Comp PDF

CHAPTER - 3
Computer as a Controller
3.1 BASIC BUILDING BLOCKS OF COMPUTER CONTROL SYSTEM
Basic functions of computer-aided process control system are:
(a) Measurement and data acquisition.
(b) Data conversion with scaling and checking.
(c) Data accumulation and formatting.
(d) Visual display.
(e) Comparing with limits and alarm raising.
CD Recording and monitoring of events, sequence and trends.
(g) Data logging and computation.
(h) Control action.
A block diagram of computer-aided process control system IS shown III
Figure s.i
Computer system
v
Final
Set-point
Controller
Digital-to- control
yet)
equation
..
~
--+ ~
u
...
analog UfC) ~ element
Process
+ (digital
(control
algorithm)
converter
p..-
valve)
Analog-to
Measured Controlled
variable
Sensor
variable
digital ....
(measuring element)
...
converter
~
(a) Schematic diagram
)
3.2
computer Control of Process
3.3
Fig 3.2 shows the Role of computer in process control. After the
technological development of digital computer system, its use for measurement
and control application has tremendously increased. The basic objective of
computer-aided measurement and control is to identify the information flow,
and to manipulate the material and energy flow of given process in a desired,
optimal way. The requirement in terms of response time, computing power,
flexibility and fault tolerance are stricter, since the control is to be carried out
in real-time. Other difficulties, as a result of computer technology developments
are, a solution to the problem of complexity, flexibility, and geographical
separation of process elements (plant equipment to be controlled).
3.1.1 Role of computers in Measurement and Control (process control)
Digital computer control application exists to-day for two major areas III
the process industries, passive and active applications. Passive application.
involves acquisition and manipulation of process data where as, active
application involves manipulation of process as well. The passive application
deals predominantly with monitoring, alarming and data reduction systems. The
process data, after captured (measured) on-line, are sent to the data acquisition
computer through interface module. The smart instruments such as smart.
sensors, smart transmitters and smart actuators (final control element), are now
rvailable which have a microcomputer built into them. The smart instruments
help the operator to get real-time process measurement information and
automatic transmission is required form for further processing by the process
process.
As shown in fig 3.l(a), the controlled variable (output of the process) is
measured as before in continuous electrical signal (analog) form, and converted
into a discrete-time signal using device called analog-to-digital converter (ADC).
The value of discrete signal thus produced is then compared with the discrete
form of the set-point (desired value) inside the digital computer to produce an
error signal (e). An appropriate computer program representing the controller
,
called control algorithm, is executed which yields a discrete controller output.
The discrete signal is then converted into a continuous electrical signal using
a device called digital-to-analog converter (DAC), and the signal is then fed to
the final control element. This control strategy is repeated at some pre
determined frequency so as to achieve the closed-loop computer control of the
uri)
I
i{DBta
'---r---'L---,.---...JL..-.---''---r--J: output
I
I
I
I
I
I
I
---- 1
Hold
device
Process
uri)
(b) Block diagram
Digital
algorithm
Digital computer
Fig. 3.1 (a) & (b) Typical Computer-aided Process Control Systems
Fig. 3.2 In Computer-based control, the analog loop controller is
replaced by Computer programming
Data }:
acquisition :
'------,---'
I
I
I
I
I
I
I
I
Measurement
3.4 Computer Control of Process Q.omputer as a Controller
3.5
control computer. The smart instruments ensure that the actuator, transmitter
or sensor function according to design.
As illustrated in Fig 3.3, the tasks carried out by each Levels in the
automation hierarchy are as follows:
Fig. 3.3 Automation of computer aided process control
(b) System coordination and reporting
Collecting information on unit production, raw material consumption,
energy consumption and transmitting to the higher Level computer (Level-S),
Providing services to the operator's man-machine interface (MMI).
(c) Reliability assurance
Performing diagnostics on the various control equipments to help in
detecting the fault and keep updating hot-back up (stand by) system if connected
in the hierarchy.
Field Level (Level-G)
Control Level (Level-l)
Supervisory Level (Level-2)
Plant Level (Level-S)
Corporate Level (Level-4)
Supervisory Level (Level-2)
(a) Control enforcement
Responding to any emergency condition in its region of plant cognizance.
Optimization of the processes under its control as per established production
schedule and carry out all established process operational schemes or operating
practices for these processes.
Control Level (Level-l)
(a) Control enforcement
Maintaining direct control of the plant units under their cognizance,
detection of any emergency condition is these units and taking appropriate
response.
Field Level (Level-D)
(a) Measurement and transmission of field parameters to the Control
Level (Level-l) computer control systems.
Process control level
control)
Supervisory control level
____ optlmlzatlon
Plant control Level
_____ Scheduling)
Corporate Level
__________ .:. __ Information
Field Level
Optlmizatlon
Level
Control Level
Level 4
Level 3
Level 2
Level 1
Level 0
- - - - - - - - - - - - - - - - -
Measurements level
________ L-- ('!n!!oJ:!l!- Transducer)

