode. Because a significantly larger number of wires is required for
the parallel mode, this limits its application from an economic point of view,
and as a result the slower serial communicating approach is more often used
instead [7].

Digital communication signals are considered to have significant advantages

over analogue communication signals. For instance, analogue transmission
methods require an individual cable to handle the connection between the
transmitter and the receiving device. So, for example, a large chemical plant of
500 instruments will need 500 cables to individually connect the transmitters
of all 500 instruments to their respective receiving devices in the control
room. This situation can be avoided through the use of digital communication
signals because the address of the instrument to be communicated with can be
incorporated into the transmission device. Hence, a single cable in this case
is all that is required to connect multiple transmitters with multiple receivers.
The fact that it has to sample digital signals might appear to be a disadvantage
of this type of communicating signal, because it does not represent the state
of the process plant in a continuous manner. However, in most chemical
processing plants the sampling time intervals are usually considerably shorter
than the speed of response of the various processes occurring in the plant [7].

A richer communication between instruments is usually achieved with

digital communication signals. This means that more information than just
the value of the instrument being measured can be transmitted to the receiving
end of the communicating line. This includes the last time routine calibration
was carried out, or the rate of change of the measurements. Moreover, with
digital signals the communication can also take place in the reverse direction.
For instance, the controller device can use the transmission line to induce the
measuring instrument to perform a self-evaluation [6].
Numerous digital communications standards are currently used in chemical
processing plants. Because most of them are not interchangeable, when using
digital communication signals it is best to utilise only one supplier [6].

Highway addressable remote transducer (HART) communication lies in

between analogue communication and complete digital communication.
This class of communication can be utilised as a two-way communication
approach between the central controller device and the measuring instrument.
It can also be fitted as a typical 4–20 mA communication line with digital
signals superimposed on top of the analogue communicating signals. The two-
way communication set-up can serve various purposes, including diagnostics
and troubleshooting, other measuring information assessment and device
configuration. Furthermore, HART communications are able to convert an
existing 4–20 mA line into a multi-drop line. This simply means that the
multiple 4–20 mA devices will carry on the communication process over the
same communication cable. The devices will have their own time schedules on
the cable, and this time sharing will be monitored by the HART protocol [6; 7].

Fieldbus is the generic name given to a number of standards that have been
established since the year 1990. The latest iteration of these standards is
known as Foundation Fieldbus, and it is a standard for complete processing
plant digital communications. This is the largest installed base of all standards.
This standard utilises series of digital lines called buses connected to multiple
transmission devices, and each device attached to a line has an address. These
addresses are used to convey operating instructions to the appropriate device.
A controller obtains a measurement from an instrument on a fieldbus according
to the following protocol [6; 7]:

(a) Initially the controller uses the instrument line address to convey an
instruction requesting measurement from the device.
(b) Afterwards, the instrument takes the requested measurement and finally
transmits the measurement recorded to the controller also using that
line’s address.

All devices connected to the bus lines constantly monitor the bus lines; however,
each one responds only to instructions directed to its address.

Ethernet refers to a group of standards used for office computer networks. In

the past this family of standards was never consider for plant control systems
applications. However, because this technology has improved so much in terms
of speed and cost effectiveness, it was recently considered a good prospect
for industrial process control systems [6].
Nowadays, wireless communications are being considered for some
industrial process applications. The inherent advantage associated with these
standard communications is the absence of wiring between the instruments
and the controller device. This makes them perfect candidates for distant and
exposed regions. However, the effectiveness of the communicating signals is
closely dependent on the location, and can easily be subjected to interference.
For example, if a large object is placed in front of the transmitter device, this
could severely affect the quality of the communicating signal [6].

The transfer function is a convenient representation of the input–output

relation for a linear time-invariant dynamic system. One of the most common
and useful methods of representing a system is by its transfer function. The
transfer function is easily determined once the system has been described as
a single differential equation [6].

The combination of transfer functions and block diagrams provides an

dimensional systems the transfer function is a simple rational function of complex

variables. It can be obtained by simply inspecting algebraic manipulations of
the differential equations that describe the systems. Transfer functions can

governed by partial differential equations. The transfer function of a system

can be determined from experiments on a system, and is easily extended to
systems with multiple inputs and/or multiple outputs [6; 7].

The primary aim of a control system is to keep the controlled variable(s) at

the set point(s). To achieve this objective, a control algorithm is used to adjust
the manipulated variable appropriately. A process control algorithm is the
mathematical expression of a control function. Control algorithms can also be
utilised to calculate the requirements of much more complex control loops.
In more complex control loops, questions such as “How far should the final
control element be adjusted in response to a given change in set point?” and
“How long should the final control element be held in the new position after
the process variable moves back toward set point?” are to be answered [7].

The block diagram of a control loop is basically a schematic illustration of

the signals representing the flow of information around the control loop.
Block diagrams can be the simplest representation of any social, economic
and engineered systems. Every block of the diagram illustrates a specific
operation indicating what happens to specific input(s) in order to achieve
specific output(s) [8].

Signals in block diagrams are represented by arrows to indicate the direction

of flow of information in the diagram. Generally, the presence of arrows in the
block diagram of a dynamic system also clearly denotes that the output signal
is closely dependent on the input signal (see figure 2.20) [8].
The closed-loop feedback system adjusts any control signal by primarily
determining the error between the system actual output and the desired output
set point. This control task is performed using a comparison element known
as the summing point of the feedback loop and the system input. Basically,
the actual value of the system output is compared with the desired set point
value and either a negative or a positive error signal is generated to regulate
the controller output signal accordingly [8].
Error = Set point – Actual

The summing point in the block diagram of a closed-loop control is represented

by means of a circle with two crossed lines, as shown here:

Summing point

This comparison element used in feedback control-loop systems can add signals
together. In this case a plus (+) symbol is used to indicate that the device is a
summing element or a “summer” (positive feedback). It can also subtract signals

the device is a “comparator” (negative feedback). Below are typical graphical

representations of a summer and a comparator.

Summing points can be cascaded together in order to sum more than one
signal at any given point of the control loop [8; 9]. Although summing points
can have more than one input signal for either addition or subtraction, they
can have only one output signal, which is the algebraic summation of the
input signals [9].
The blocks forming the diagrams of control loops are interconnected via signals.
Below are examples of the most common algebraic operations performed on
signals of control loop block diagrams [8; 9]:

(a) Addition operation

(b) Branching

(c) Subtraction operation

The basic derivable transformation techniques used when deriving the transfer
function of a control loop from its block diagram are summarised as follows [8]:
(a) The combination of blocks in cascade

(b) The combination of blocks in parallel (feedforward loop)

(c) The combination of blocks in parallel (feedback loop)

A transfer function is an algebraic expression representing the dynamic relation
between selected inputs and outputs of a process. Consider the differential
equation with x(t) as input and y(t) as output. The transfer function is the ratio
of the Laplace transform of the system output to that of the system input, both
taken with zero initial conditions. The polynomial that forms the denominator
of the transfer function is called the characteristic equation. It largely determines
the nature of the system’s behaviour (oscillatory or heavily damped, slow or
fast, etc.) [10; 11].

The transfer function of a single unit process depicted in the block diagram
above can be expressed as follows [11]:

Transfer function (TF) =

For a process of multiple units corresponding to a block diagram involving

more than one component connected in series, the transfer function of each
unit is calculated as follows [11]:

The overall transfer function is obtained as the product of the transfer function
of each individual process unit of the block diagram only if the units are
connected in series [11].

The overall transfer function is therefore expressed as follows:

G= G1 G2

The dynamic behaviour of processes operating using feedback control

mechanisms is described by a closed-loop control system. Here the controlled
variables of the process are initially measured, and the resulting measurements are
transmitted as corresponding signals to the controller, where they are compared
with the desired set point. Then, depending on the nature of the error generated,
an appropriate corrective signal is sent to adjust the manipulated variables of
the process [11].
The transfer function of a closed-loop control system refers to a mathematical
expression (algorithm) describing the net result of the effects of a feedback
control loop on the input signal to the circuits enclosed by the loop. In this case
the transfer function is measured at the output. The input signal waveform is
used together with the transfer function to calculate the output signal. Consider
the typical control loop illustrated below [11]:

Block G1 represents, in the forward pathway, the open-loop gains of the control
system or controller. Block H represents, in the feedback pathway, the gain
of the sensor, transducer or measurement system. The transfer function of the
closed loop above is derived by calculating the output signal in terms of the
input signal. The equations of the block diagram provided can be written as
follows [10; 11]:

The output signal from the process can be determined as: Output = G1 Error

In this case, note that the error signal is also an input to the feedforward block,
G. Therefore, the output signal from the summing point is equal to: Error =
Input – G2 Output

Eliminating the error term from the expression of the output signal from the
system results in the following equation: Output = G1 (Input – G2 Output)

Therefore: G1 Input = Output + G1 G2 Output

G1 Input = Output (1 + G1 G2)

By rearranging the above expression, we obtain the closed-loop transfer function

equation as follows [11]:

The plus sign (+) in the denominator of the derived transfer function equation
of the closed-loop system indicates negative feedback. A positive feedback
system would be represented by a minus sign (–) in the denominator, and the
equation above would be written as [11]:
When G2 = 1 (unity feedback) and G1 is very large, the equation of the transfer
function approaches unity:

Moreover, in the case of unity feedback and the steady state of the system, a
decrease in the system gain G1 will result in a much slower decrease in the

expression of the transfer function.

This simply means that the system is fairly insensitive to variations in the
system gain (G1), which is one of the main advantages of a closed-loop control
system [11].

Although we applied the basic transfer function principle to a single-input,

single-output closed-loop system above, it also applies to more complex multi-
loop systems. Real-life feedback circuits usually have some form of multiple
loop control. The transfer function between the main controlled variable and
the main manipulated variable for a multi-loop configuration depends on what
types of feedback control loop (open or closed) the configuration is made of.
Study the multi-loop system below [11].

The multi-loop system contains an inner loop identified as follows [11]:

To determine the overall transfer function of this complex multi-loop control

system, we need first to reduce any blocks in series, such as g1 and g2 as
shown below [11]:

Afterwards, blocks of the inner loop can be reduced to yield the following
block diagram [11]:
Further application of the algebraic reduction technique to the inner loop
results in the diagram of a single loop with two blocks in series [11].

The above block diagram can further be reduced to a final block diagram
which resembles that of the previous single-loop closed-loop system [11].

Hence, the transfer function of this complex multi-loop system can be expressed
as follows [11]:
The basis of an efficient control system including the feedback and the
feedforward control of process variables is regulatory controls. Effective control
of process variables such as temperature, load level and pressure is vital, as they
have a tremendous impact on the quality of the final product manufactured by
the processing plant. The introduction of regulatory control at each successive
unit operations varies as a function of quality, which accumulates throughout
the manufacturing cycle and is reflected in the overall production cost and
specifications of the final product [1–11].

The feedback control system is based on a reactive strategy. Therefore, an error

must be detected by the control system before corrective actions are taken.
In some situations, it is possible to measure a disturbance before it affects
the process. The effect of the disturbance is then reduced by measuring it
and generating the appropriate control signal that counteracts it. This type of
controlling system is known as a feedforward control system [2–10].

The transfer function is an equation used to represent system dynamics, which

can also be done via s representation from Laplace transforms. Transfer functions
show the flow of signal through a system, from input to output. They are
basically the ratio of output signal to input signal (in the s-domain via Laplace
transformation). Transfer functions can also represent the dynamics of a given
system in the form of ordinary differential equations or state equations. The
control of processes can be done either manually or automatically, and both
analogue and digital processing are used in process-control applications [6 – 11].

(1) Abate, J. and Valkó, P.P. “Multi-precision Laplace transform inversion”.

International Journal for Numerical Methods in Engineering 60(5):
979, 2004.
(2) Cohen, A.M. “Inversion Formulae and Practical Results”. Numerical
Methods for Laplace Transform Inversion. Numerical Methods and
Algorithms. 5. p. 23, 2007.
(3) Davies, B.J. Integral Transforms and their Applications, 3rd ed., Springer-
Verlag, Berlin, New York, 2002.
(4) Distefano, J.J., Stubberud, A.R. and Williams, I.J. Feedback Systems and
Control, Schaum’s Outline Series, 2nd ed., McGraw-Hill, New York,
p. 78, 1995.
(5) Lipschutz, S., Spiegel, M.R. and Liu, J., Mathematical Handbook of
Formulas and Tables, Schaum’s Outline Series, 3rd ed., McGraw-Hill,
New York, p. 183, 2009.
(6) Luyben, M.L. and Luyben, W.L. Essentials of Process Control, McGraw-
Hill, New York, NY, 1997.
(7) Marlin, T.E. Process Control: Designing Processes and Control Systems
for Dynamic Performance, 2nd ed., McGraw-Hill, New York, NY,
2000.
(8) Matley, J. (Ed.). Practical Instrumentation and Control II, McGraw-Hill,
New York, 1986.
(9) Ogunnaike, B.A. and Ray, W.H. Process Dynamics, Modelling and Control,
Oxford University Press, Oxford, 1994.
(10) Riley, K.F., Hobson, M.P. and Bence, S.J. Mathematical Methods for Physics
and Engineering, 3rd ed., Cambridge University Press, p. 455, 2010.
(11) Seborg, D.E., Edgar, T.F. and Mellichamp, D.A. Process Dynamics and
Control, 2nd ed., Wiley, Hoboken, NJ, 2004.

2.1 Briefly describe the operation of (i) a feedback control system and (ii) a
feedforward control system. List the advantages and disadvantages of
both control strategies.
2.2 Briefly define the following terms:

(a) Control system, control system lag, dead time and capacity
(b) Control algorithm
(c) Process upset, process load, process demand, steady and transient state
(d) Load disturbance and response
(e) Set point and disturbance change

2.3 List the three fundamental tasks required for process control to occur.
2.4 Use block diagram algebra to simplify and determine the overall transfer
functions of the following block diagrams of control systems:
 (a)

(b)

(c)

2.5 Consider a gauge pressure measuring device calibrated to a given span

of 20–120 Pa. It can be connected to any analogue signal transmitter
device, as the conversion is linear. If the gauge pressure measured by this
instrument is 80 Pa:

(a) Calculate the absolute pressure if the atmospheric pressure at the

location is given to be equal to 1 012 350 Pa.
(b) Calculate the output of a pneumatic signal and a current-loop signal.

2.6 List and briefly describe the commonly used process plant
digital communication standards.
2.7 List and describe the different transmission approaches of digital signals.
What are the advantages and disadvantages of each?
2.8 List and briefly describe the different type of controller communication
signals.
2.9 How many kinds of analogue signals are commonly used in industrial
plants? Give a brief description of each kind.
We have already spoken about the fundamentals of process control and
instrumentation in chapters 1 and 2 of this study guide. The aim of process
control has been explicitly defined; the basic control loop and its characteristics
elements have been described; the role of instruments in ensuring effective
process control has been highlighted, and finally the symbology used as a
means of communication in engineering drawings has been presented and
discussed. However, very simple types of control loops have been used thus
far [1].

For the majority of industrial processes, external disturbances tend to cause

a number of process variables to vary all together in a continuous manner.
Consequently, most control systems generally do more than just monitor a
process-controlled variable. Instead, they regulate a sequence of events that
result in the production of a specific product from a set of raw materials.
These monitoring systems are often quite complex, and the techniques used
to maintain individual process units at their respective set points can be
fairly complicated. Moreover, controlling the effects of these various set points
on one another and on the overall process control system can be a difficult
and delicate operation [1].

In this chapter, we will talk about the most common strategies and methods
used in complex process control loops. We will also discuss the characteristics
of these techniques, along with various applications. After completing the
chapter, you should be able to:

• Differentiate between multi-step, continuous and discrete controllers

• Describe the characteristic mechanism of the following modes of
controller action:

– Proportional action
– Integral action
– Derivative action

• Discuss the advantages and disadvantages of the three modes of con-

troller action
• Describe typical applications of each mode of controller action
 • Discuss the operation of the proportional + integral (PI) control action,
the proportional + derivative (PD) and the proportional + integral +
derivative (PID) control action
• Explain the importance of tuning the controller for a particular process
•
• Identify the various control loop characteristics used to measure the
performance of a control system
• Describe control loop stability criteria
• Explain the fundamental principles of controller tuning
•
application
• Select the most popular tuning techniques for both open- and closed-
control loops

A controller is basically a device consisting of a main control component and

a comparator known as a reference element (see figure 3.1). The role of the
controller is to maintain the controlled system in such a way that most of the
time, the value of the controlled process variable (PV) is very close to the
desired value or set point (SP). Regardless of the type of controller, the first
step in the algorithm of most controllers is to compare a process measurement
with a desired set point and to generate an error [1].

Error = process measurement-set point

This calculated error is used to implement the corresponding compensating

change in the controller output or the controller response. The change is
then transmitted to the final control element in order to adjust the process
accordingly. The adjusting action of the control system has a final objective
of reducing the error calculated to zero [1; 2].

The process error (e), which is the difference between the process variable (PV)
and the desired value or set point (SP), is generated from the reference element.
Based on the magnitude and sign of the error produced, the control component
will generate an appropriate control action in order to bring the process variable
as close as possible to the desired value or set point. The control action can
be initiated in many different ways by the control component of the controller.
Based on the process error calculated by the reference element, the following
are the most common ways to generate the control action [2]:

• mechanically
• electrically
• digitally
• using analogue signal
• with or without auxiliary energy

These various means of generating the controller action do not have any
influence on the type of action produced. They are factors to consider purely
during the selection of the controller device. Based on the roles and properties
of the control mechanisms of industrial controllers, they can be classified as
either (see figure 3.2) [2].

(a) Discrete controllers,

(b) Multi-step controllers, or
(c) Continuous controllers.

Each type of controller is suitable for meeting the specifications of its corresponding
applications, and has its own particular advantages and disadvantages. Depending
on the type of controller, the control signal can be either continuous or discontinuous
[2].

Regardless of the type of control system in place, every type of controller

requires energy to operate. Controllers that require external electrical energy,
pneumatic energy or hydraulic energy to operate are categorised as controllers
with auxiliary energy. Self-operated regulators are commonly used when no
energy transfer medium is available at the point of installation [1; 2].

These regulators derive the energy necessary to help the control system to
generate the appropriate corrective action. Conventional controllers are rugged,
cost effective and popular nowadays for pressure flow, pressure differential and
temperature control. They are fairly effective in applications where the distance
between the point of measurement and the point of change is relatively short.
They can also be very useful in systems where process fluctuations caused by
energy withdrawal are acceptable [2].
These types of controllers are the simplest, and have only two control modes
or control positions: on and off. For this reason, they are sometimes referred
to as two-step controllers. Discrete controllers do not steadily keep the process
variable at the desired set point. They generally maintain the process variable
near the set point in a region known as the dead zone (see figure 3.3). A
household heating system is a typical example of a discrete control system [2].

At the moment that the temperature of the water in the tank reaches the desired
value, the heat supplying device turns off. As soon as the temperature of the
water in the tank starts to fall below the set point, the heating device turns on.
As soon as the water reaches the set point temperature again, the cycle will
begin again. Discrete controllers can be very useful in domestic applications
such as the typical household heating system described above. However, in
industrial process applications, where there is generally a large number of
individual controllers involved, the continuous oscillating nature of the response
of this ON and OFF controller is not desirable [2].

Multi-step controllers have a similar mechanism to discrete controllers, but

they have at least one other possible position in addition to the on and off
positions. Multi-step controllers operate in such a way that as the set point is
approached, the controller takes intermediate steps. As a result, a less drastic
fluctuation is registered around the set point when multi-step controllers instead
of discrete controllers are utilised (see figure 3.4) [1; 2].
Continuous controllers are controllers that automatically detect any existing
error occurring in the process by comparing the measured value of the process
variable with the desired set point. Once the error is detected and its magnitude
evaluated, the output action of the controller is adjusted appropriately to match
the parameters that have been implemented in the controller device. For
instance, consider the example of the liquid level in a tank. A tank containing
a process liquid is illustrated in figure 3.5. A control valve is used to monitor
the flow of liquid into the tank; meanwhile, a certain amount of the liquid
flows out of the tank [1; 2].

A transducer measures the level of liquid in the tank and transfers the value
of the measurement to a control system whose objective is to ensure that the
level of liquid in the tank is constantly at some pre-set or set point value. The
controller continuously operates according to certain control principles and
algorithms to maintain the level of the liquid in the tank, despite numerous
fluctuations generated by external influences. Therefore, if the flow of the liquid
out of the tank increases, the control system will increase the opening of the
input valve to counteract the effect of the disturbance by increasing the input
flow rate, hence balancing and regulating the level of liquid in the tank [1; 2].
This type of control system is classified as a continuous variable control system,
because both the level of liquid in the tank and the valve opening setting can
vary over a range. Although the controller operates in an OPEN/CLOSED mode
when it comes to the input valve, and this induces the oscillation of the level
of the liquid in the tank as the input valve is constantly opened and closed to
eliminate the effects of disturbances such as output flow variations, there is
still variable regulation [2].

To describe the operation of a continuous controller in a general way, it is

better to express the process error as a percentage of the measured variable
range known as the span of the instrument measuring the variable. The error
can therefore be calculated as the percentage (ep) of the span of the measuring
instrument over a range of measurement using the following equation [2]:

For example, consider a process set point of 14 mA and a process measurement

of 13.7 mA. If we use a measuring instrument with a span range of 5 mA to
25 mA, the percentage error will be calculated as [2]:

A negative error simply indicates that the process measurement is below the
set point, while a positive error indicates that the measurement is above the
set point [2].
Controller modes, also known as control principles, are the different ways in
which controllers react to error changes in any process. They can also be defined
as the ways in which a control system makes corrections relative to process
deviations. From a control instrumentation point of view, the mode of control
is the way in which a controller changes its output action relative to process
deviations. The three most popular controller modes are the proportional
(P), derivative (D) and integral (I) control modes. Each of these three modes
responds differently to the error. The amplitude of the response produced by
each control mode is adjustable by changing the controller’s inherent settings [3].

The algorithms of the proportional (P), derivative (D) and integral (I) control
modes were developed for the first time by an engineer in charge of the design
of the automatic steering system of a US navy ship. When the captain gave a
human helmsman a bearing to steer, the engineer observed that the human
helmsman considered the following three factors when adjusting the position
of the ship’s wheel [3]:

(1) How far the current bearing was from the desired bearing: Because a
huge difference would lead to a significant change in wheel position,
the corrective action is considered to be proportional to the difference
in bearing.
(2) How long the current bearing had been different from the desired bearing:
If a variation in bearing persisted, then the human helmsman would turn
the wheel further; the corrective action effectively integrated the detected
error.
(3) How quickly the actual bearing was fluctuating: If the actual bearing
varied quickly towards the desired bearing, then the human helmsman
would observe that the time-based derivative was too high, and would
adjust the wheel accordingly.
The proportional control action is the most basic control mode, and is also, in
most cases, the primary driving force in a controller. It changes the controller
output in proportion to the error input signal to the controller. This simply
means that detecting a bigger error will result in a corrective action of higher
amplitude. This is common sense, because large errors will obviously require
corrective actions of greater intensity [3].

This elementary control mode reacts without delay to any deviation from the set
point or to any measured disturbance or load change with the sole purpose of
keeping the process variable (PV) within an acceptable range around a desired
system set point (SP). The adjustable settings for the proportional mode are
expressed as follows [3]:

• proportional gain
• proportional band

(a) Proportional gain

The proportional or the controller gain Kc is defined as the ratio of controller

output (CO) and error signal (e). Therefore, the proportional gain usually answers
the question: “What is the percentage change of the controller output relative
to the percentage change in controller input?”

In electronic controller devices, the proportional gain of the controller typically

represents the proportional action, and can be expressed mathematically by
the following equation [3]:

In the proportional control algorithm, the controller output is proportional

to the error signal. This simply means that the output of a proportional only
controller is the multiplicative product of the error signal and the proportional
gain. This is expressed mathematically as [3]:

where:

• P(t) is the output of the proportional controller

• P0 is the bias value (this value is adjustable)
• Kp is the proportional gain
• e(t) is the instantaneous process error at time t e(t)= |PV - SP|
• PV is the process-controlled variable
• SP is the desired set point
(b) Proportional band

The proportional control band (PB) is another way of expressing the same
information. It is basically the change in the input required to produce a full
range of change in the output due to the proportional control action. It is
also referred to in simple terms as the percentage change of the input signal
required to change the output signal from 0% to 100%. The proportional band
answers the following question: “What percentage of change of the controller
input span will cause a 100% change in controller output?”

This adjustable setting of the proportional control mode can also be expressed
mathematically using the following formula [3]:

For 100% full span, the controller gain can be directly converted into the
proportional band using the following mathematical expression [3]:

The relationship between the controller gain and the proportional band is
further demonstrated in figure 3.6 and table 3.1 below [3].
Example 1:

A temperature controller with a 70% PB has an input range of 0 to 60 °C and

an output range of 5 to 30 mA. Calculate the controller gain.

Solution

Example 2:

A proportional only controller has a controller gain of 5. What is the percentage

steady-state error signal required to maintain a controller output of 40% when
the normal bias value is 0%?

Solution

Generally, providing a speedy response and stability to correct the process

error arising from load change is the primary action of the proportional control
algorithm. This is achieved by generating a controller output that is proportional
to the magnitude of the system error (see figure 3.6). The higher the magnitude
of the error, the stronger the corrective (proportional) action (see figure 3.7).
Proportional controllers are quite effective for simple control loop applications
where the process variable is not required to match precisely with the set point.
The effect of a change in controller gain on the response of the proportional
only controller is illustrated in figure 3.7 [3].

However, an optimum value of proportional gain should be determined. If the

controller gain is increased beyond a critical limit, it will induce instability in
the control loop (see figure 3.9) [3].
(c) Determining the controller output

In a proportional only controller, the output is a function of the change in error

and controller gain [3].

(Output change %) = (Error change %) (Gain)

Consider as an example a sudden set point change of 10% with a proportional

band setting of 50%. The output will change as follows [3]:

Gain and Gain =

=2

Gain

2 = 20%

(d) Proportional action – closed loop

Every process control loop has a critical or natural frequency. This is the
frequency at which cycling may exist. This critical frequency is determined
by all of the loop components. The control loop gain should not be either
too high or too low. At high proportional band, the loop gain is too high; the
process variable will cycle around the desired set point and cause instability
in the control loop (see figure 3.10). However, the process variable is still not
on set point. When the proportional band is high, the control loop gain is low.
This results in a very stable loop, but an error remains between the process
variable and the desired set point (see figure 3.11) [3; 4].

(e) Advantages of proportional control mode

Proportional controllers are simple to understand and exhibit the following

advantages [4]:
• The proportional control action is a type of linear feedback control mode
that is more sophisticated than the ON/OFF control system, but simpler and
easier to operate than a proportional-integral-derivative (PID) control system.
• The proportional action mode generates a fast response to set point or
disturbance change.
• It induces instantaneous change of magnitude proportional to the size of
the error entering the system in the controller output.
• It provides a very stable control process, under the condition of effective
selection of the controller gain (Kc).
• Proportional control action overcomes the instability caused by the ON/
OFF controller by modulating the output to the controlling device, such
as providing a linear output able to position the final control valve at an
intermediate position as well as the fully open or fully closed positions.

However, the use of this type of linear control mode has several drawbacks [4]:

• An extremely small controller gain could cause the final control valve not
to fully open or fully close.
• An extremely high controller gain would induce instability in the process
control loop.
• The proportional control action does not return the process control variable
exactly to the desired set point. It does, however, bring the process-controlled
variable back to a value that falls within a well-defined span (proportional
band) around the desired set point.
• Steady-state error (also known as offset) occurs when disturbances occur.
Offset is defined as a sustained error that cannot be eliminated by proportional
control, since it only responds to a change in the magnitude of the error
entering the system.

To fully understand the last disadvantage associated with the use of a proportional
only controller, consider the example of a control system monitoring the level
of a process liquid in a storage tank illustrated in figure 3.5 [4]. A proportional
only controller is used. The level of the process liquid in the storage tank remains
constant if the process liquid flows out of the storage tank at a constant rate. If
the flow of process liquid out of the storage tank begins to increase, the level of
the liquid in the tank will start to drop because of the lack of balance between
the inflow and outflow of the process fluid. This will generate a process error.
As the level of the process liquid drops in the storage tank, the error entering
the proportional controller increases. As a result, a greater controller output
proportional to the error entering the control system is produced.

Subsequently, the final control element, which in this case is the valve controlling
the inflow of the process liquid into the storage tank, opens wider to enable
more liquid to flow into the storage tank. The valve controlling the inflow of
the liquid into the storage tank will continue to open wider as the level of the
process liquid continues to decrease, until a point is reached where the inflow
of process liquid into the storage tank is equal to the outflow of the process
liquid from the storage tank [3; 4].

At this stage the tank level is not at the desired set point, but it remains constant,
meaning that a constant process error is now entering the control system.
Because a constant error enters the control system, the proportional controller
generates a constant controller output as a response to the constant error. This
causes the control valve to also hold its position and the system now remains
balanced, although the tank level remains below the desired set point. This
residual sustained error is called the offset, and it cannot be eliminated by the
proportional controller [3; 4].

Under proportional only control, the offset can only be eliminated manually.
This is done by an operator who manually changes the bias on the controller
output to reduce the offset to zero. Generally, the operator has to switch the
controller to manual mode first, in order to change the controller’s settings
manually until the steady-state error is reduced to zero. Afterwards the operator
can return the controller to automatic mode. This procedure is known as the
manual resetting of the proportional controller [3; 4].

Generally, the integral control mode is not commonly used. However, it is

very useful to understand the behaviour of this mode of control, as it provides
essential knowledge about the contribution the integral control action makes
when combined with other control modes. This mode of controller has to do
with another component of the system error, namely the duration of the error [4].

For how long has the error existed in the controlled process? The controller
output action from the integral mode is directly dependent on the duration
of the error. In this control mode the control system has the unique ability to
return the process variable to the exact desired set point by acting in such a
way that the control effort produced is proportional to the integral of the error.
The integral only controller in this case behaves as an integrator [4].

The integral control action is basically used to fully remediate system deviations
at any operating point. The integral only controller will generate an output signal
causing the manipulated variable to change as long as the error is non-zero.
The control process is considered balanced either when the desired set point
and the controlled variable are equally large, or when the manipulated variable
reaches its system-specific limit value (i.e., maximum allowable value) [3; 4].
The development of the automatic reset or the integral control mode, as it is
currently known, was inspired by the need to manually reset proportional only
controllers. As long as the process variable is not equal to the desired set point
(system error present), the integral control mode will continuously increase or
decrease the controller’s output action in order to reduce the system error to
the greatest extent possible. After a sufficient period of time, the action of the
integral only controller will drive the output far enough so as to completely
eliminate the system error (see figure 3.12) [4].

A fast controller output response is obtained when the system error is large, and
a slower output action is generated for a smaller system error. The controller’s
integral time variable (tI) sets, for a specific system error, the speed of the
integral action. A long integral time is characterised by a large value of tI and
results in a slow controller integral action. By contrast, a short integral time is
represented by a small value of tI, resulting in a fast integral action [4].

The integral time setting is an essential factor of the integral only controller.
If it is set too long, the controller will be sluggish. If it is set too short, the
process control loop will begin to oscillate and become unstable. The time
interval required to execute the controller algorithm is sometimes known as
the sampling time or scan time. Integral only controllers use integral time in
minutes as the standard unit of measure for integral control. Other controllers
use integral time in seconds as the standard unit of measure for integral control.
In the algorithm for the integral only controller, the integral gain is defined as
the repeats per minute or repeats per second (see table 3.1) [3; 4].
Mathematically the actions of the integral only controller are expressed by the
equation below: the value of the manipulated variable is changed proportionally
to the integral of the system error e [4].

where K I is the integral action coefficient and I(0) is the bias value.

From the mathematical expression above it is then clear that the magnitude of
the increment or decrement of the manipulated variable is directly dependent on
the system error e and the integral time tI (the reciprocal of the integral action
coefficient KI; in other words, a long integral time is equivalent to a small integral-
action coefficient). The higher the integral action coefficient K I, the greater the
integral action of an integral only controller; hence the lower the integral time
value tI. Conversely, the lower the integral action coefficient K I, the weaker the
action of the integral only controller; therefore, the higher the value of the integral
time tI [4].

Compared to proportional only controllers, the dynamic behaviour of integral

only controllers reflects a slower system response. The manipulated variable
of the process increases sluggishly in an integral only controller, whereas in a
proportional only controller, the manipulated variable of the system reaches
its final control value immediately. Consequently, the response of integral only
controllers to disturbances and step changes in the system with reference to the
desired set point is generally very slow. The value of the integral time could
be decreased in order to increase the controller response speed. However,
this should be done carefully, because if the integral time is too short after
adjustment, it will induce a rapid increment or decrement in the manipulated
variable, easily causing oscillation in the system and generating instability in
the process control loop at the end (see figure 3.13) [4].