The major application of digital computers is in process control and plant
optimization. Computer control systems, once prohibitively expensive, can now
be tailored to fit most industrial applications on a competitive economic basis.
The advances in the use of computer control have motivated many and changed
the concepts of the operations of industrial processes. Video display terminals
now provide the four for operators to supervise the whole plant from a control
room. Large panels of instruments, knobs and switches are replaced by a few
keyboards and screens. Control rooms are now much smaller and fewer people
are required to supervise the plant.
Process control computers now have the capability to implement
sophisticated mathematical models. Plant managers and engineers can be
provided with comprehensive information concerning the status of plant
operations to aid effective operation. With the use of microprocessor-based
instruments and new emerging techniques, it is possible for automatic tuning
of controller parameters for best operating performance. The expert systems and
advanced control techniques such as model based predictive control, are being
applied with the help of computer for optimization of the process operation.
3.1.2 Detailed Task listing of computer control systems
3.6 Computer Control of Process
3,7
Multiplexer
(Telemetry)
Signal
Conditioner - 1
Signal
Conditioner - 2
Signal
Conditioner - 3
Signal
Conditioner - 4
Transducer - 1
e
19
<Il
E
e! Transducer - 2
m
a..
1:
m
a::

Transducer - 3
'!2.
s:
a..
Transducer- N
(b) Finance and Accounting
Logging and monitoring of company's sales and expenditures, annual
accounting, profit and loss account for the management information.
BlockDiagram of DataAcquisition system
3.2 DATA ACQUISITION SYSTEM (DAQ)
A data acquisition (DAQ) system is used for the measurement and
processing of plant signal data before it is displayed on the operator desk or
permanently recorded. A block diagram of a PC (computer) based data
acquisition is shown in fig 3.'1. It consists of individual transducers (sensors)
for measurement of physical plant parameters (such as temperature, pressure,
flow, etc.i). After measurement, the transducer data is fed to the signal
conditioning device to bring the signal level up to a sufficient value to make it
useful for conversion, processing, indicating and recording. Signal conditioner
is used to amplify, modify or select certain portion of signals.
(b) Plant coordination and operational data reporting
Collection and maintenance of process/production database. Keeping
inventory of raw material and energy consumption for the units. under its
control. Maintaining communication with the higher and Lower Level computer
systems. Providing services to the operator's man-machine interface (MMI).
Plant Level (Level-S)
Performing diagnostics on the various control equipment to help in
detecting the fault and keep updating hot-back up (standby) system if connected
in the hierarchy.
(a) Production planning and scheduling
Prepare the immediate production schedule is its area of control. Locally
optimize the costs for its individual production area as a basis for modifying
the production schedule established by the production control computer system
(minimize energy usage and maximize production).
(b) Plant coordination and operational data reporting
Preparing production reports, maintenance of plant inventory about
materials and energy usage, maintaining communications with higher and
Lower Level computers, operational related data collection and off-line analysis
for future prediction and usage, providing services to the operator's
man-machine interface (MMI).
Performing diagnostics on the various control equipment to help in
detecting the fault and keep updating hot-backup (standby) system if connected
in the hierarchy.
Corporate Level (Level-d)
(a) Sales and Marketing
Customer order management, order booking, and transferring order
information to plant Level (Level-S) computer. Market forecasting and market
intelligence, customer database management, market survey.
Fig. 3.4 Block Diagram of Data Acquisition System
The output of the signal conditioner is fed to the multiplexing (telemetry)
device. With the help of multiplexing all individual signal data (called lower
bandwidth communication channels) are combined and transmitted over a
higher bandwidth channel. At the receiving end, de-multiplexing recovers the
original lower bandwidth channels. It scans across a number of analog signals
and time-sharing them sequentially into a single analog output channel. The
multiplexed data is converted into digital signal with the help of analog-to-digital
converter. The converted digital signals are fed to the computer for further
3.8 Computer Control of Process Computer as a Controller
3.9
processing, mathematical computation, storage, etc. The final and processed data
is either displayed on electronic digital display panel or recorded on magnetic
media and/or chart recorders.
transducers, and pressure transducer which convert force, rate of flow, and
pressure to electrical signals. In each. case electrical signals produced are
proportional to the physical parameters they are monitoring.
Data acquisition system is generally a process of collecting input signal
in analog or digital form, as rapidly,as accurately, completely and economically
as necessary and possible. Data acquisition system can be of two types: an
analog data acquisition system and a digital data acquisition system. Analog
data acquisition system deals with information in analog form where as digital
data acquisition system handles information in digital forms.
Today, most scientists and engineers are using personal computers with
expansion buses for laboratory research, industrial control, and test and