For an integral only controller, it is fairly pointless to adjust an operating value

because the integral action component of the controller will rectify any set point
deviation detected in the system. It is therefore safe to state that the automatic
change implemented in the process-manipulated variable until the detected
system error is reduced to zero is comparable with an automated operating
variable correction. For a steady-state process (when e = 0) the manipulated
variable of the integral only controller would be at a value which would have
to be entered manually via the operating point adjuster in a proportional only
controller [4].

(a) Analysis of integral control action in open loop

In this control scheme, the purpose of the integral action is to return the
process variable to the desired set point. The integral only controller achieves
this by continuously repeating the proportional mode correction action for
as long as the error continues to exist in the system. Consequently, except
in some electronic controller devices, the integral or reset control algorithm
is usually combined with the proportional control algorithm. The controller
integral or reset action is sometimes characterised in terms of either repeats
per minute (how many times the proportional action is repeated each minute)
or minutes per repeat (how many minutes are required for one repeat to occur)
[4; 5].
(b) Analysis of integral control action in close loop

Process control systems involving a closed loop in the integral or reset algorithm
add one more gain component to the control loop. The faster the reset or integral
action, the greater the gain. Consider the example illustrated in figure 3.14. The
control loop appears to be stable because at the control loop critical frequency,
the total value of the control loop gain is not too high. In this example you
can see that the process variable reaches set point due to the reset action [5].

In the case where the value of the total control loop gain is too high at the
critical frequency of the control loop, the controller reset action will be too
fast. As a result, the process variable will be cycling around the desired set
point (see figure 3.15) [5].
(c) Windup

The integral or reset windup, also simply known as windup, is defined as

the situation in which a large difference between the desired set point and
the process variable causes the controller output to be driven from a desired
output level. This situation is very common in process plant shutdown. At
shutdown the process variable will probably go to zero, but the set point will
not change; consequently, this large generated error in the system will drive
the output to one extreme [5].

Similarly, at the start-up of a processing plant, significant overshooting of the

process variable may occur because the reset speed prevents the output from
reaching its desired value fast enough. However, it is possible to introduce
an anti-reset windup device into the controller. The purpose of the anti-reset
windup device in the controller would be to allow the output to reach its
desired value faster, therefore minimising the overshoot (see figure 3.16) [5].

where

IVP = controller output response

ARW = Anti-reset windup

(d) Advantages of the integral control mode

• A pure integral controller will respond to a constant system error by

generating an output action with an intensity that increases linearly with
time and with a gradient which is directly dependent on the size of the
system error and on the integral gain [6].
• It is very helpful to compensate for the fact that an error has existed for a
period of time [6].
• No error at steady state; as long as an error is detected in the process being
controlled, the integral action will always be adjusting the controller output
response in an attempt to eliminate the error. The integral action will stop
adjusting the controller output signal only when the system error is reduced
 to zero. Any controller that has an integral control mode would not permit
the system being controlled to experience steady-state offsets because these
will always at some point be eradicated by the integral component of the
controller algorithm [6].

(e) Disadvantages of the integral control mode

• At high integral time values the integral only controller generates a sluggish
output response [6].
• At small integral time the control loop tends to oscillate, which may induce
instability in the process control loop [6].
• Integral saturation/reset windup; in a real system there are lowest and
highest limits on the range of any manipulation that can be applied on
a process final control element. For example, a control valve cannot go
beyond fully closed or fully open – in other words, the valve cannot be
more open than fully open or more closed than fully closed. However, this
cannot be detected by an integral only controller. As long as the system
error is detected, the controller will continue to integrate the error even
though the controller output has reached its lowest or highest limits. This
state of the integral only controller is known as output saturation. When
the system error magnitude reduces, miscalculations occur in the controller.
The large error integral that has been built up will hold the controller output
against the limit for a considerable period of time. As a result, the measured
value significantly overshoots the desired set point. This behaviour of the
controller is termed integral saturation or (reset) windup. In extreme cases
reset windup can drive the controller output to oscillate between its lowest
and highest limits [6].

Example 1:

An integral controller has a value of KI equal to 0.5 s-1. If there is a sudden

change to a constant error of 10%, what will the output be after a period time
of 2 seconds if the bias value is zero?

Solution

Since e(t) is constant, it is safe to write:

Certain types of large and/or slow processes respond quite poorly to small
changes in the controller output signal. For example, a large liquid level
process or a large thermal process, also known as a heat exchanger, may react
very slowly to a small change in their respective controller output signals. To
improve the response of these controlled processes, a large initial change in
the controller output signal must be applied, and this can only be done by
means of the derivative control algorithm. A derivative only controller’s output
is fundamentally based on the derivative of the error detected in the system
and can be mathematically represented by the following expression [5; 6]:

From its mathematical expression it is clearly noticeable that the derivative

action changes the controller’s output only if the derivative of the error registered
in the system is not zero. In other words, the derivative action is initiated only
when there is a change in the rate of change of the error (see figure 3.17). This
simply means that in the situation where a constant error is registered in the
system, its derivative will be calculated to be equal to zero. As a result, the
derivative control action will not change the controller’s output despite the
existence of an error in the controlled system [6].

Consequently, the derivative only controller is not useful at all when it comes to
reducing errors in a controlled system, and is never used alone. The derivative
mode is sometimes called the rate, and has an adjustable setting called derivative
time (tD). The larger the derivative time setting, the more derivative action is
produced. Two units of measurement are used for the derivative setting time
of a derivative only controller, namely minutes and seconds [6].
Basically, the derivative only controller generates a corrective output change
that is proportional to the rate of change of the error detected in the controlled
process. If the detected error in the controlled process increases rapidly, a
large and positive derivative of the error will be calculated. In this case the
derivative algorithm will make a large positive change in the control action.
This will generate an effort of great magnitude that will then attempt to reduce
the future error detected in the controlled process [5; 6].

On the other hand, if the error detected in the system is decreasing rapidly, a
large and negative change will be implemented in the control action by the
derivative algorithm of the error. This could probably be attributed to an excess
control action that has already been applied, causing the measured value of
the process variable to significantly overshoot the desired set point. With the
large negative derivative value of the error, the derivative action generated will
induce a large negative change in the control action, with the main purpose
of reducing the rate of approach of the desired set point value [6].

The magnitude of the derivative action is determined by the setting of the

derivative only controller, which is expressed in terms of minutes. During its
operation, the derivative only controller initially compares the current measured
value of the process variable with the latest measured value of the process
variable. Afterwards, if there is a change in the way the measured values of
the process variable are fluctuating, the derivative only controller determines
what its output would be at a future point in time, which is obtained by the
value of the derivative setting, in minutes. Finally, the derivative control action
immediately increases the output by that amount [6].

Derivative only controllers can provide significant improvements to control

systems. However, they are not popularly used for the following reasons:
Derivative control actions are quite difficult to tune because they are only
effective in improving the response of the control system through its interactions
with the other control algorithms. The second drawback to the application of
this control mode is more serious, as it has to do with its sensitivity to system
noise. For this reason, it is recommended that derivative only controllers
be used on systems where any noise can be filtered out. This can easily be
achieved when the noise is of a totally different frequency from that of real
process changes [6].

For instance, in the control system of temperatures in large tanks, the noise
frequency is generally a multiple of the unit hertz, which is different from the
noise of the vessel real time constants of several hours. Therefore, it is possible
to completely eliminate the noise without really affecting the process signal.
However, in faster control systems such as flow control, the noise and real
time measurement frequencies change at relatively the same rate. As a result,
it is not possible to completely remove the noise without seriously damaging
the process measurement signal. In this situation, the use of derivative control
action is not recommended [6].

(a) Analysis of derivative control action in a closed loop

To understand the derivative control algorithm, consider this example of a

closed-loop temperature control system. In this scenario, to illustrate the action
of the derivative only controller in a slow process, the time scale has been
extended. Also, there is no reset time, and we can assume a proportional band
setting of 50%. A 10% change in set point is induced by a change in controller
output of 20% with an acting proportional gain of 2. Because the temperature
process is typically a slow one, the setting time after a change in error is quite
long. However, in this scenario, because of the absence of the reset setting,
the process variable never becomes exactly equal to the desired set point [6].

To study the effect of the rate of the derivative action, a time setting of 1
minute is considered. Initially at time t = 0, a fairly large controller response is
obtained, inducing a high output spike as a result of the rate of the derivative
action (see figure 3.18). It is important at this stage to remember that the change
in the output of the controller caused by the rate of the derivative action is
directly dependent on the rate at which the system error changes, which is in
a step nearly infinite [6].

(a)
 (b)

However, adding the derivative action rate only is not sufficient to cause the
process variable to match the desired set point. Now let’s imagine that the
derivative action setting time is increased to 10 minutes. This implies that the
controller gain is now much higher. As a result, the process variable will not
settle at the exact desired set point, and will start cycling. This is why it is
imperative to add a reset action, in order to bring the process variable back
to the exact desired set point (see figure 3.19) [6; 7].

Because the derivative action is directly dependent on the speed of change

of the input or error, its sole application on fast processes or processes with
noisy signals will be very unpredictable. The controller output, as a result of
derivative action rate, will have the highest change when the input changes
rapidly. However many controllers, especially the digital kinds, are designed
to respond to changes in the process variable only, and consequently ignore
changes in the desired set point. This feature of controllers eliminates a major
upset that would occur because of a change in the set point [6; 7].

(b) Advantages of the derivative control mode [6; 7]

• In derivative only controllers a rapid output response is obtained, which

reduces the time that is required to return the process variable to the desired
set point in a slow process. Basically, these types of controllers, unlike
proportional only controllers, generate the manipulated variable from the
rate of change of the system error rather than from the magnitude of the
change of the system error. Consequently, derivative only controllers respond
to changes in system error much faster than proportional only controllers.
Moreover, for small errors, the derivative only controller will produce in
anticipation large control amplitude as soon as a change in the magnitude
of the system error occurs.
• Because of the immediate control action obtained with derivative only
controllers, whenever there is a change in the error signal, the control
dynamics of derivative only controllers are significantly higher than those
of proportional only controllers.
• When combined with other modes of control, the derivative action tends
to maintain stability in the process control loop.

(c) Disadvantages of the derivative control mode [6; 7]

• Derivative only controllers do not recognise steady-state error signals,

regardless of the magnitude of the error. As long as the error signal is
constant, its rate of change is zero, and so it will not be identified by the
derivative only controller.
• It is quite a difficult control mode to implement and adjust. Its successful
application is limited to processes in which there is a considerable amount
of lag.
• The derivative only controller dramatically amplifies noisy signals, which
can cause too high a setting time cycling and extensive instability in fast
process control loops. These types of controllers are very sensitive to
measurements of very noisy signals.
• Some kind of signal filtering process is required within the software or
hardware implementation.

Proportional + integral controllers are by far the most common type of feedback
controllers utilised in the chemical process industry. The combination of
both control modes is obtained simply by connecting one proportional only
controller and one integral only controller in parallel, in that way generating
a proportional + integral only algorithm. The total output of the controller is
the sum of the output of the two individual control models (proportional +
integral). The proportional + integral only algorithm can be mathematically
expressed by the following expression [7]:
In this combined control algorithm, the controller gain affects both the
proportional and integral components of the process controller. The integral
time tI is the adjustable setting in this case when it is necessary to change the
relative balance between the two control modes. The dynamic behaviour of
the control loop is characterised by the proportional-action coefficient KP and
the reset/integral time tI (see figure 3.20). In the proportional + integral control
algorithm, the proportional action generates a fast and immediate response
to errors detected in the system and the manipulated variable is promptly
adjusted accordingly. P(0) is the bias or adjustable value of the controller [7].

Meanwhile, the integral action starts to gain influence only after some time
and eliminates any remaining steady-state error present in the system. In
proportional + integral controllers, the reset time represents the time required
until the integral component of the controller produces a control amplitude
that is equal to the one generated by the proportional component from the
beginning. Similarly to the integral only controller, the reset or integral time in
a proportional + integral controller must be reduced in order to generate an
amplified integral action component of the combined modes controller [7].

Generally, proportional + integral control is used where temporary error readings

that do not truly reflect the process variable condition (process noisy signal)
might exist, where no offset can be tolerated and where excessive time after a
disturbance before control action takes place (dead time) is not really an issue. In
processes where no offset can be tolerated, where no noise is present, and where
dead time is an issue, a proportional + integral + derivative control system is a
better alternative [7].

The proportional + integral controller, when properly designed, combines the

advantages of both control modes such as no steady-state error, rapid response
and stability and rapidity. As a result, the disadvantages of the two control
algorithms are compensated for at the same time. The proportional + integral
controller exhibits the following main operating features [7]:

• Most popular feedback controller

• No offset
• Better performance than proportional only controller
• Better dynamic response than integral only controller
• Quite complex to tune
• Possibility of instability due to time lag introduced

In proportional + derivative controllers the output response results from the

addition of the individual proportional and derivative components of the
controller and the algorithm of the combination of both control modes can
be expressed mathematically by the following equation [7]:

The derivative action coefficient KD is a measure of the influence of the

derivative component of the controller (see figure 3.21). A high value of the
derivative action coefficient implies strong derivative control action [7].

In a way similar to the mathematical expression of the algorithm for the

proportional only controller, the summand P(0) stands for the operating point
adjustment (bias value). It is the preselected value of the manipulated variable
which is issued by the controller in steady state when e = 0. The influence
of the derivative component increases at a rate that is directly proportional to
the derivative action coefficient KD [7].

Steady-state error occurs in proportional + derivative controllers just as it does

in proportional only controllers. However, because of the immediate derivative
control action generated, as soon as a change in the error signal takes place, the
control dynamics in the proportional + derivative controller are higher than in
proportional only controllers. Although changes in the controlled variable can
occur quite rapidly, the probability of oscillation occurring in the control loop is
significantly reduced due to the stabilising effect of the derivative component.
This also means that a higher value of the proportional action coefficient
KP can be safely set in the proportional + derivative controller. This can be
very useful in reducing residual steady-state error. Proportional + derivative
controllers generally exhibit the following characteristics [7]:

• Stable control loops when they are appropriately tuned

• Less offset than in the case of a proportional only control algorithm (the
use of higher gain is possible)
• Faster controller output response (reduced lags)

The proportional + integral + derivative controller, commonly known as the

PID controller, has a controller output which is made up of the sum of the
proportional, integral, and derivative control actions (see figure 3.22). The
mathematical expression of the proportional + integral + derivative algorithm
is given by the following equation [7]:

The proportional + integral + derivative controller provides a higher magnitude

of control action sooner than what is possible with other types of controllers.
This reduces the effect of a disturbance on the process and shortens the time
it takes for the level to return to the desired set point. The PID controller has
the following general characteristics [7]:

• Very complex type of controller to implement and operate

• Rapid controller output response
• No offset
• Quite difficult to tune
• This is the best process controller when it is tuned appropriately

The conventional type of proportional + integral + derivative control

algorithm is just one of many algorithms that exist. In the past individual
process manufacturers could have their own take on the algorithms of the
supplied controllers thanks to these built-in ideas. However, most modern
controllers manufactured nowadays provide manufacturers with a choice in
terms of the kind of algorithm the controller can implement. Generally, there
are no distinctive reasons for implementing one algorithm in preference to
any of the others available. It is usually just a matter of preference and which
algorithm the process control engineer is comfortable and familiar with [7].

PID controllers are used to process control loops with second- or higher-order
systems that require fast stabilisation and do not tolerate residual steady-state
error. The controller output response in PID controllers is advantageous in
higher-order controlled systems made up of components storing large amounts
of energy. This is because these systems require very rapid control action and
do not tolerate steady-state error [7].

Of the types of controllers we have already discussed, the PID controller exhibits
the most sophisticated control response. The controlled variable reaches the
desired set point promptly and stabilises within a very short period of time,
and oscillation occurs only slightly about the set point. The three control
parameters, namely proportional gain, integral time constant and derivative time
constant, offer tremendous versatility in terms of adjusting the control response
in accordance with amplitude and control dynamics (see figure 3.22) [7].

(a)

(b)

(a) Parallel or non-interacting algorithm

This conventional algorithm of the proportional + integral + derivative is

mathematically represented by the following expression [6; 7]:
In this algorithm, the proportional, integral and derivative components of the
controller have their own independent gains. Adjusting the gain of any control
mode will not have any effect on the other two modes [6; 7].

(b) Series algorithm

This form of the PID algorithm is expressed mathematically by the formula

below:

The series algorithm of the PID is considered a historical artefact, although

mathematically it seems a little bit messy. However, it can be implemented
using a single electronic or pneumatic amplifier. The disadvantage associated
with this form of the PID algorithm is that all three control modes interact
with each other at the same time. This makes it difficult to relate the output
response of the controller to any individual control action, and it is therefore
quite difficult to tune [6; 7].

(c) Algorithm tweaks

In some controllers using the PID algorithm, a number of both minor and
significant tweaks need to be made. The derivative action can be tailored to
be based on the derivative of the measured process variable instead of the
system error. This overcomes the issue associated with the standard derivative
control algorithm (derivative of system error) in situations where a step change
in the desired set point is introduced (infinite error derivative!) [7].

In practice, derivative control action is usually implemented with a built-in

first-order filter. In the past this was the only way that the derivative action
could be successfully implemented, and so it was very important. This is
because in the past analogue rather than digital mechanisms were utilised, and
analogue mechanisms did not have the operating feature of being able to save
old values in the way that digital systems can. Currently, modern controllers
use the filter for protection against sudden fluctuations in the desired set point
when the derivative is based on system error and spikes in the measurement
of the process variable, even though the measurement of the process variable
should already be filtered sufficiently before reaching the process controller [7].

The proportional control action can also be tailored. Sometimes use of the
square of the error is considered. However, it significantly increases the change
when large system errors are involved, and reduces the proportional control
action when dealing with smaller system errors. As a result, this generates a
non-linear control algorithm that can be quite complex to tune [7].
To successfully implement the classical PID algorithm in digital control systems,
some alterations are required. The operation of digital control systems is
based on sampling the process. This simply means that the measurements are
numbers representing snapshots at precise time intervals of the process state.
The analogue PID algorithm can be converted into two digital equivalents [8]:
(a) The velocity algorithm – this is the most commonly utilised digital
equivalent of the PID analogue algorithm. It is represented by the following
mathematical expression [8]:

where
COk and COk-1 = controller output at the current kth and (k-1)th samples
respectively

ek and ek-1 = system error at current kth and (k-1)th samples respectively

ts
= sample interval time between successive snapshots for the controller

tI = integral or reset time setting

tD = derivative time setting

The above digital PID control algorithm basically expresses the magnitude of
the change necessary in the controller output response, and hence should be
added to the previous controller output response. Moreover, with this algorithm
a smooth transfer from manual to automatic operation is easily feasible. This
is the reason why its application is very popular. In most control processes,
new process variables are used when operating under manual control, and
once the process gets near to its desired state the operator then switches the
controller over to automatic mode [8].
The velocity algorithm does not require a zero initial error to avoid a sudden
change in the controller’s output response. This is because in this case the
proportional component of the controller operates in response to the change
in the system error. Furthermore, the integral component of the controller,
when tuned to be drastically less aggressive than the proportional action,
will gradually change the controller output response in order to eliminate the
detected system error [8].
(b) The positional algorithm – This alternative form is a direct conversion of
the analogue PID algorithm into its digital equivalent, and produces the actual
value of the controller output [8].
COk = controller output at the current kth sample
CO(0) = bias value
ek and ek-1 = system error at current kth and (k-1)th samples respectively
ts = sample interval time between successive snapshots for the controller
tI = integral or reset time setting
tD = derivative time setting
The positional algorithm operates successfully when the initial system
error is exactly at zero. When the initial error is not exactly zero, the
proportional component of the controller will cause a sudden change in the
controller output; this will be conveyed to the process, generating a rapid
disturbance known as a “bump” [8].

The successful selection of an effective controller is closely dependent on

the parameters of the system to be controlled. A thorough assessment should
be conducted before a specific type of controller is considered suitable for
the process it is supposed to control. To achieve the control objective, it is
necessary that the controlled system first of all be analysed, and a suitable
controller then be selected and designed. The most important properties of
the widely used proportional, integral, proportional + integral, proportional
+ derivative and proportional + integral + derivative control modes can be
summarised in the following table [9]:

The following factors should be considered during the selection of a

controller [9]:

• With or without self-regulation, is the system based on the integral or

proportional control mode?
• How high is the process lag in terms of time constant or/and dead time?
• How quickly must system errors be corrected?
• Is steady-state error acceptable in the controlled process?

In any industrial process, the ultimate goal is to generate a final product that
meets the desired specifications. Consequently, various process control loops
are implemented within the process to ensure that it operates in accordance
with the design criteria in order to convey the desired characteristics to the
end product. The evaluation of the performance of a process control system is
based entirely on whether or not the control delivers a product that is within
design specifications. The entire process will be pointless if the desired quality
of the end product is not achieved [9].
In brief, assuming that in any real industrial situation the control system
always delivers a final product that meets the design specifications, it is
nevertheless essential to understand how well the system performs the task.
It is also fundamental to study the fluctuations in process parameters and the
percentage of rejected production batches associated with routine operations.
The performance of a control system is evaluated by measuring the quality of
the controller’s response to system disturbances and then analysing how the
control loop characteristics influence these measurements. There is no absolute
conclusion when analysing the performance of a control system, because a
control system that is considered suitable for one industrial process might be
unsatisfactory for another [9].

A typical control loop is a basic requirement in almost all industrial processes.

This is attributed to the fact that the process variables are dynamic in nature
and therefore change easily under the influence of various disturbances.
Consequently, monitoring and regulatory actions are needed to keep the process
variables at their respective design values. However, in industry it is not really
practical to bring back the process variables to precisely their design values,
and so the objective is to bring them as close as possible to their respective
desired set points [8; 9].
Moreover, the fact that initially the process variable will deviate from the set
point before the controller can generate an appropriate corrective response
to counteract the error in the system is inevitable. As a result, it can safely be
said that perfect control is impossible because at some point process variables
are going to fluctuate from their design values. The practical definition of
control quality refers to the evaluation and measurement of these deviations
that prevent the process from generating the desired end product [9].
A typical process control engineer does not possess the necessary skills to
evaluate how the deviation of the process variable from the desired set point
will affect the overall process in terms of (i) affecting the specifications of
the end product, (ii) inducing end product rejection or (iii) influencing the
cost effectiveness of the overall process. Consequently, a set of conventional
criteria or measurements has been developed to establish a common language
between product experts and process-control experts. This set of measures
provides a communication link that serves as a control quality measure, relating
the dynamic characteristics of a control loop to the specifications of the final
product of the process. To understand how the performance of a control
system is measured, it is important first to define control quality in terms of
the characteristics of the control loop [9].

(a) The loop disturbance

Because a typical control system has the main purpose of generating regulating
actions in order to prevent disturbances in the system from causing significant
fluctuation of the controlled variable from the desired set point value, the
quality of the process-control system can be defined by the extent to which
the fluctuations that are induced by the disturbances are minimised.

Three basic kinds of disturbances can usually occur in a typical control system.
These are:

(1) load change disturbances

(2) set point change disturbances
(3) transient disturbances

The first possible type of disturbance is a step-function change in the process

load that occurs instantaneously. The load change could be induced by a rapid
permanent change in any of the parameters of the process that contribute
directly to the overall process load. The second type of disturbance refers to a
step-function change in set point. This is an instantaneous change in the loop
set point from an old value to a new value. Finally, a transient disturbance
is the result of a momentary variation in any process variable that affects
the process-controlled variable. The quality of the control system in place is
measured by evaluating the way in which the control system responds to both
load and set point (sudden) changes. In practice, steady and regular types of
disturbances are useful to define the performance of a control system. Thus,
transient disturbances, which can easily vary in amplitude, shape, duration
and peak, cannot be used to define the level of quality of a typical control
system [8; 9].

(b) Optimum control

Universally, the best control quality obtained from any system is referred to as
optimum control. When disturbances are introduced into the process, the most
negative influence involves the deviation of the process-controlled variable from
either the process load or the desired set point. Therefore, the main purpose
of setting the control system is the provision of optimum control. The control
would not necessarily be perfect, but should be as good as it can be. For this
reason, the ability of the system to reach optimum control can represent the
control quality of this system. The following three effects resulting from set
point and process load changes can be used to define the control quality of
the system [9]:

• stability
• minimum deviation
• minimum duration
The quality of control provided by a control system is arrived at through the
evaluation of stability, minimum deviation, and minimum duration following
a disturbance of the dynamic controlled variable [9].

(c) Stability

The capacity of a control loop in terms of providing stable regulation of the

dynamic controlled variable is the most fundamental property used in defining
control quality of the system. In most cases the disturbance causes the dynamic
variable to either simply increase linearly without limit or to execute growing
oscillations with amplitude increasing infinitely. Thus, stable regulation by the
control system prevents the dynamic variable from growing infinitely. However,
in special cases, such as in a two-position controller, the controlled variable
oscillates between two limits under nominal load conditions. The controlled
variable may be cyclic, but it is stable. A change in the process load may
influence and hence change the period of oscillations, but if the amplitude
swing remains essentially the same, the variable is under stable control [9].

(d) Minimum deviation

In a control system where the primary objective of the output response of

the controller has been calibrated to adjust the process controlled variable to
a desired value (set point), the quality of the control provided by the system
is represented as the extent to which a disturbance of the process will result
in a deviation of the controlled variable from the set point. In the case of a
disturbance registered as a set point change, the quality of the control provided
by the system is regarded as any undershoot or overshoot of the controlled
variable in reaching the new set point. In this case the general objective is to
minimise any deviation of the dynamic controlled variable from the desired
set point value [9].

(e) Minimum duration

Generally, a disturbance occurring in a process will induce some deviation of

process variables. The length of time required before the controlled variable
achieves set point value or at least falls within the acceptable range of that
desired value after the disturbance represents another factor used to define
the performance of the process control loop [9].

In practice, adjusting a process control to provide a stable operation with

minimal deviation and minimum duration all at the same time is an idealistic
aim that is quite difficult to achieve. For instance, if a stable operation is
obtained with minimum deviation, a less than minimal duration usually occurs;
alternatively, the system could be adjusted so as to achieve a faster production
rate at a tolerable reduction in the final product specification. Consequently,
various types of measurement are used to accommodate such scenarios and
illustrate how close the control circumstance is to ideal conditions. Consider
a controlled process where stable operation has been achieved. Under the
influence of a disturbance to the process, the dynamic controlled variable in
the process-control loop can respond in one of three ways, depending on the
controller gains and lags [9].

(a) Overdamped

In this case, after a disturbance is introduced into the process, the deviation
approaches the desired set point value gradually without oscillations. However,
minimum duration and minimum deviation are not successfully achieved. This
type of control response is considered safe, as it ensures that no instabilities
occur in the process-control loop and that rare cases of maximum deviation
never take place [8; 9].

(b) Critically damped

This control response is the result of a careful adjustment of the process-control

loop. This is generally considered an optimum control response in the case
where no overshoot, no undershoot or no cycling is desired following a set
point change. In this case, minimum duration is achieved for a non-cycling
response [8; 9].

(c) Underdamped

Once a critically damped response of the control loop is obtained, if it is

adjusted further, a cyclic response will result, with the deviation executing
a number of oscillations about the desired set point. Attaining minimum
deviation and minimum duration is feasible in some cases. If the cycling can
be acceptable in the control loop, this kind of response can be tolerated [8; 9].

Although controllers are meant to operate successfully for quite a while after
being properly tuned, it is very common for their effectiveness to decrease
with time. Phenomena such as catalyst degradation in reactors or fouling
can completely modify the originality of the process. As a result, the gains
and dynamic parameters of the process controller response may no longer be
suitable. It will then be necessary to change the controller’s tuning parameters in
order for them to be appropriate for the new process conditions. Furthermore,
other factors such as change in production rate, change in final product grade
or change in properties of raw material(s) can also completely redefine process
conditions [8; 9].

These changes in process conditions would initially affect the gains and dynamic
parameters of the controller and reduce control system performance gradually.
However, the effects will be noticeable only at high intensity. For this reason,
modern scanning control and data acquisition (SCADA) systems are designed
in such a manner that they are able to monitor the performance of the control
loop in a chemical plant and generate warning alerts when potential problems
arise. SCADA systems generally monitor the following features of typical control
loops implemented in most manufacturing plants [8; 9]:
• The length of time the loop remains in automatic mode. When the loop
remains in manual mode for a fairly long time, this is a clear indication that
the performance of the control loop is unsatisfactory [9].
• All process controllers can potentially reach saturation. Saturation happens
when the controller output response reaches the system’s design limit. As
soon as the controller is saturated, it is no longer able to monitor the process.
Consequently, if the controller remains accurate for an extended period
of time, the control response action is no longer accurate for the specific
process. In this case it is possible that the final control element (valve) may
need cleaning or resizing [9].
• The number of oscillations in the output response of the controller can be
determined using a standard unit of measurement known as the oscillation
index. Significant changes in the oscillation index are a straightforward
indication that some disturbances are driving the control loop out of tune [9].
• Various evaluation techniques, such as standard deviation calculation, can
be used to measure the inconsistency of the process manipulation and
controlled variable over an estimated period of time. Because any process
is generally assumed to operate at approximately constant conditions, any
changeability of the controlled variable is a strong indication of the presence
of disturbances in the process-control loop [9].
• Process steady time period can also assist when evaluating the consistency
of the controlled variable. However, this measuring method is effective
only for processes operating at roughly constant conditions over time [9].
• Recording the activities of plant operators could also be considered as a
technique for evaluating the performance of the process-control loop. If the
process set point has been adjusted multiple times and the controller has
been switched over from automatic mode to manual mode, the increasing
attention given to the control loop is a clear suggestion of the existence of
potential disturbances in the process-control loop [9].

Once a potential problem has been detected in the process-control loop, it

is important first to eliminate the other possible sources before thinking of
retuning the controller. Generally, other sources of the problem could be any
of the other components of the control loop, such as interference in one of
the communication lines, any fault in any of the process instruments, a faulty
control valve or calibration required in one of the instruments. Readjusting the
tuning parameters of the controller might cover up other existing problems for
a short while, but will not eliminate them completely. Only after a thorough
examination of the other reasonable causes of changes in the process could
a decision be taken to retune the process controller [8; 9].

In theory it is often assumed that the loop response is stable for the application
of process control principles. In practice, a great deal of expertise in multiple
engineering fields is applied in the design and development of process-control
loops to achieve this stability [8].
(a) Source of instability – Occasionally, the transfer function of the process-
control loop is such that feedback to the error summer actually increases the
system error because of the phase shift and the gain of the control loop. If there is
any frequency for which this condition exists, then oscillations will undoubtedly
start and grow at that frequency. The control system is therefore unstable, and
something like random noise will eventually maintain the control system into
growing oscillation. Consequently, in the design of a process-control system,
emphasis should be placed on its ability to regulate the process-controlled
variable without instability occurring in the control loop [8].

One way of deriving the stability criteria of a control system is by defining

the exact process conditions in terms of system gain and phase lag that are
capable of increasing the intensity of the system error. The following conditions
apply [8; 9]:

• The gain must be greater than 1

• The phase shift (lag) must be –180°

This simply means that at the frequency at which the system’s gain is greater
than one and the phase shift is –180°, the system is considered unstable.
From this reasoning, in order to determine when the control loop is stable,
the following two basic rules have been developed [8; 9]:

• A process control system is stable if and only if the system gain is less than
one (unity) at the frequency for which the phase shift (lag) is –180°.
• A process control system is stable if the phase shift (lag) is more positive
than –180° at the frequency for which the gain is one (unity).

The practical application of these rules to an actual process requires determining

the system’s gain and phase shift for all frequencies to verify whether both
rules are applicable. This is easily done if a plot of gain and phase versus
frequency is provided. A specific type of diagram known as a Bode plot is
generally used to plot the gain and phase of a typical control system versus
the system’s frequency (see figure 3.23). The frequency is usually expressed
as rad/s and commonly plotted on a log scale along the horizontal axis. This
makes the display of a large range of frequency possible. In this case we have
two vertical axes: the system’s gain is plotted on a log scale on one axis, while
the phase lag is linearly plotted as degrees on the other vertical axis (see figure
3.23) [9].

Another mathematical approach commonly used to assess the stability criterion

of a process control is the sign of the roots of its characteristic equation. The
characteristic equation of a process-control loop is generally as follows [9]:

CE = 1 + G(s)K

where

CE stands for characteristic equation

G(s) is the control loop transfer function
K is the controller overall gain
A control system is considered stable if all the roots of the characteristic
equation (CE = 1 + G(s)K = 0) are negative or if they all have negative real
parts in cases where they happen to be complex numbers. Otherwise the
control system is said to be unstable [8].