measurement. Obtaining proper results from a PC-based DAQ system depends
on each of the following system elements, as shown in figure 3.5 personal
computer, transducers, signal conditioning, DAQ hardware and software.
(ii) Isolation
(iii) Filtering
(iv) Excitation
Transducers
Transducers change physical phenomena into electrical signals. For
example, thermocouples, Rl'Ds, thermistors, and IC sensors convert temperature
into a voltage or resistance. Other examples include stain gauges, flow
(v) Linearization
Amplification
The most common type of conditioning is amplification. Low-level
thermocouple signals, for example, should be amplified to increase the resolution
and reduce noise. For the highest possible accuracy, the signal should be
amplified so that the maximum voltage range of the conditioned signal equals
the maximum input range of the analog-to-digital converter (ADC). Very high
resolution reduces the need for high amplification and provides wide dynamic
range.
Isolation
Another common application for signal conditioning is to isolate the
transducer signals from the computer for safety purposes. The system being
monitored may contain high-voltage transients that could damage the computer.
An additional reason for needing isolation is to make sure that the readings
from the DAQ device are not affected by differences in ground potentials or
common-mode voltages. When the DAQ board input and the signal being
acquired are each referenced to "ground", problems occur if there is potential
difference between the two grounds. This difference can lead to what is known
Transducers
DataAcquisition
Hardware
Signal
Conditioning
Fig. 3.5 Typical Data Acquisition (DAQ) System
@
Software
3.10
Computer Control of Process
3.11
as ground loop, which may cause inaccurate representation of the acquired
signal, or if too large, may damage the measurement system. Using isolated
signal conditioning modules eliminates the group loop and ensures that the
signals are accurately acquired.
Filtering
Filtering removes unwanted noise from the signal that are being tried to
be measured. A noise filter is used on DC-class signals such as temperature to
attenuate higher frequency signals that can reduce the accuracy of the
measurement. AC-class signals such as vibration often require a different type
of filter known as an antialising filter. Like the noise filter, the antialising is
a lowpass filter; however, it has a very steep cut off rate, so that it almost
completely removes all frequency components higher than the input bandwidth
of the board. If the signals- are not removed, they erroneously appear as signals
within the input bandwidth of the board.
Excitation
Signal conditioning also generates excitation for some transducers. Strain
gauges, thermistors, and RTDs,for example, require external voltage or current
excitation. Signal conditioning modules for these transducers usually provide
these signals. RTD measurements are usually made with a current source that
converts the variation in resistance to a measurable voltage. Strain gauges are
resistance devices in a wheatstone bridge configuration, which often require
bridge completion circuitry and excitation sources.
Linearization
Another common signal conditioning function is linearization. Many
transducers, such as thermocouples, have a nonlinear response to changes in
the phenomenon being measured. Therefore, linearization routines for
thermocouples, strain gauges, and RTDs are included in the DAQ application
software.
3.2.1 Data Acquisition Hardware
Data acquisition hardware includes the following functions
Analog Input
The analog input specifications can give information on both the
capabilities and the accuracy of the DAQ product. Basic specifications, which
are available on most DAQ products, tell about the number of channels,
sampling rate, resolution, and input range. The number of analog channel inputs
will be specified for both single-ended and differential inputs on boards that
have both types of inputs.
Single ended inputs are referred to a common ground point.
Non-referenced single-ended inputs are all referenced to AISENSE, not a
common ground. The potential at this node can ~ with respect to the system
ground. These inputs are typically used when the input signals are-high level
(31 V), the leads from the signal source to the analog input hardware are short
(15 ft), and all input signals can share a common ground reference.
If the signal do not meet these criteria, differential inputs should be used.
A differential, or non-referenced, measurement system has neither of its inputs
tied to a fixed reference. An ideal differential measurement system responds
only to the potential difference between two terminals-the (+) and (-)inputs.
Any voltage measured with respect to the instrumentation amplifier ground
present at both amplifier inputs is referred to as a common-mode voltage. The
term common-mode voltage range describes the ability of a DAQ board in
differential mode to reject common-mode voltage signals. In differential mode,
noise errors are reduced because the common-mode noise picked up by both
leads is cancelled out.
Sampling Rate
This parameter specifies how often conversions can take place. Using a
faster sampling rate, more inputs can be acquired in a given time, providing a
better representation of the original signal. As shown in figure 3.6 (a), all input
signals must be sampled at a sufficiently fast rate to faithfully reproduce the
analog signals obviously, if the signal is changing faster than the DAQ board
is digitizing, errors are introduced into the measured data. In fact, data that
is sampled too slowly can appear to be at a completely different frequency. This
distortion of signal is referred to as aliasing, as shown in figure 3.6(b).
3.12
3.13
(a) Adequately Sampled