Controller tuning refers to the adjustment of the parameters of a controller to

suit a particular controlled process. The initial approximation of the tuning
parameter of a PID controller can be carried out in various ways. However,
none of these tuning techniques is completely fool proof, and even the best
procedure for successfully tuning a controller involves some trial and error.
Controllers operating without any tuning procedure are commonly found in
industry. In this case, the appropriate control action will have initially been
set by the instrumentation engineers, and the controller operates under the
default factory settings. This is because tuning a controller is a time-consuming
task that usually requires highly skilled engineers, who may be difficult to find.
Nevertheless, efficient controller tuning often costs less than what it costs to
repeat the task over and over [9; 10].

The selection of an effective controller is a critical aspect to factor in for

a successful control system result. The proper adjustment of the controller
parameters (proportional gain, reset/integral time and derivative time constants)
to the controlled system response is of even greater importance. However,
adjusting the controller components is usually a balancing act between a
fairly stable and very slow control loop. The compromise can also be made
between very dynamic but irregular control responses which easily result in
oscillation, causing the process control loop to be unstable at the end [9; 10].

In the case of non-linear systems that need to be monitored so as to always

operate in the same operating point, in other words, at a fixed desired set
point control, the components of the controller in place should be fitted to
the system’s response at this specific set point control. Conversely, when a
particular fixed control set point cannot be fixed or defined, as in the case of
follow-up systems, the parameters of the corresponding controller should be
adapted to provide sufficiently fast and stable control action within the entire
process operating range [9; 10].

In practice, the process controllers are tuned to determine the following [10]:

• The appropriate magnitude of the corrective action: how much correction

should be made? The required change in controller output is determined
by the proportional mode of the controller.
• The period of time required to adjust the controller output response: for how
long should the correction be applied? The duration of the adjustment to
the controller output is determined by the integral mode of the controller.
• The rate at which the controller output response should be adjusted: how
quickly should the correction be applied? The speed at which a correction
is made is determined by the derivative mode of the controller.

When it comes to proportional + integral + derivative controllers, there are a

number of stages involved in setting up this particular controller for a specific
application [10].

(1) It is quite important to define the direction of the control action. The
direction of the control action is usually directly dependent on the process
gain and the type of final control valve. Regardless, the control switch in
the controller in place should be set appropriately (in either the “direct”
or “reverse” position).
(2) Of the three available control modes, the most useful to the particular
process needs to be selected. The simplest choice would be a proportional
control action; however, there is an issue of residual offset. The residual
offset issue can be overcome with a PI controller, but this controller is
rather complex to tune. The best control results can be obtained easily
with the PID controller. However, a noise filtering system should be
implemented. If derivative control action is a component of the controller,
then extreme care needs to be taken with regard to the system noise
before tuning takes place.
(3) The parameters of the controller need to be appropriately adapted to the
controlled process. Low proportional gain usually results in high process
gain. An integral time constant of a high value will generate long time
constants in the controlled process. A controller is generally tuned by
means of the trial-and-error method.
Regardless of the type of controller, a tuning approach that is commonly used
is a method first proposed by Ziegler and Nichols, known as the ultimate
tuning technique. It makes possible simple tuning that can be applied in many
controlled systems that allow sustained oscillation of the controlled variable.
During tuning using the Ziegler-Nichols method, the values of the proportional,
integral and derivative parameters of the controller are determined from the
ultimate gain and the ultimate period of the process control loop. To determine
these ultimate parameters, the closed control loop is disturbed; the disturbance
response of the system is investigated, and the values of the ultimate constants
are extracted [9; 10].

Basically, the controller must first be switched to automatic mode and set up
for proportional only control. After this, a series of slight set point changes
are introduced into the system and the control loop response is examined.
The primary objective is to obtain a sustained oscillation as the control loop
response by fluctuating the controller gain. If a set point change causes the
oscillation of the loop’s response to disappear, the controller gain should be
increased. As new oscillations start growing (meaning that the control loop
is starting to be unstable), the controller gain should be decreased rapidly.
As soon as sustained oscillations are achieved, the controller gain is noted as
“ultimate gain” and the period of oscillation can also be measured as “ultimate
period”. The steps below should be carefully followed when applying the
Ziegler-Nichols tuning procedure [9; 10]:

• At the controller, set KP and tD to the lowest value and tI to the highest value
because it has the smallest possible influence on the controller output.
• Adjust the controlled system manually to the desired set point and start up
the process-control loop.
• Set the manipulated variable of the controller to the manually adjusted value
and switch the controller to automatic operating mode.
• Increase the value of the proportional gain until the point of instability
is reached and sustained oscillations occur in the control loop (constant
amplitude limit cycle takes place).
• Determine the ultimate time period tu and the ultimate control loop gain
Ku from the constant amplitude limit cycle. The ultimate time period tu is
obtained by determining the time span for one full oscillation amplitude and,
when necessary, by taking the time of several oscillations and calculating
their average.

Calculate the tuning parameters using appropriate equations summarised in the

table provided below [9; 10]:
• Multiply the values of Ku and tu by the values according to the table above
and enter the determined values for KP, tI and tD at the controller.
• When necessary, readjust KP and tI until the control loop shows satisfactory
dynamic behaviour.

One of the difficulties associated with the Ziegler-Nichols tuning technique is

the potentially aggressive resulting control action. Tyreus and Luyben suggested
an alternative way of extracting controller parameters (see table 3.4) [9; 10].

The Ziegler-Nichols technique is a fairly attractive technique when the process

remains under control during the operating period. In practical situations the
method is rarely applied due to the following drawbacks associated with its
application [9; 10]:

• Most processes have slower dynamics and it usually takes longer to obtain
substantial oscillation patterns for use in evaluating whether they are growing,
shrinking, or steady and consistent.
• All chemical processes are always subject to random disturbances, which
will after a while definitely interfere with the control process.
• Because the technique is based mostly on inducing instability in the control
process, without proper monitoring it can cause serious problems in the
control loop.
 When we are dealing with an open-loop control system, the information for
tuning can be obtained by conducting a step test. This is achieved by switching
the controller over to manual mode and inducing a small step change in the
controller’s output. This results in a controller output response from which
the process gain kp, process time constant and process dead time td can be
extracted. Therefore, the controller’s parameters can be calculated using the
mathematical expressions set out in table 3.5 [9; 10].

Subsequently, Cohen and Coon derived a more complex mathematical

expression that can be used to calculate the controller’s parameters from the
process gain kp, process time constant and process dead time td (see table
3.6) [9; 10].
This tuning technique is a model-based control design called international
model control (IMC). It has become quite popular in industries around the
world, especially in the USA, because it has been reported to yield excellent
performance. A number of advantages are associated with the application
of the lambda tuning method. This model-based control technique makes it
possible to design a system controller that will produce a specific response
instead of correcting an undesired controller output response after installation
by modifying the controller’s parameters. A conventional model-based design
would require high-order, accurate models, but the IMC tuning method
approximates the model of the process by using the first-order plus dead time
(FOPDT) model [9; 10].
The approximation procedure used in this model is quite similar to the one
used in the Cohen-Coon and Ziegler-Nichols tuning techniques. However, the
former and the latter produce tuning parameters with the primary purpose of
minimising some error measured in the system. The lambda tuning method,
on the other hand, produces tuning parameters that are meant to always
generate an overdamped controller response to set point changes. Therefore,
it is still necessary to determine the process gain Kc, process time constant
and process dead time td. An additional design constant, , enables the control
engineer to alter the controlled response to suit different applications [9; 10].
The first step in the process tuning that utilises the lambda tuning method requires
the control engineer to perform a step change test in order to approximate the
model gain Kc, process time constant and process dead time td. Afterwards,
it is necessary to select the value of parameter . When the approximated
model is perfect, meaning that it is a good representation of the real process,
is basically the time constant of the closed control loop. In this case (but only
in this case!) the controller’s response to a set-point change will be complete
after 3 to 5 . The efficiency of the lambda tuning technique can be seriously
affected if errors occur in the process modelling, resulting in controller’s
responses with oscillations. As a prevention method, large values of lambda
are commonly used to provide some sort of robustness to the design of the
control system [9; 10].
This will enable the controller of the closed control loop to generate output
responses predictably despite the presence of fairly large errors in the process
model. It is usually recommended that the value of should be three times either
the approximated process time constant or the process dead time, whichever
is the larger value. However, over the years numerous authors have suggested
modifications to the conventional lambda tuning method to suit different types
of industrial processes. Regardless of the kind of process, the closed-loop
response is always slower than the open-loop response. Once an appropriate
value of has been selected, the following mathematical expressions are used
to determine the parameters for the PI controller tuning [9; 10]:

tI =
where Kc = controller gain
tI = integral time constant
= lambda
Kp = process gain
= process time constant
td = process dead time

The lambda tuning method has the following advantages [9; 10]:

• Because of the slow set point response, the changes in set point have a
negligible effect on other control loops in the process. This feature of the
lambda tuning technique has made it very popular in the paper industry.
• With the lambda tuning method it is possible to synchronise multiple control
loops in order to obtain similar closed-loop responses.
• By allocating smaller values of to important control loops, precedence
can be given to them.
• The lambda tuning technique is robust and tolerates fairly poor process
testing and model fitting.

However, various drawbacks are also associated with the application of this
tuning procedure [9; 10]:

• The controller’s response is extremely slow – in fact, it is slower than the

open-loop response.
• Because of the really slow response, the elimination of the process’s
disturbance is almost ineffective.
• For non-standard processes (oscillatory open-loop processes, integrators,
etc.) it is necessary to apply special tuning rules.

In practice, numerous controllers available in the market come with an auto-

tuning feature. Generally, this feature has to be activated to run for a specific
period of time and then stop. Controllers that alter their parameters to match
varying process conditions on a continuous basis are referred to as “adaptive
controllers.” They are most common in academia, and are rarely used in industry.
The use of an adaptive controller in industry requires absolute confidence
that the algorithm of the controller will adapt efficiently to changing process
conditions [9; 10].

From basic tuning methods (automatic version of the Ziegler-Nichols) to expert

tuning systems (evaluating controlled responses at an expert level), there is
a wide range of auto-tuning methods used by various manufacturers. Auto-
tuning can be a useful tool for the calibration of controller devices; however,
for the reasons I will explain below, it cannot be considered to be a complete
solution to controller tuning problems [9; 10].

Like the manual tuning methods used in closed loops and open loops, the
mathematical expressions used to calculate controller parameters from response
constants are not guaranteed to give the best results in all situations. In process
industries, interacting systems (systems with multiple manipulations and
controlled variables) are very common. The various manipulations in these
systems can affect the numerous controlled variables. Therefore, the changes
in controller behaviour will depend on whether the controller is in automatic
or manual mode. Unfortunately, automatic tuning does not take such situations
into consideration [9; 10].

This tuning technique requires someone with expert skills to experiment

with the control loop by modifying the tuning parameters until the desired
control loop performance is achieved. Expert tuners are potentially able to
generate good control loop performance in less time than that required to run
the Ziegler-Nichols tuning procedure or reaction curve tests. Moreover, the
tuning of complex control loops always ends with some manual adjustment.
This is attributed to the fact that more tailored tuning can be achieved with the
help of experts compared with the general algorithms used in conventional
tuning methods or the auto-tuning algorithms. Various techniques are used,
depending on the expert performing the manual tuning. One of the commonly
used manual tweaking procedures is set out below [9; 10]:

(1) Set up the controller in proportional-only mode.

(2) Modify controller gain until approximately ¼ decay ratio on small set
point changes.
(3) In the case of unacceptable offset, reduce the controller gain by
approximately 20% and add integral action.
(4) Modify the controller gain and integral time constant according to the
type of controller response obtained.

Special considerations are required for derivative control. For the adjustment
of the controller gain and integral time constant in step 4, the map below (see
figure 3.24) could be quite useful [9].

From the given tuning map, you can see clearly that the ideal tuning is located
at the centre of the map. This region of the map is characterised by fast rise
time, settling time and quarter decay ratio. At the bottom left of the map
the controller gain is too low, and as a result the sudden change in output
response to set point change is small. Meanwhile, the integral time is too long
– in other words, it takes too long for the controlled variable to rise to the set
point. The controller gain at the top left of the map is too high. This indicates
that instability has been introduced and that the process has been driven into
sustained oscillations. However, the integral time constant is still too long, and
so the centre of the oscillations takes a long time to centre around the new
desired set point. The right side of the tuning map can be clearly interpreted [9].
In practice, chemical plants are complicated systems, usually with numerous
control loops operating simultaneously. The selection of an effective controller
has a significant influence on the successful operation of the control system.
Therefore, the suitability of a certain type of controller must be thoroughly
investigated to accommodate the process it controls [1–4].

Proportional only controllers are generally utilised in fairly simple control

systems where residual steady-state error (offset) is tolerated. With this type of
controller stable and dynamic control response is reached at minimum effort.
Meanwhile, integral only controllers are most suitable in systems with lower
requirements in terms of the control dynamics and where the system does
not exhibit significant lag. The advantage of this controller is that the system
errors are completely eliminated [5–7].

Proportional + derivative controllers, on the other hand, can be suitable

for applications in systems with significant lag and where residual steady-
state errors are acceptable. The derivative only component of the controller
increases the rate of response so that control dynamics improve compared with
those of proportional only controllers. In proportional + integral controllers
the advantages of both the proportional only controller and the integral only
controller are combined and their characteristic disadvantages are overcome
too. A dynamic control response is generated by this mode of control action
without residual system offset [7–9].

Usually, most process control tasks can be achieved with this type of controller.
However, when the system requires as fast a speed of response from the
controller as possible despite the presence of significant lag, a proportional +
integral + derivative controller will be more suitable. Proportional + integral +
derivative controllers are quite useful when it is necessary to eliminate significant
lag in the controlled system as quickly as possible. The derivative component of
the controller improves the control dynamics obtained with just the proportional
+ integral controller, while the integral component completely eliminates
any residual steady-state errors in the system. The operating properties and
characteristic equation of the various controllers are summarised in tables 3.4
and 3.5 below [8–10].
The control system is the central point of any chemical plant. It provides sensing,
analysis, and control of the various processes taking place in the plant. When
the control system is properly tuned, the effects of process disturbances are
reduced, efficiency is maximised, operating costs are minimised, and production
rates are increased. Controller tuning refers to the process of selecting suitable
tuning parameters to ensure the best controller output response [9; 10]. When
the tuning procedure is too slow, the controller’s output response becomes
almost inactive, making the controller unable to handle upsets, and the system
takes longer to reach the desired set point. Conversely, if the tuning process is
too aggressive, the control loop overshoots and becomes unstable. Controller
tuning can be achieved rapidly and accurately using efficient techniques.
However, in practice engineers often believe that controller tuning is “part art,
part science”. For this reason, many engineers and technicians combine proven
scientific tuning methods to the “tune by feel” conventional practice [9; 10].

(1) Anderson, N.A. Instrumentation for Process Measurement and Control,

Chilton Book Company, Radnor, PA.
(2) DIN 19226 Parts 1 to 6 “Leittechnik: Regelungstechnik und
Steuerungstechnik” (Control Technology), Beuth Verlag, Berlin.
(3) Distefano, J.J., Stubberud, A.R. and Williams, I.J. Feedback Systems and
Control, Schaum’s Outline Series, 2nd ed., McGraw-Hill, New York,
p. 78, 1995.
(4) “International Electro-technical Vocabulary”, Chapter 351: Automatic
Control. IEC Publication 50.
(5) Luyben, M.L. and Luyben, W.L. Essentials of Process Control, McGraw-
Hill, New York, NY, 1997.
(6) Marlin, T.E. Process Control: Designing Processes and Control Systems
for Dynamic Performance, 2nd ed., McGraw-Hill, New York, NY,
2000.
(7) Murrill, P.W. Fundamentals of Process Control Theory, Instrument Society
of America, Research Triangle Park, NC, 1981.
(8) Ogunnaike, B.A. and Ray, W.H. Process Dynamics, Modelling and Control,
Oxford University Press, Oxford, 1994.
(9) Seborg, D.E., Edgar, T.F. and Mellichamp, D.A. Process Dynamics and
Control, 2nd ed., Wiley, Hoboken, NJ, 2004.
(10) Terminology and Symbols in Control Engineering Technical Information
L101EN; SAMSON AG.
3.1 Briefly discuss the difference between continuous controllers and discrete
controllers.
3.2 Describe the characteristic mechanism of the three main control modes.
Briefly state the advantages, disadvantages and applications of each control
mode.
3.3 Discuss the operation of the proportional + integral (PI) control action, the
proportional + derivative (PD) and proportional + integral + derivative
(PID) control action.
3.4 Briefly discuss the main objective of the controller tuning process.
3.5 Consider a process with a desired set point of 10.5 V. The measurement
of the process variable obtained from the instrument is 13.2 V. What is
the process error?
3.6 If the reading of the process variable was obtained using the standard
measurement with a range of 3 V to 15 V, what will the percentage error
of this particular process be?
3.7 The process controller of a temperature variable has a 70% proportional
band. The controller output range is 5 to 25 mA and its input range is
from 1 to 45 °C. Calculate the controller gain.
3.8 What will be the residual steady-state error (offset) signal as a percentage
of a proportional only controller required to maintain the controller’s
output response at 35% when the bias value is set at 0%? The measured
controller gain is 5.
3.9 Consider a control system for the level of process liquid in a storage tank.
A proportional only controller is utilised. The liquid in the tank is at the
desired level when the flow of the liquid into the tank is 5 m3/h and the
flow out of the tank is 5 m3/h. The controller output is then at 50% and
its operating gain is 10%.h/m3. Calculate the controller output and the
system steady-state error if the flow of the liquid out of the tank is changed
to 7 m3/h.
3.10 The gain of an integral only controller has a value KI of 0.3 s-1. The bias
value of the controller is set at 0. If there is a sudden change to a constant
error of 15%, what will the output of the controller be after: (a) 5 seconds
and (b) 10 seconds?
3.11 An integral only controller is used to maintain a process speed to 16 rpm
within a range of 15 to 20 rpm. The controller output is 25% initially. The
controller constant is found to be -0.2% (% error). If the speed suddenly
changes to 14 rpm, what will the output of the controller be after 5 seconds
if the system error is constant?
3.12 The ultimate time period tu and the ultimate control loop gain Ku for the
tuning process of a PID controller were found to be equal to 8 mins and
14 s respectively.

Calculate the parameters of P only, PI and PID controllers using the

Ziegler-Nichols method.
In the previous chapters, we spoke about the components of a control system
of any process and the instrumentation commonly utilised to perform the
process monitoring task. By now you know quite a lot about all the elements
of a standard control system. However, this does not imply that you are able
to tune and operate a process control loop installation. In this chapter, you
will learn how process control elements and control algorithms are combined
to create an effective process control system [1].
In practice, it is often necessary to monitor several process variables at the
same time, and each variable can have a significant impact on the entire
process. Consequently, control systems must be designed to counteract any
disturbances at any point in the system and to remediate the effect of those
disturbances throughout the system. This chapter focuses on the various ways
of implementing a process control loop system and tuning the controller
parameters in order to obtain optimum control performance. However, the
main objective of the chapter is to familiarise you with carefully selected
general concepts and as much background knowledge as possible relating
to the interpretation and utilisation of the conventional process control loop
theories [1].
Normally, complete understanding of typical process control optimisation
requires a detailed mathematical study of control loop stability theory, detailed
knowledge of the process, and many years of experience in the field. The
emphasis is not on the theoretical development of process control loop
characteristics, but on equipping the process control technologist (you) with
substantial foundation knowledge concerning the real-life implementation of
control loops for effective process control. After completing this chapter, you
should be able to [1]:
• Explain the difference between a single variable control loop and a multi-
variable control loop
• Describe how a feedback control loop is different from a feedforward
control loop and explain their intrinsic properties in terms of:
– Operation
– Design
–
– Limitations
• Do the following for each type of standard process control loop (temperature,
pressure, level and flow):

– State the typical control system used and explain why it is used
– Identify and explain considerations for equipment selection (e.g., speed,
noise)
– Identify typical equipment requirements
– Draw the schematic diagram of the process control loop using ISA
symbology

• Explain the fundamentals of the control implementation process and briefly

describe the equipment considerations and requirements for each of the
following types of control systems:

– Cascade
– Batch
– Ratio
– Selective
– Fuzzy

• Explain the advantages and disadvantages of each type of control system

listed above
• Identify examples of process applications in which each type of control
system described in this chapter might be suitably used
• Describe the elements of the final control operation
• Describe the various type of control valves
• Understand the steps involved in the procedure of valve sizing and selection

Typical process control loops can be divided into two main categories: single
variable loops and multiple variable loops [1; 2].

The single variable control loop is the most elementary process-control loop.
The operation of the control loop is based on monitoring and maintaining
control of a specific process variable at the desired set point by manipulating a
controlling variable, irrespective of the other process parameters. This family of
process-control loops can also be divided into various subcategories, depending
on the type of process single variable [1; 2].

(a) Independent single variable – This involves process control systems

where certain specifications are fundamentally unrelated to the other process
variables. The desired set point is usually implemented into the system in order
to activate the action of the controller, and the control system is allowed to run.
As an example, consider a control system established to regulate the flow of a
process fluid into a tank where it is heated at a fixed rate implemented by the
system’s set point (see figure 4.1). Regardless of any other changes happening
within the overall process, this control system would adjust the final control
element of the loop (valve in this case) to maintain flow rate at the set point
value only if a load change affecting the consistency of the fluid flow into the
tank is introduced into the system [2].

(b) Interactive single variable – Considering the example illustrated in figure

4.1, a second single variable control loop can be implemented to monitor
and maintain the temperature of the process fluid in the tank at the desired
value (set point) by regulating the heat input into the tank (see figure 4.2). At
ideal operating conditions, the flow of the process fluid into the tank and the
temperature in the tank should both remain at their respective set point values
at any time. However, a change in the flow of process fluid into the tank would
represent a load change in the temperature control system [2].

This is attributed to the fact that with the change in the amount of process fluid
in the tank or the flow rate of the process fluid into the tank, the temperature
system should now adjust to the new conditions by readjusting the rate of
heat input to accommodate the load change and return the temperature to the
desired set point value. For this reason, the two loops are said to be interactive
with each other. This simply means that any sort of instability in the flow-
control loop would induce instability in the temperature system because of
this interaction between the two loops. In practice, almost all processes where
numerous variables are under control present control loops that interact [2].
(c) Compound variable – This refers to the situation in which the single
process control loop is used to monitor and control the relationship between
two or more process variables. In the case where measurements from two
sensors represent the input to the controller of a process control system, a
signal conditioning system can be used to scale the measurements from the
two sensors and add them before their input into the process controller for
evaluation and generation of the corrective action. These systems can become
quite complex to analyse. A good illustration of this type of control system is
in a section of a manufacturing plant where a chemical reaction occurs and
it is necessary to monitor the ratio of two reactants [1; 2].

The procedure usually consists in letting one reactant flow into the reactor
without any adjustment; however, the flow rate should be measured at all
time. This allows the adjustment of the flow rate of the second reactant after
its measurement in order to constantly obtain the desired ratio of reactants (set
point) (see figure 4.3). The flow rate of the reactant flowing uncontrollably is
measured and added, with appropriate scaling, to the measurement of the flow
rate of the controlled reactant and the resulting signal is sent to the controller,
which reacts by adjusting the control valve of the controlled reactant input
line [1; 2].
(d) Examples of single control loops

(i) Level control loops

The control of the level of a process fluid in a vessel is influenced by several

factors. One of these is the size or the shape of the vessel, because a smaller
vessel will fill faster than a larger one. Consequently, manufacturers may
use various measurement instruments to determine the level of the process
fluid in the vessel. These include ultrasonic, float gauge, radar and pressure
measurement. One of the most critical phenomena to avoid with level control
is the overflow of the process tank or vessel. For this reason, redundant level
control systems are usually employed. A valve on the input and/or output
connection pipes is the most common final control element used in typical
level-control loops (see figure 4.4) [2].
(ii) Flow control loops

Generally, a flow-feedback control loop consisting of a flow sensor, a transmitter,

a controller, and a valve or pump is used in flow control (see figure 4.5). The
speed of changes in a flow-control loop is generally considered the fastest
type of control response. This makes it critical for flow-control instruments to
have fast sampling and response time periods. Typical flow transmitters are
very sensitive instruments, and so they generate noise or fluctuations rapidly
in the control signal [2].

Modern flow transmitters are equipped with filters with a damping function
that eliminates signal noises. In some cases, the filters are installed between
the transmitter and the control system. The temperature of the process fluid
has a direct influence on its density, and so the temperature of the process
fluid is also measured when flow measurements are taken and fluctuations in
temperature readings are compensated for during flow calculations [2].

(iii) Temperature-control loop

Changing the temperature of a process fluid takes a fairly long time. Consequently,
temperature-feedback control loops are regarded as quite slow. To increase
the speed of the temperature-control loop response, feedforward control
strategies are usually preferred. Although temperature could be directly wired
to the input interface of the system controller in some cases, the control system
usually involves temperature transmitters and controllers (see figure 4.6).
Thermocouples or resistance temperature detectors (RTDs) are the temperature
sensors that are typically used. The final control element for a temperature-
control loop is generally the valve of the fuel line to a burner or a valve to a
specific type of heat exchanger. A cool process fluid can also be added to the
process to maintain temperature (cooling operation) [2].
(iv) Pressure-control loops

The speed of the pressure-control loop response to changes in load varies;

these loops can respond or produce control action either slowly or quickly,
depending on process parameters such as the volume of the process fluid.
Large-volume systems such as huge storage facilities for natural gas tend to
change less quickly than low-volume systems. A typical pressure-feedback
control loop is illustrated in figure 4.7 [2].
Multivariable control loops are referred to as advanced process control loops
because the operation of this kind of control system is quite complex. These
types of control loops consist of a primary controller that monitors one process
variable by conveying signals to a secondary controller of a different control
loop that influences the process variable of the primary control loop. Up to
now in this study guide we have only considered examples of control systems
with a single manipulated variable and a single process-controlled variable.
However, in real processes far more complicated structures are involved
consisting of multiple disturbances, manipulations and controlled outputs. For
instance, consider a process unit where the temperature of a process fluid is
to be kept constant in a tank that is heated by a pressurised steam chamber
surrounding the tank (steam jacket). The temperature of the process fluid in
the tank can be regarded as the primary process-controlled variable, which is
controlled by a primary controller sending signals to a secondary controller that
has the purpose of monitoring the pressure of the steam (see figure 4.8) [2; 3].

The primary controller basically manipulates the set point value of the secondary
controller in order to maintain the primary process controlled variable, which
in this case is the temperature of the process fluid in the tank at the desired set
point value. The multi-variable control system described previously assumes the
controlled variables to be predetermined, which is not in fact the case. In real
plants, the identification of the process variables to be controlled, especially
when we are dealing with multi-variable control systems, is one of the first
tasks of the control system designer. In chemical engineering the degree of
freedom in a system is defined as all the possible ways the system can be
tuned without violating any design constraint. Consequently, the number of
variables that could be controlled is obviously limited by the existing degree
of freedom (DOF) in the system under control. The fundamental definition of
DOF encountered when learning the basics and fundamentals of chemical
engineering is represented mathematically by the following equation [2; 3]:

DOFC = DOF – Number of disurbances – Number of equations relating

them

In process control the disturbances to the process, which are external factors
that cannot be controlled, represent additional limitations to the DOF of the
system. In this case, the DOF can be expressed mathematically as follows [2; 3]:

DOFC = DOF – Number of disurbances included as independent variables

In multi-variable control loop design the primary objective is to reduce the

number of control degrees of freedom to zero. So, if an additional control loop
is added to the multi-variable control system, the DOF of the control system is
reduced by one by fixing one means of movement of the system. It can then
be reasonably concluded that the number of control loops in a multi-variable
control system should be equal to the DOF of the control system. When the
control loops are fewer than DOFC, the system will not be controlled effectively
and the control system will not generate consistent results. However, when
the number of control loops is higher than the DOF, the system is known to
be overdetermined. This simply means that some of the control loops are
unnecessary, and will not operate [3].

Because determining the number of degrees of freedom of control systems

of real industrial processes is an extremely tedious task, control engineers
generally do not bother to do it. As an alternative solution, they draw on their
work experience to estimate the number of control loops required for each
process. This method is simple and it is not dangerous, because less experienced
control engineers tend to use a rule of thumb that states that the maximum
number of variables that can be controlled in a system is equal to the number
of variables of all manipulable variables entering or leaving the overall process
to be controlled. The worst thing that could happen here is that the control
engineer could implement more control loops than are necessary [3].

The tuning process for multiple variable loops is quite different from the standard
tuning procedure for a single control loop. However, it consists in tuning the
secondary loop before tuning the primary loop, because alterations made to
the secondary loop have an immediate effect on the primary loop, whereas
adjustments made to the primary loop have no influence on the secondary loop
tuning. The primary process control loop is referred to as the master control
loop, and the secondary process control loop is known as the slave loop. For
this reason, you need to apply a very efficient technique when combining the
multiple disturbances, manipulation and controlled outputs in a multi-variable
control loop. Unfortunately, no universal method to develop multi-variable
control systems with outstanding performance has been reported so far.
The most commonly used technique involves eliminating the worst control
loop combinations using the following rules of thumb [3]:

• Select control loops where the cause-and-effect relationship between the

disturbance and manipulation is very strong.
• Preference should be given to control loops that minimise the dead time
between the disturbance and the control manipulation.
• Select control loop combinations that minimise the interaction between
the control loops.
• Preference should be given to control loops with reasonable process gain,
as this reduces the control manipulation range required.
• Although this is almost impossible in practice, control loops with an almost
linear relationship between the disturbance and the control manipulation
are preferable.

The use of a complete multi-variable controller is another effective option

when controlling multiple-disturbance, manipulation and output systems.
These controllers have been used in the oil industry for a number of years.
They provide control to the system as a whole without breaking down the
control complex into lots of single loops. These controllers are receiving
increasing attention in other applications [3].
A feedback control loop is the most popular standard single variable control
loop, where the process device, the measuring instrument, the controller and
final control element are all interconnected in a loop. The feedback loop
measures the actual value of the process-controlled variable, and the reading
is sent to the controller, where it is compared against the system’s desired set
point. When the process variable differs from the desired set point, a system
error is generated and control action is initiated to return the process-controlled
variable to set point. A typical feedback control loop is represented in figure
4.9. In this case study, the temperature of the fluid is measured by a transmitter
and compared with the set point value. Afterwards, a control action is taken
by either opening or closing a hot steam valve to regulate the temperature of
the fluid [3; 4].
Feedback loops have the advantage of directly controlling the desired process
variable. However, the fact that an error should be generated in the first place
for the control action to be produced represents a disadvantage of this control
technique. Nevertheless, feedback control loops are commonly used in the
process control industry. The most commonly encountered are the pressure,
flow, level, and temperature feedback control loops [3; 4].

Feedforward control loops operate under the principle of predicting the

process load disturbances and eliminating them before they are able to affect
the process-controlled variable. It is essential for the control engineer to have
an excellent mathematical understanding of how the manipulated variables
could influence the process variable in order for the control system to be
fairly effective. A typical feedforward control system intended to control the
temperature of the process fluid in a tank by either opening or closing the
valve of a hot steam line, depending on how much cold fluid passes through
the flow sensor of the control system, is illustrated in figure 4.10 [3; 4].
What makes the feedforward control system an attractive control technique is
the fact that the system error is prevented rather than corrected, although the
technique is not effective enough to account for all possible load disturbances in
the system. A number of external factors could easily represent a disturbance to
the process and generally cannot be effectively accounted for in the feedforward
control system design. These include build-up in pipes, temperature, humidity,
moisture content in raw materials and inconsistency in grade of raw materials.
The effectiveness of this type of control system is clearly perceptible in cases
where the controlled variable is a potential major load disturbance to the
process variable ultimately being controlled. In light of the complexity and
expense associated with the use of the feedforward control technique, the
benefits resulting from this control system may not be worthwhile in the case
of a variable that could probably represent just a minor load disturbance [3; 4].

Accounting effectively for every possible load disturbance in a process

is a very complex task. Consequently, feedforward control systems are
often combined with feedback control systems. This combined control system
uses controllers with summing functions in order to sum up the inputs from
both the feedforward loop and the feedback loop, and send just a single
signal to the final control element. The combination of a feedback and a
feedforward control loop is illustrated in figure 4.11, where a flow transmitter
and a temperature transmitter generate the information required to adjust the
valve of a hot steam line [3; 4].
Up until now we have discussed particular types of control loops, their
characteristic components and typical applications (e.g. pressure, flow and
temperature). However, in industry, combinations of many independent and
interconnected loops are used to monitor and control the outputs of a typical
processing plant. Next I will say a bit more about the control techniques
most commonly used in various process industries at present [5].