low-frequency signals, such astemperature and pressure, while maintaining the
cost benefits of continuous scanning. This method scans the input channels at
one interval and uses a second interval to determine the time before repeating
the scan.' The, input channels are scanned within microseconds, creating the
effect' of simultaneously sampling the input channels.
10.00 r ~ _ _
140 60 40
~ ~
20 o
Fig. 3.7 Digitized sine wave with 3 bit Resolution
7.50 -
6.75
3.75
5.00
6.25
1.25
2.50 --------------
Multiplexing
A common technique for measuring several .signals with a single ADC is
multiplexing, A multiplexer selects and routes one channel and repeats. Because
the same ADC is sampling many channels, the effective rate of each individual
channel is reduced in proportion to the number of channels sampled.
Resolution
The number of bits that the ADC uses to represent the analog signal is
the resolution. The higher the resolution, the higher' the number of divisions
the voltage range is broken into and therefore, the smaller the detectable voltage
change.
Figure 3.7 shows a sine wave and its corresponding digital image as
, . .
obtained by an ideal 3-bit ADC. A 3-bit converter divides the analog range in
/to 2
3
or 8 divisions. Each division is represented by a binary code between 000
I . . .
. and 111. Clearly, the digital representation is not a good representation of the
original analog signal because information has been lost in conversion. By
(b) AllasedDueto Under-sampling Rate
Sampling Methods
When acquiring data from several input channels, the analog multiplexer
connects each signal to the ADC at a constant rate. This method, known as
continuous scanning, is significantly less expensive than having a separate
amplifier and ADC for each input channel. BeC\ause the mUltiPlexer, switches
between channels, a time skew is generated between each channel sample. An
analog multiplexer connects one of the input signals to the ADC for processing.
This method is appropriate for applications where the time relationship between
sampled points is unimportant. For those applications where the time
relationship between inputs is important (such as phase analysis of AC signals),
simultaneous sampling is needed. DAQ products capable of stimultaneous
sampling usc sample-and-hold circuitry for each input channeL Simultaneous
sampling can be simulated without paying for additional sample-and-hold
circuitry. Internal scanning creates the effect of simultaneous sampling for
Fig. 3.6 Effects of too Low a Sampling Rate
, /
According the Nyquist theorem in control system, inputs must be sampled
at least twice the rate of the maximum frequency component in that signal to
prevent aliasing. The frequency at one-half the sampling frequency is referred
to as the Nyquist frequency. Theoretically, it is possible to recover information
about signals with frequencies at or below the Nyquist frequency. Frequencies
above the Nyquist frequency will alias to appear between DC and the Nyquist
frequency. For example, audio signals converted to electrical, signals by a
microphone commonly have frequency components up to 20 kHz. A board with
a sampling rate greater than 40 kS/s is needed to properly acquire this signal.
3.14
Q?mputer as a Controller 3.15
increasing the resolution of 16 bits, however, the number of codes from the ADC
increases from 8 to 65, 536 and therefore, an extremely accurate digital
representation of the analog signal can be obtained, if the rest of the analog
input circuitry is designed properly.
3.2.2 Data Acquisition Software
DAQ hardware without software is of little use-and DAQ hardware with
poor software can be worse. Following two types of software is used:
Driver software
Driver software is the layer of the software that directly programs the
registers of the DAQ hardware, managing its operation and its integration with
the computer resources, such. as processor interrupts, direct memory access
(DMA), and memory. Driver software rides the low-level, complicated details of
hardware programming while preserving high performance, providing the user
with an easy-to-understand interface.
Driver functions for controlling DAQ hardware can be grouped into analog
1/0, digital I/O, and timing I/O. Although most drivers will have this basic
functionality, driver software does more than simply getting data on and off the
board. If has the following functionality..
(i) Acquire data at specified sampling rates.
(ii) Acquire data in the background while processing in the foreground
(continuous data acquisition)
(iii) Use programmed 1/0, interrupts, and DMA to transfer data.
(iv) Stream data to and from disk.
(v) Perform several functions simultaneously.
(vi) Integrate more than one DAQ board.
(vii) Integrate seamlessly with signal conditioning equipment.
Application Software
Application software uses driver software to control the DAQ hardware
efficiently. It adds analysis and presentation capabilities to the driver software.
The application software also integrates instrument control such' as GPIB (as
general purpose interface bus), RS-232, PXI (peripheral component interact
extensions for instrumentation), and VXI (virtual extension for instrumentation)
with data acquisition.
3.2.3 Data Logging
The efficient operation of a manufacturing process may involve the
interplay of many factors, such as production rates, material costs, and
efficiencies of control. When the process requires implementation of many
process-control loops, then the interaction of one stage of the system with
another often can be analysed in terms of the controlled variables of the loops.
An example of this is. the rate of production of one loop, expressed as a. flow
rate, which serves as a determining factor in the production rate of a following