We have already discussed the fact that in a single control loop, an operator
inputs a set point into the controller and, based on the set point, a response
is generated by the controller in order to drive the final control element
appropriately. Meanwhile, a cascade control system involves two or more
controllers. In a cascade loop, the output response generated by one controller
represents the set point of another controller. The controller generating the
set point is referred to as the primary, master or outer controller, while the
controller obtaining the driven set point is referred to as the secondary, slave
or inner controller. Let’s consider the classic example of a level-control system.
In a single loop control, a level transmitter sends a signal to the controller
that compares it with its set point before generating an appropriate signal that
will drive the final control element (a control valve) to keep the level at its set
point (refer to figure 4.4) [5].

However, the cascade control arrangement for the same process consists
of a primary controller (level controller) that generates an output response
representing the set point of a slave controller (flow controller), which will then
adjust the control valve so that the flow corresponds to the set point driven by
the level controller in order to keep the level of the process fluid at its design
set point (see figure 4.12) [5].

(a) Advantages and disadvantages of cascade control

The main advantage of the cascade control arrangement is its ability to isolate
the response of the slow control loop from non-linearities in the final control
element. In the level control scenario above, the faster flow control loop is
utilised to prevent or/and minimise the control valve problems that could arise as
a result of the relatively slow level control loop. For instance, the control valve
might have a static friction problem. This type of valve problem arises when
the internals of the valve are sticky, and it usually causes the valve to stick in
one position. This can lead to the level-control loop continuously oscillating
in a stick–slip cycle with slow periods, which is very likely to have a negative
effect on the process downstream. The problem of the sticky control valve
will also cause the fast flow control loop to oscillate. However, because of
the inherently fast dynamic behaviour of a well-tuned flow control loop, the
oscillation will be much faster and could be attenuated by the downstream
process without generating much of a negative effect on it [5].

Another valve crisis that can occur relates to non-linear flow characteristics of
the valve. This makes the control loop driving the valve sluggish or unstable,
and so this control loop must be detuned as a way to maintain stability across
the available range of flow rates. If the faulty valve was driven by the level
controller, the detuning process required to maintain stability would probably
lead to very poor level control. Fortunately, in a cascade control loop the
valve is generally driven by the flow controller, which in this case will be
detuned to maintain stability. Consequently, very poor flow control will be
obtained. However, because the flow-control loop is dynamically much faster
than the level-control loop, the level-control loop will barely be affected
negatively. Because of these characteristics of the cascade control system,
it is recommended for use where processes that are dynamically slow, such
as composition, humidity, level and temperature, have to be controlled through
the manipulation of dynamically fast processes such as liquid and gas-flow [5].

Three main drawbacks are generally associated with the use of the cascade
control arrangement. The first disadvantage involves the need for additional
measurement such as flow rate in order for the control strategy to work. The
second limitation is the need to tune an additional controller, and the final
drawback relates to the fact that the engineer and operators have to deal with
a more complex control technique. So, before deciding whether or not this
control system should be implemented, engineers need to assess these various
disadvantages against the predicted improvement that the control system
will provide. For example, the cascade control arrangement is beneficial
only when the inner loop is dynamically much faster than the outer loop. If
the dynamics of the inner loop are not at least three time faster than those
of the outer loop, the improved performance obtained will not be worth the
added complexity of the system. Furthermore, when the inner loop in the
cascade control arrangement is not that much faster than the outer loop, there
is a significant risk of interaction between the two loops, and this could result
in major instability in the system. The situation is worsened when the inner
loop is tuned quite violently [5].

(b) Tuning procedure for cascade control systems

When tuning a cascade arrangement, it is best to tune the innermost loop first.
After this tuning process is complete, the loop should be placed in external
set point mode, also known as cascade control mode. The next step involves
tuning the loop driving the set point of the previous loop. When tuning control
loops in a cascade arrangement, use of the quarter-amplitude-damping tuning
techniques such as the unmodified Ziegler-Nichols or the Cohen-Coon tuning
techniques is not recommended. This is because of the risk of instability if the
process dynamics of the inner and outer loops are similar [5].

Batch control systems are control structures applicable to processes that

are conducted in batches throughout (see figure 4.13). The mixing of the
ingredients required for the production of a soft drink, for example, is generally
performed as a batch operation. Manufacturing processes of this kind can
only be effectively accomplished in batches because determined quantities of
ingredients are to be mixed for a specific period of time, and so continuously
running processes are not suitable in this case. Flow, level, temperature, mass
and pressure are key parameters that need to be constantly monitored at various
stages of the process, because batch processes typically consist of introducing
the correct proportions of ingredients into a batch reactor in order to obtain a
final product of the desired specifications after residence time [4; 5].
One of the main features of a batch process that represents a drawback for
batch control systems is the fact that at the beginning of each batch process
start-up is required. These restarting procedures may present control problems,
because during the start-up period all process measurements are below the
system design set point(s). The fact that in order to improve final product
specifications it could be necessary to change the nature or proportions of
raw materials and therefore also to recalibrate the measurement and control
instruments constitutes another disadvantage of this kind of control system [4; 5].

Ratio control is an elementary type of feedforward control, widely applied

in the process industry almost exclusively to flow rate control. The primary
objective is to maintain the ratio of two process variables to a specified value
(set point). The process of mixing two or more fluids in defined proportion
in order to control the composition of the product stream is a fairly common
application of ratio control. For instance, ratio control can be used to control
batch processed quantities of food prepared according to a very strict recipe.
The fuel-to-air ratio in a combustion engine can also be monitored by means
of ratio control. Finally, a chemical reaction requiring a specific stoichiometric
balance can also be an application of a ratio control system. The two process
variables are usually process flow rates, one flow rate as the manipulated
variable u and the other flow rate as the disturbance variable d; hence the
ratio [5; 6]:

This type of control system involves a controller that receives measurement input
from a flow reading instrument on an unregulated flow to a specific process
unit. The controller performs the relevant calculations in terms of the ratio
of the process raw materials. Afterwards an output signal from the controller
carrying the desired set point is sent to another controller that adjusts the flow
of the second fluid into the process unit so that the desired proportion of the
second fluid stream is added to the unit operation [5; 6].

For instance, consider a process during which a monomer is to be dispersed in

an organic liquid in the proportion of one part monomer to two parts organic
liquid in a mixing vessel. The typical process set up would be to have the
monomer supply line on one side of the mixing unit and the organic liquid
supply line on the other side. It would therefore be practical to install a control
system to maintain the ratio of monomer to organic liquid at the desired value
(set point). In a typical process, ratio control can be implemented in two main
ways, both of which require constant monitoring of the two process variables.
Ratio control systems are also suitable for continuous processes in which an
additive is being introduced into the process unit. An example of this is the
chlorination of water in a typical potable water production plant [5; 6].

(a) Flow fraction controller – This technique is a straightforward and simple

application of the ratio control algorithm. A divider instrument, generally
included in the controller, instantaneously calculates the ratio Rm of the two
process flow rates by dividing the measured flow rate d by the measured flow
rate u. Afterwards, the controller compares the calculated ratio with the set
point Rsp and generates a corresponding controller output in order to adjust
the opening of the control valve appropriately to maintain the design ratio R
constant. In this case, the controlled variable is actually the ratio Rm of the
two other process variables [6].

(b) Ratio relay controller – This ratio control method is basically similar to
the flow fraction controller technique; however, the control algorithm entails
measuring one flow rate, for instance d and using the measurement signal
obtained to compute what the flow rate of the other process variable (d in
this case) should be to maintain the design ratio R constant. The output of the
ratio relay, which is equal to usp = d/R, is then utilised as the set point for the
controller. Once the controller receives the measurement signal of the flow
rate of the other process variable u, it compares it with the given set point usp
= d/R and adjusts the generated output appropriately to keep the flow of the
manipulated stream at the desired set point, hence keeping the ratio R constant.
Sophisticated ratio control systems have become increasingly popular. The
control algorithm of these control designs can be based either on feedback
control of the load-stream flow rate or the analysis of the composition of the
mixed stream, because it could have a feedback effect on the constant ratio
R of the ratio relay [6].

(c) Advantages and disadvantages of ratio control

Because the control algorithm of the system calculates the actual ratio of process
variables, some of the advantageous key features of ratio control systems are [6]:

• Pre-setting of the relative quantities of components in blending operations

• Maintaining a stoichiometric ratio of reactants to a reactor constant, thus
reducing the effect of feed flow rate changes
• Maintaining a desired reflux ratio for a distillation column
• Maintaining the fuel–air ratio to a furnace at the optimum value

However, the major disadvantage associated with the use of this control is the
fact that a divider element must be included in the control loop, and this causes
the process gain to fluctuate in a non-linear way. Because of this significant
drawback it is preferable to implement the ratio relay control technique in
most cases [6].

The selective control technique is commonly used in applications where it is

extremely important to protect or maintain the safety of process equipment,
even if this means compromising the control of the desired value of the
optimal process variable. This type of control system basically maintains the
most important process variable at the design value. For example, in a boiler
the fuel flow should not outpace air flow because unreacted fuel can build
up in the boiler and cause an explosion. Consequently, a selective control is
required to always favour an air-rich mixture, but never a fuel-rich mixture [6].

This is a form of adaptive process control technology that is relatively new,

and uses computer logic to make decisions about how to adjust the process.
In determining whether something is or is not included in a set, this form
of control system is based on a rating scale in which numerous factors are
taken into consideration and graded by the computer algorithm. Because the
monitoring and adjustment of the process is performed by a machine, without
any human intervention, fuzzy control removes from the operators some of
the control abilities, but not the responsibility for controlling the process. The
basic principle underlying fuzzy control entails developing a form of artificial
intelligence that will consider multiple variables to formulate a theory on
how to make operational improvements to the process, to adjust the process
appropriately, and to learn from the result obtained after adjusting the process
[6; 7]; for example: changing the cooling water flow rate to control condenser
pressure (vacuum), or changing steam flow rate to control heat exchanger
outlet temperature. In both cases, flow-control loops should be used as inner
loops in cascade arrangements. Fuzzy control systems have the following
advantages [7]:
• Fuzzy controllers are more robust than PID controllers because they are
able to cover a much wider range of operating conditions than PIDs can,
and are able to operate with noise and disturbances of different natures.
• Developing a fuzzy controller is cheaper than developing a model-based
or other controller to do the same thing.
• Fuzzy controllers are customisable, since it is easier to understand and
modify their rules, which not only use a human operator’s strategy but also
are expressed in natural linguistic terms.
• It is easy to learn how fuzzy controllers operate and how to design and
apply them to a concrete application.

•
•
•
•
•

The final control operation converts the control signal produced by the controller
into proportional action on the process itself. For instance, in order to convert
a characteristic 4–20 mA control signal generated to rectify the large flow rate
of a system, some final adjustments would certainly be required. It would
be necessary to generate enough power to actuate changing the state of the
controlled-process state and to convert the control signal (4–20 mA). The
following diagram illustrates the various stages involved in the final control
operation and the sequencing of the operation [7; 8].

From the control signal to the process, each stage plays a fundamental role in
efficiently applying the corrective action generated by the controller in order
to eliminate the error detected in the process [8].

At this stage it is necessary to convert the corrective signal produced by the

controller so that it can be effectively interpreted at the next stage by the actuator
device. Depending on the type of actuator device, various types of conversion
can be made. For instance, a current signal can be converted into a pressure
signal, or modifications can be made from current to voltage. For example, if
an electric motor actuator is used to operate a final control element (valve),
the signal from the controller should be converted into a 4–20 mA signal in
order to operate the electric motor actuator. Transducers are the instruments
that are usually used to convert control signals from one form to another.
This process is sometimes referred to as analogue signal conditioning. The
most commonly used methods of signal conversions for final process control
action include the following [8]:

• Digital to analogue conversion

• Current to voltage conversion
• Amplification
• Current to pneumatic pressure conversion

The amplified or converted signal generated by the signal conversion device

has the main purpose of actuating or appropriately operating the mechanism
required to adjust the controlling variable in the process accordingly. The
actuator is a device that has the ability to amplify and convert the control
signal produced into corrective action relating to the process control element.
In practice the direct effect is usually obtained by a device or instrument in
the process, such as a heater or valve, which must be adjusted by a precise
mechanism. For instance, if a valve is to be operated in order to eliminate an
existing error in a controlled system, the actuator is the device that converts the
corrective signal generated into the physical action of closing or opening the
valve. The most common actuation devices used in chemical plants include [8]:

• Pneumatic actuators
• Electrical actuators such as stepper motors, relays and solenoids
• Hydraulic actuators

The control element (usually a valve) is a designed integral part of the process
itself and it has a direct effect on the process variable. For instance, if the flow
of a process stream is to be monitored, a suitable control element would be a
valve to be installed directly within the flow system of the process plant. In the
case where the temperature of a certain process is the variable to be controlled,
it is necessary to implement an appropriate mechanism in the process so that
the control element directly regulating the temperature will be involved in the
control system algorithm. The control element could be a heater or a cooler,
or a combination of both a heater and cooler that is electrically operated by
pneumatic valves or relays [8].

A signal generated from the controller is conveyed to an appropriate transducer,

also known as an actuator, and is converted to a stream of pressurised air.
This pressurised air stream pushes against one side of the valve’s diaphragm
and works against the spring to move the valve stem to the desired position.
Within the body, the valve stem is moved up; this movement is known as a
stroke, and is initiated by the actuator of the valve. The strokes of the valve
stem open and close a gap between the valve plug and the valve seat. The
valve trim is the section within the valve body that comes in contact with the
process fluid, including the valve seat and plug [8].

Any type of valve can be manufactured with different trim designs, and each
design is meant to regulate and “shape” the flow according to its specific
intrinsic characteristics. The mechanical and geometry characteristics of these
various flow control modes are discussed later in the chapter. However the
three most common ways in which the valve opens, in other words, the three
most common ways in which it strokes or travels according to the flow of the
process stream to be controlled, are summarised in table 4.1 below [8].

Now that we have briefly discussed the most common types of valve control,
we need to talk about the various types of valves commonly utilised in the
field of process control. There is an extremely wide variety of categories of
valve available for implementation in manufacturing systems. However, the
most commonly used ones in manufacturing processes are those summarised
in table 4.2 below [8].
Some more information about these valves and their various applications is
summarised in table 4.3 below [8].

In most chemical processing plants, the globe valve is the type of valve
most commonly used to regulate the flow of process fluids in pipelines. It
generally consists of a moveable disk-shaped element and a stationary ring seat
in a spherical body. Like ball valves, where the plug is moved in and out of
the globe, globe valves can also have a stem or a cage (see figure 4.14) [7; 8].
The plug is generally designed in such a way that the flow of process fluids can
be properly controlled. Potential leakages through the valve can be prevented
by means of a seal. These valves have design features such as a cover that can
be easily removed, exposing the plug and seal and making it easy to maintain
the equipment. These valves are very efficient for ON/OFF flow control and
when accurate throttling is required. Moreover, the use of these types of valves
is particularly suitable where cavitation and noise are factors to consider. A
typical example of the use of a globe valve is to control the cold and hot water
flows to a kitchen sink or bathroom basin [7; 8].

A fail-safe control system is designed in such a way that if a malfunction

involving instrumentation failure resulting in an interruption to the supply
of either electricity or air forces the plant to shut down, this will take place
safely. The purpose of this type of control design is to ensure the safety of
plant employees and of products, equipment and the environment at all time
during plant operation [7; 8].

In most process control loops, an increase in air pressure or electrical signal

generated from the controller usually causes the final control element (valve)
to open. Conversely, the conceptual algorithm of the fail-safe design conditions
the process-control loops so that the most probable failure in terms of loss of
instrument air pressure or loss of electrical power will result in a less dangerous
outcome [7; 8].

There are two categories of fail-safe control valves used to ensure fail-safe
operation. These are air to open/fail closed and air to close/fail open control
valves. The design algorithm of these valves is fundamentally the reverse of
the process-control loop. Consequently, the type of valve used has a direct
effect on the required output action of the process controller. Moving from
one class of fail-safe control valves to the other basically entails changing
the output manipulation of the process controller from one direction to the
opposite direction. In the event of a plant instrument air failure it is essential
for all control valves to fail in a safe position. For instance, it is important that
the feed valves or possibly just one of the feed valves of an exothermic reactor
should fail closed (air to open) and its coolant system valves should fail open
(air to close) in the event of a failure in the system [7; 8].