control system. Historically, an understanding of this type of interaction
required analysis, after the fact, of strip-chart recordings taken from process
parameters during a production run. Such analysis, carried out by trained
personnel, may then dictate settings of operational limits of future production
runs.
With the development of high-speed digital computers with mass digital
storage, it became possible to record such data continuously and automatically,
display the data on command, and perform calculations on the data to reduce
it to a form suitable for evaluation by appropriate technical individuals.
Fixed Loggers
Let us assume the process is under the control of many analog
process-control loops and there is provision for analog process variable
measurements to be available as a commonly scaled voltage. Thus, some signal
conditioning converts all measurements into a given range, often a specified
voltage range, as required by a data-acquisition system. A brief accounting of
the elements of the system is given later.
3.16 computer Control of r o e s ~
.computer as a Controller
-
3.17
Fig. 3.8 General features of a data-logging system using a computer
Data-Acquisition System (DAS)
Alarms
An important part of any data-logging circuit is an alarm system} that
monitors inputs from excursions beyond some specific limits. Withscan rates
of the data as high as 5000 per second, it is possible for a computer to maintain
tight vigilance over variable values. Everytime the computer inputs a particular
variable, the value is compared to its preset limits,which, if exceeded, trigger
an alarm.
Computer
The computer, of course, is the central element in the system. Through
programming, the computer accepts inputs and performs prescribed reductions
of the data through mathematical operations. The results are evaluated by
further programmed tests to oversee the operation of the entire process from
which the inputs are taken. Projections of future yields, evaluations of efficiency,
deviation trends, and many other operations can be performed and made
available to process personnel.
Peripheral Units
. The peripheral units are the support equipment to communicate computer
operations to the outside world. These units include the operator console where
the programs are entered and through which commands can be given to initiate
specific actions, such. as calculations and data outputs by the computer. The
console usually has a CRT/keyboard and a typewriter unit for inputs and
outputs. A mass-storage system, such as magnetic tape, is used to store data,
such as periodically sampled inputs from the process, that can be used in later,
more detailed analysis of process performance.
SP
Alarms
Address
lines
Process
}-i
Line printers,
CRTs and so on
Data
acquisition
. system
Analog {
process-control
loops
The data-acquisition system, IS the switchyard by which the computer
inputs samples of process variable values. The concept of "samples" of these
values is an important topic. The reason for concern over this point is that there
are situations where the sample rate can be such that erroneous information
about the variable time dependence results. The rate at which samples of a
process variable can be taken depends on how long it takes for the DAS to
acquire a value, how long it takes the computer to process the value, and how
many other variables are to be sampled.
3.3 SUPERVISORY CONTROL AND DATA ACQUISITION SYSTEMS
(SCADA)
3.3.1 Introduction
A natural extension of a computer data-logging system involves computer
feedback on the process through automatic adjustment of loop setpoints. As
various loads in a process change; it is often advantageous to alter set points
in certain loops to increase efficiency or to maintain the operation within certain
3.18 Computer Control of Process Computer asa Controller
3.19
precalculated limits. In general, the choice of setpoint is a function of many
other parameter in the process. In fact, a decision to alter one setpoint may
necessitate the alteration of many other loop set points as interactive effects
are taken into account. Given the number of loops, interactions, and calculations
required in such decisions, it is more natural and expedient to let a computer
perform these operations under program control.
After having dealt with the basic hardware modules of a real-time system,
let us first concentrate on Supervisory Control and Data Acquisition (SCADA)
system, since it is the first step towards automation. The basic functions carried
out by a SCADA system are:
Channel scanning
Conversion into engineering units
Data processing.
Figure 3.9 & 3.10 the block schematic of SCADA. Before considering other
features of SCADA, let us discuss the basic functions.
Fig. 3.9 In computer supervisory control, the computer monitors
measurements and outputs the loop setpoints
-
--+
Display
module
j[~
T
Micro
Signal Analog
processor
Serial
conditioner
I
Input --. interface I +
!e
module
.L,
module
Memory
mcdute
--+
(/)
Alarm
----+ annunciator
'C
module
c
lUOl I Digital
gj.fi
I
Input
I
!
Communication
:9""
I Module
~
with central
computer
External
TImer
~ Counter
clock
module
-
Fig. 3.10 Supervisory control and data acquisition system
P
R
o
C
E
S
S
3.3.2 .Channel scanning
There are many ways III which microprocessor can address the varIOUS
channels and read the data.
Polling
The microprocessor scans the channel to read the data, and this process
is called polling. In polling, the action of selecting a channel and addressing it,
is the responsibility of processor. The channel selection may be sequential or in
{
Data
output
system.
Alarms
Address
Line Printers
eRTs and so on
,--- - - -------------
}i
Data
acquisition
system
Analog process- {
control loops
3.20 Computer Control of Computer as a Controller
3.21
any particular order decided by the designer. It is also possible to assign priority
to some channels over others, i.e. some channels canbe scanned more frequently
than others. It is also possible to offer this facility of selecting the order of
channel addressing and channel priorities to the operator level, these
facilities as dynamic.
The channel scanning and reading of data requires.. the following actions
to be taken.
Sending channel address to multiplexer
Sending start convert pulse to ADC
Reading the digital data.
For the digital data at ADC output, the end of conversion signal
of ADC chip can be read by processor and when it is 'ON', the digital data can
be read. Alternatively, the microprocessor can execute a group of instructions
(which donot require this data) for the time which is equal to or greater than
conversion time of AVC and then read the ADe output. Another modification
of this approach involves connecting the end of conversion line to one of the
interrupt request pins of the processor. In this case the interrupt service routine
reads the ADC output and stores at predefined memory location.
The channels may be polled sequentially, in which case the channel
address in first step above increases by one every time or they may be .scanned
in some other order. In the later case, a channel scan Array can be maintained
in memory (fig The scan array contains the addresses of the channels in
Channel
number
9
10
1
2
ASCN 5
(Scan array) 9
.3
4
2
6
8
9
n 7
Fig. 3.11 Channel scan array
the order in which they should be addressed. The ASCN array in fig. 3.11 has
. 9, 10, 1, 2.... as entries in sequence. Thus the first channel to be scanned will
be channel 9, followed by 10, 1, 2.... As the pointer reaches the last entry in
the array, the first entry is again taken up (i.e channel 9 is scanned).
. If a channel number is repeated in the array, then that particular channel
will be scanned repeatedly. Thus it is possible to scan some channels more
frequently than others. This gives them higher priority over others. In figure
(3.11) the channel 9 is scanned 3 times, channel 2 is scanned 2 times while
other channels are scanned once during a cycle. Figure 3.12 shows the flowchart
for one scan cycle.
Fig. 3.12 Scan cycle flow chart
The processor may scan the channels continuously in the particular order
as illustrated by the flowchart or the channels may be scanned after every fixed
time period. The. second approach requires a timer/counter circuit whose output
is connected to interrupt request input. The scan routine for one channel is
incorporated in 'Interrupt Service Routine'. It is also possible to make the time
gap between two channels as variable. This would require a n x 2 dimension
scan array (fig The Interrupt Service Routine fetches the time gap value
for next channel, loads the timer/counter with the value and initiates the
timer/counter before returning to main program.
/
3.22 computer Control of Process Computer as a Controller
3.23
The scan array may be decided at the design stage of SCADA and fused
permanently in (ROM) Read Only Memory. Thus the channels are always
scanned in that particular order. However, it may be desirable to offer the
facility of changing the sequence at the operator level. The operator may like
to take this action depending on the condition of the plant being monitored. As
an example, if at any instant operator finds out that the transducerconnected
to channel 9 has generated some .fault, then he might take decision to by pass
the channel 9, as otherwise its data will be taken into analysis to produce
incorrect result. In another situation, it may become necessary to scan some
channel more frequently for some time to observe its response to some
modifications incorporated in the system. The operator should thus be able to
insert new scan array at any time. This facility may be provided through a key
switch which may be connected to interrupt request input of the processor. The
Interrupt Servicing Routine will accept new scan array and store in place of
the old one.
Channel Time
number gap
9 FF (H)
10 OF (H)
1 FO (H)
2
5
9
3
4
2
6
a
9
7
n
Fig. 3.13 Scan array with time
Interrupt Scanning
Another way of scanning. the channels may be to provide some primitive
facility after transducer to check for violation of limits. It sends interrupt request
signal to processor when the analog signal from transducer is not within High
and Low limits boundary set by Analog High and Analog Low signals. This is
also called Scanning by Exception. When any parameter exceeds the limits then
the limit checking circuit would send interrupt request to microprocessor which
in turn would monitor all parameters till the parameter values came back within
pre-specified limits. This allows a detailed analysis of the system and the
problems by the SCADA system.
'I'he limit checking circuit for one channel is shown in Fig 8.14. Two analog
comparators check whether the input signal is within high and low limits. The
output is ORed and the final output is used as interrupt request to
microprocessor. This limit checking should not be constructed an alarm.
condition, but the condition for the start of any abnormality which may generate
alarm condition or may be controlled by the system before any alarm condition
is reached. Therefore, the system should be watched closely during this time.
3.3.3 Conversion to Engineering Units
The data read from the output of ADC should be converted to the
equivalent engineering units before any analysis is done or the data is sent for
display or printing. For an 8-bit ADC working in unipolar mode the output
ranges between 0 and 255. An ADC output value will corresponds to a particular
engineering value based on the following parameters.
1. Calibration of transmitters.
2. ADC mode and digital output lines.
Fig. 3.14 Interrupt request generation on limit violation