The air to close/fail open valve, denoted AC/FO, is a control valve that is
designed to switch to the open position when a control or air failure signal
is generated. This type of control valve is generally denoted as FO in most
process flow diagrams. The valve is basically held by a spring, and needs air
pressure to move it towards the closed position. The valve is designed such that
the valve is fully open when there is no signal produced by the controller or
when there is no air pressure. However, as the controller generates a signal or
the air pressure gradually increases, the valve closes progressively. The output
of the controller needs to increase in order for the valve to close further [7; 8].
Air to open/fail closed valves are normally held closed by a spring and require
air pressure or the output signal from a controller to open them. They are
denoted FC on most process flow diagrams. They are designed such that they
open gradually as the air pressure increases or the output signal generated by
the controller increases [7; 8].

The other important thing you need to understand is the direction of control
action. Generally, the output response from the process controller is directly
dependent on the action of the process under control. Most of the time the
primary purpose of the output signal generated by the controller is to counteract
the current situation detected in the system under control. For example, when
the measured flow of a process increases beyond the design set point, the
output response generated by the controller has the objective of reducing the
flow of the process fluid. The action of the process controller can be defined
in two different ways [7; 8]:

(a) Direct acting – An increase in the measured variable results in an increase

in the controller output generated. In such cases the air pressure inlet is applied
to the diaphragm of the valve, forcing the stem to move. This movement of
the stem causes the plug to close the gap or opening, enabling fluid to pass
through the valve. The opening or gap is designed to reduce as the pressure
input increases. If the inlet pressure drops to zero as a result of some failure
occurring in the system, the spring forces the valve to open and causes it to
fail in a wide-open position [7; 8].

(b) Reverse acting – In this case the output response from the process controller
decreases as the results of the process measured variable increase. In a reverse
acting situation, the operating force of the control valve is derived from
the compressed air pressure which is exerted on the flexible diaphragm of
the control valve. The actuator of the control valve is designed in such a way
that the force generated from the compressed air pressure, multiplied by the
area of the diaphragm, overcomes the force applied in the opposite direction
by the springs of the valve [7; 8].

The various control (globe) valves are categorised based on the link between
the position of the stem of the valve and the flow rate obtained through the
control valve. Consequently, there are three main types of control (globe)
valve [7; 8]:

(a) Linear valve – As indicated by its name, the flow rate generated by this
kind of globe valve varies linearly on the basis of the position of the stem. In
this case everything is perfectly set and designed for the pressure drop to be
solely dependent on the valve. For this reason this type of valve is commonly
utilised when it is necessary to maintain a constant pressure across the valve
and its travel is directly proportional to its stroke action. The relationship
between the stem position of the valve and the flow rate generated can be
mathematically expressed by the following equation [7; 8]:
where Q is the flow rate generated by the valve (m3/s); Qmax is the maximum
flow rate that can be generated by the valve (m3/s); S is the stem position (m)
and Smax is the maximum stem position (m).

In the case where a relatively constant pressure drop is maintained across

the valve, the flow rate varies linearly with the stem position according to the
following relationship:

where P is the pressure in pressure units (Pa, N/m2) and

Qmin is the minimum flow rate.

(b) Quick opening valve – This family of valves is most suitable for full ON/
full OFF industrial control applications, because in quick opening valves, high
flow rates are achieved with only a small change in valve stroke. They can,
for instance, generate 90% of the maximum flow rate with just 30% travel of
the stem [7; 8]. A quick opening valve is very useful for safety where a quick
opening action is required. It is also known as the square root trim, according
to the following equation below:

(c) Equal percentage valve – This type of valve plays a very important role in
flow control application due to its characteristic of inducing a change in the
outflow rate which is equivalent in percentage to the change in stem position of
the valve. This simply means that equal increments of the stem travel generate
an equal percentage in flow change. These valves are designed in such a way
that they do not completely shut off the flow of fluid when the stem is positioned
at one of its travelling limits. Hence, Qmin denotes the minimum flow rate
produced when the stem is at one of its travelling limits. At its other travelling
limit, the valve generates a flow Qmax, which represents the maximum open
valve flow rate. For this type of valve, the characteristic rangeability, denoted
as R, is the ratio of the maximum flow rate and the minimum flow rate. It is
expressed by the following equation [7; 8]:

The flow relationship is represented mathematically by the expression below

[7; 8]:
Example

An equal percentage valve with a full travelling path of 50 mm has a minimum

flow rate of 3 m3/s and a maximum flow rate of 75 m3/s. Calculate the flow
rate generated by a stem travel of 2 cm.

Solution

The rangeability of the valve can be calculated from the information provided
in the problem statement.

The generated flow can then be calculated as follows:

The task of determining the correct valve size for a specific control system is
essential, and requires the consideration of various operating factors. However,
the capacity parameter of the valve or the flow coefficient, denoted CV, is the
most important characteristic to take into consideration when selecting the
valve for a control system. This characteristic is used to evaluate how efficient
a valve is at allowing a fluid to flow through it [8; 9].
The flow coefficient of the valve CV is the number of US gallons of water per
minute that flows through a fully open valve with a pressure differential of 1
pascal per square inch. It is generally determined experimentally. However,
most valve manufacturers provide customers with the CV chart of the valve
they manufacture. Another technique used to estimate the values of the flow
coefficient when sizing an appropriate valve needed for a specific system is
through the valve flow equation [8; 9].

The task of valve sizing relies on the following four fundamental characteristics
[8; 9]:

• Flow rate through the valve

• Stem position
• Fluid properties
• Pressure drop through the valve

The volumetric flow Q through the control valve can be calculated as follows
[8; 9]:

In the case of existing design specifications such as maximum flow rate through
the valve and maximum pressure drop allowable, the following form of the
equation is utilised to calculate the flow rate across the valve [8; 9]:

PV is
the pressure drop across the valve measured in pascals per square inch; CV is
the valve coefficient; SG is the specific gravity compared with water and f(x) is
the fraction of the total area of the valve. The table below represents a typical
CV chart usually provided by the valve manufacturer [8; 9].
Before we follow the standard algorithm for control valve (globe) sizing and
selection, we need to identify the type of control valve (globe). The following
key variables need to be carefully considered [8; 9]:

• The design maximum flow rate

• The design minimum flow rate
• The total pressure drop
• The operating flow rate
• The pipe diameter
• The specific gravity of process fluid

Some other considerations for valve sizing and selection are [8; 9]:

• It is not advisable to use a control valve that is less than half the pipe size.
• It is not advisable to use the lower 10% of the valve stroke and the upper
20% of the valve stroke. The globe valve is much easier to control within
the 10–80% stroke range. One of the common rules of thumb is that the
selected control valve should handle 10–15% of the total pressure drop or
10 psi, whichever is greater.

The following steps are followed when sizing and selecting a control valve
(globe) [8; 9]:
Step 1: Calculate CV at the design maximum flow rate using the flow equation
for a control valve.

Step 2: Perform preliminary valve selection using the chart provided and
taking into consideration all key sizing and selection criteria. At this
stage it is necessary to determine which valve is suitable to provide
the design maximum flow rate and ensure that the valve position is
within 80 to 85% travel.

Step 3: Calculate CV at the design minimum flow rate using the flow equation
for a control valve.

Step 4: Verify that the stroke percentage for minimum flow does not fall
below 10% at the design minimum flow rate. If it does, it is advisable
to choose a smaller valve, especially if the valve is likely to operate
near the minimum flow rate for long periods of time.

Step 5: Check the gain across applicable flow rates and remember that the
gain should never be less than 0.5.

In a section of a chemical plant, water at 25 °C is pumped from one tank to

another through a piping system with a total pressure drop of 160 psi. The
design maximum flow rate is 160 gpm, the operating flow rate is 120 gpm
and the minimum flow rate is 30 gpm. The pipe diameter is 5 inches. (Take
specific gravity of water at 25 °C to be equal to 1.)

Solution

1 – Calculating the CV at the maximum flow rate using the flow equation for
a control valve.

Flow equation:

From the flow equation we can make CV the subject of the formula and obtain:
Before we begin the calculations, we need to determine the pressure drop
allowable across the valve. For this case study, 10% of the total pressure drop
is equal to 16 psi, which is higher than 10 psi. Therefore, the valve selected
should be able to handle a pressure of 16 psi.

Calculating the valve coefficient at the maximum design flow:

2 – Preliminary valve selection

Looking at the values from the chart provided (table 4.2), we can see that a
two-inch valve appears to be quite suitable for the calculated value of CV (40)
at maximum design flow rate. A valve size of 3/2 inches will not be suitable
for this maximum flow rate because at a stroke percentage of 100% the value
of CV is only 35.8.

The valve of size 2 inches has a CV value of 25.4 at a stroke percentage of 70%
and a CV value of 59.7 at a stroke percentage of 100%. Using the interpolation
technique we approximate from the given parameters that the calculated CV
value of 40 corresponds to a stroke percentage of 83%, which falls within
the recommended range of 80–85% for maximum design flow rate. For this
reason, at this stage we select a valve size of 2 inches.

3 – Calculating the valve coefficient at the minimum design flow

4 – Verifying that the stroke percentage for minimum flow does not fall below
10% at the design minimum flow rate

Looking at the value of CV calculated while referring to the valve chart (for size
2 inches), we observe that it corresponds to a stroke percentage of between
30 and 40%. Using the interpolation technique, we see that the calculated
value of CV obtained at minimum flow (7) corresponds approximately with the
valve size of 2 inches preselected to a stroke percentage of 35%. This is above
the lowest limit of 10% recommended. Consequently, this size is certainly still
acceptable for the design minimum flow rate.
It is usually almost impossible to determine the perfect valve size. However, the
selection criteria mentioned above make it possible to determine a valve that
will operate well most of the time. For this reason, we need good judgement
skills. For instance, during valve sizing and selection, it is essential to understand
whether the system is more likely to operate closer to the minimum design
flow rate than to the maximum flow rate, or whether the system is designed
to operate near the maximum design flow rate for extended periods of time.
A maximum pressure drop of 16 psi has been used in all our calculations.
However, at flow rates lower than the maximum allowable flow rate, the
pressure drop across the valve will actually be lower, and so by using the
maximum pressure drop, we are considering the worst case scenario. For
the preselected valve size (2 inches), even a CV value of approximately 1.5
at minimum design flow rate would not really be an issue. The preselected
valve has a CV value of 1.66 at a stroke percentage of 10%. In our calculations
we used the maximum pressure drop, and looking at the flow equation, we
notice that at lower pressure drops the calculated values of CV increase, which
is advantageous, in that our estimation remains acceptable.
5 – Calculating the gain across applicable flow rates
(a) Calculating the valve coefficient at the operating design flow rate:

Using the chart values in table 4.2 and the interpolation technique, we find that
the value of CV calculated above corresponds to a travel percentage of 74%.
(b) Calculating gain across applicable flow
The gain of the system is mathematically defined as:

The three design flow rates for this specific system are as follows:
Qmin = 30 gpm
Qop = 120 gpm
Qmax = 160 gpm

The corresponding CV values are as follows:

CV(min) = 7.5
CV(op) = 30
CV(max) = 40
The corresponding stroke percentages are as follows:

Stroke at minimum flow rate = 35%

Stroke at operating flow rate = 74%
Stroke at maximum flow rate = 83%

From all the calculated parameters above, the following table can be constructed:

Gain 1 = 90/39 = 2.3

Gain 2 = 40/9 = 4.4

We can see that none of the gains calculated is less than 0.5, which follows
the rule mentioned previously.

It is also recommended that the difference between the values calculated

should be less than 50% of the higher value obtained.

Therefore: 4.4 – 2.3 = 2.1

0.5 × 4.4 = 2.2
2.1 < 2.2. Consequently, the chosen valve (2 inches) should operate well most
of the time, and there should be no problem in controlling the valve.

The recent advances in process control refer to the implementation of a broad

range of novel and more sophisticated technologies within traditional process
control systems to improve their efficiency. However, advanced process controls
are usually utilised when necessary, and usually in addition to existing basic
process controls systems. During the design and construction of most chemical
plants, basic process controls are incorporated to monitor and facilitate the basic
operation, control and automation of processes for economic improvement
opportunities [8; 9].

Process control systems can be considered information processing systems,

because they collect information regarding the status of the process by means
of the instruments in place. The information gathered is then compared with
design specifications and a decision is taken on what to do to move the process
to a desired state. The decision is ultimately sent to a final control element
(usually a control valve) that is used to implement the desired changes in the
process [8; 9].

Final control devices are elements of the control system that gather the flow
of information from the controller in the form of output signals and convert
it into a desired action that is able to physically and appropriately influence
the behaviour of the plant. The most commonly used final control element
in the chemical industries is the control valve. However, other final control
elements, such as agitator speed controls and conveyor belt speed controllers,
are also sometimes used [8; 9].

(1) Anderson, N.A. Instrumentation for Process Measurement and Control,

Chilton Book Company, Radnor, PA.
(2) DIN 19226 Parts 1 to 6 “Leittechnik: Regelungstechnik und
Steuerungstechnik” (Control Technology), Beuth Verlag, Berlin.
(3) Distefano, J.J., Stubberud, A.R. and Williams, I.J. Feedback Systems and
Control, Schaum’s Outline Series, 2nd ed., McGraw-Hill, New York,
p. 78, 1995.
(4) “International Electro-technical Vocabulary”, Chapter 351: Automatic
control. IEC Publication 50.
(5) Luyben, M.L. and Luyben, W.L. Essentials of Process Control, McGraw-
Hill, New York, NY, 1997.
(6) Marlin, T.E. Process Control: Designing Processes and Control Systems
for Dynamic Performance, 2nd ed., McGraw-Hill, New York, NY,
2000.
(7) Murrill, P.W. Fundamentals of Process Control Theory. Instrument Society
of America, Research Triangle Park, NC, 1981.
(8) Ogunnaike, B.A. and Ray, W.H. Process Dynamics, Modelling and Control,
Oxford University Press, Oxford, 1994.
(9) Seborg, D.E., Edgar, T.F. and Mellichamp, D.A. Process Dynamics and
Control, 2nd ed., Wiley, Hoboken, NJ, 2004.

4.1 Consider the process flow diagram provided below. The objective of the
process is to heat the oil from temperature T0 to a desired temperature TD
using process steam. In this case, what is the fail-safe design of the valve?
4.2 Study the automated cooling system for a large power-generating engine
illustrated by the process flow diagram presented below. What is the fail-
safe design of the valve?

4.3 Clearly describe the valve fail-safe design for the following process in the
event of utility failure.

(a) Outlet valve from cement processing

(b) Oxygen transportation to the furnace
(c) Furnace fuel
(d) Final product from gold processing plant
(e) Coolant valve in exothermic reactor
(f) The heating process discussed in the previous sections

4.4 Because the economic viability of an overall process is based significantly

on product purity, it is important for distillation columns to maintain
stable operation. Changes in composition and flow rate of the feed
stream are common disturbances in distillation column operation.
Incorrect functioning of controllers can undermine the effectiveness of
the product composition. Indicate the fail-safe design suitable for each
valve in a typical distillation process.
4.5 A control valve operates from 3 psi to 15 psi. The control signal is to have
a 40 gal/min flow rate. Express the signal input in psi if it is a linear valve
from 0 gal/min to 90 gal/min.
4.6 An equal percentage valve has a maximum flow of 50 m3/s and a minimum
flow of 2 m3/s. If the full travel is 3 cm, calculate the flow at 1 cm opening.
4.7 An equal percentage control valve has a rangeability of 32. If the maximum
flow rate is 100 m3/s, determine the flow rate at 2/3 open setting.
4.8 An actuator has a stem movement which at full travel is 30 mm. It is
mounted with a linear valve which has a minimum flow rate of 0 m3/s
and a maximum flow rate of 40 m3/s. What will the flow rate be when
the stem is at the following positions?

(a) 5 mm
(b) 200 mm

4.9 Calculate the pressure drop across the valve when the flow rate is given
as Q = 90 gpm and the flow coefficient is given as CV = 51.
4.10 Calculate the flow rate generated if the pressure drop across the valve is
10 ft and the flow coefficient is given as CV = 51.
4.11 Consider the process flow diagram below. Determine the size of the valve
required to control the liquid water flow when the maximum flow rate
is 63 gpm. The pressure of the heat exchanger and the total pressure of
the system are 6 and 49.4 psi, respectively. Size the valve to be used.

4.12 A system is designed to pump water at 70 °F from one tank to another

through a piping system with a total pressure drop of 150 psi. There is a
maximum design flow rate of 150 gpm, operating flow rate of 110 gpm,
and minimum flow rate of 25 gpm. The pipe diameter is 3 inches and
the water at 70 °F has a specific gravity of 1.0.