The transmitter output should be in the range of 0-5 V or 4-20 rnA range.
Depending on the input range of measurand value for transmitter, a calibration
factor is determined. If a transmitter is capable of measuring parameter within
the input range Xl and X
2
and provides 0-5 V signal at output ,then calibration
factor is
Computer as a Controller' 3.25
If we are converting this signal to digital through an 8-bit ADC (Input
range 0-5V) in unipolar mode then 5V =255 and OV=0,
. 255
ie 1 Volt=-
. 5
Thus the conversion factor is ADC output
. HigherLimit
10
20
LowerLimit
02
09
ADCoutput
255
5 -
X2 -Xl . . .
5 . engmeermg urnts
Fig. 3.15 Limit array
1
X2-XI ... ..
= 255 engmeermg units
If the ADC output is Y then the corresponding value in engineering units
. . . .
(X
2-XI
)
Conversion factor is therefore -=------==-
255
The conversion of ADC output to engineering units, therefore,involves
multiplication by conversion factor. The conversion factor is based on the ADC
type, mode and the transmitter range. This multiplication can be achieved by
shift and add method in case of 8-bit microprocessor; For 16-bit microprocessor,
a single multiplication instruction will do the job.
3.3.4 Data Processing
The data read from the ADC output for various channels is processed by
the microprocessor to carry out limit checking and performance analysis. For
limit checking the Highest and Lowest Limits for each channel are . stored in .
an array (Fig 3.15), When any of the two limits is violated for any channel,
appropriate action like alarm generation, printing etc., is initiated. The limit
array shown (Fig 3.15) simplifies the limit checking routine. Through this; the
facility to dynamically change the limits for any channel may also be provided,
on the lines similar to scan array. .
In addition to limit checking, the system performance may also be
analysed and report could be generated for the manager level. This report will
enable the managers to visualise the problems in the system and to take decision
regarding system modification or alternate operational strategy to increase the
system performance. The analysis may include histogram generation, standard
deviation calculation, plotting one parameter with respect to another and so on.
The software can be written depending on the type of analysis required.
The facilities like scanning, limit checking, etc, could be incorporated in
a simple way using arrays in software. Let us first examine channel scanning,
using scan array in memory which contains the channel numbers in one
. particular order of scan.. For two applications using the same hardware, only
the scan array will have to be changed for the scanning. The same argument
is valid for limit checking. The conversion to engineering units can also be
achieved by storing the conversion factors for different channels in any array.
'I'he conversion factors stored in the array will change with the transducers i.e
it will be different for different applications. The simple print programme which
takes the data in particular format from memory and sends for printing may
also be same for different applications. However, the designer will have to write
specialised routine for arranging the data in memory in any particular format
required. The analysis and report generation programs will be. application
dependent and will have tobe written separately for different applications. Here
also the routines for histogram etc. may be used by different applications if they
are written in generalised manner.
3.26
3.3.5 Distributed SCADA System
Computer as a Controller 3.27
In any application, if the number of channels are quite larger then in
order to interface these to processor, one has to use multiplexers at different
levels. Figure 3.16 shows the interfacing of 256 channels, using 17 multiplexer
of 16 channels each.
The 8 address lines are used to address 256 channels. Out of the 8-address
lines, upper four are used to select a particular multiplexer and lower four lines
are used to select a particular channel in the multiplexer.
The 8-bit channel address thus directly maps into channel number and
can be manipulated in any way. The other parts are same as described earlier.
This approach will be suitable for the processor which are basically slow. Even
if a channel is scanned only once in every scan, it is only after 255 channels
have been scanned, limit checking and analysis have been performed,. a
particular channel will be addressed again. This is not acceptable in many
processes.
For the process plants where the structure does not suit, the only
alternative is to use more than. one SCADA system and distribute the channels
among them. But, for performance analysis on the process plant, it is mandatory
that the data from various channels should reach a central location where it
can be consolidated and analysed to generate the reports on plant performance.
Figure (3.17) (a) and (b) show the interfacing of number of SCADA systems
with central computer in star configuration and Daisy chain configuration
respectively.
The SCADA system directly connected to transducers are called nodes and
are the same as the systems described earlier. They scan the channels using
one of the techniques discussed, earlier; convert the data into engineering units,
perform the limit checking, generate alarm, if data item crosses the limit and
generate print out.
In addition to these functions, the data regarding the channels in the node
are transferred to central computer which analyses the system performance and
generates print out. The print outs are generated by exception i.e unnecessary
data is not printed at any point. At the node level, the print out is required for
the operators to run the system. Depending on the node performance, operator
.---
I"- '5

113
.-
L/)
w IQd

.... -
<i! " ,-+-l"-
Gl
E

....
ld
'---
F

( .
<OOOCW(/)(/) CD::;)(/) )
( 01- CD::!(/) )
t t

IrE

113
...
'5",113
1-
113 '5",
1l'3
1
-
'5:;; c.t: ,9-t:

"8.
E&. 8&. I'- 00.
0
IJ
f-+-
o
r;-
UOISJ811UOO
lOPU3 <"0

7
lH
113

'"

xI'-

.

'"

------ "0<:
"0-
<C
I
I
x
*-=-

x
I
x

::ii
e- ",;---.;

",1----;,;1 ",r--"il ",r ",-I
3.29
The concept of local area networks or microprocessor interconnections can
be used in case of distributed SCADA system very effectively. Thus we conclude
that the distributed SCADA is the ultimate solution for complex process plant
monitoring.
the input and output material flow and the fuel consumption etc. may be more
helpful.
3.4.1 Introduction
3.4 DIRECT DIGITAL CONTROL
CHI
n-l 2 2
Sensors
CPU
-----------
Serial
interface
module

J
Serial Serial Serial
. Serial
..-
interface interface interface interface
module module module module
-------- -------- -------- --------
CAS,
CAS
2 DASn-'
CAS
n
.
--- --- --- ---
.....-
-
Process
CH
. CH-M
1 CHI CH CHM CH CHM
CH2 CHM
n
CH
(b)
Fig. 3.17 (a) Distributed SCADA structure (Star configuration), and
(b) Distributed SCADA structure (Daisy chain configuration)
may decide to monitor any channel more frequently, change the limits etc. The
print out at the central node is required for the managers to take long term
. decision to optimise the performance. The details on the channel performance,
limit violation arc not required at this level. On the other hand, histogram on
[3pervisors ..
Supervisory
control level
Console
computer
-----c-
1
\
--
Operator's Analog
displays controllers
1
Specialised
I
controllers
1
Process
The advent of microprocessor has changed the field of process control
completely. The tasks which were performed by complex and costly
minicomputers are now easily programmed using microcomputers. In the past
computer was not directly connected to the process but was used for supervision
of analog controllers: The analog controllers were interfaced to the process
directly as well as through specialised control for dedicated functions (Fig
The analog controllers and specialised controllers were called level 2 and level
1 control respectively.
From other
CAS
Serial
interface
module
To other
DAS
(a)
Serial
interface
module
CH
1
Serial
interlace
module
DAS,
CPU
Serial
Interface
module
Process
CH
1
CH
2
CH-M,
Sensors
Fig. 3.18 Supervisory computer level
/
Qomputer as a Controller 3.31
'rho emergence of economical and fast microprocessor has made analog
controllers completely out-dated, as the same functions can be performed by
digital computers in more efficient and cost effective way.
3.4.3 DDC Software
The main part, DDCsoftware is program for control loops. There are two
algorithms for programming a three mode PID control loops.
(1)
median valve position
value position at time ti
Proportional constant == 100/ PE
(where PE == proportional band in percent)
integral constant
derivative constant
error at instant t
n
== (8 - V,J
value of controlled variable at instant til
set-point.
ri
S: i : f
Yn==Kpcn+KD. 6.t +kI .
o
Position Algorithm
Velocity Algorithm.
where
3.4.3.1 The Position Algorithm
The three-mode controller can be represented by,
The integral term at any given instant t
n
is equal to the algebraic sum
of all the control forces generated by the integral control action from the
beginning to that instant.
The PID control can be realised with a microprocessor based system, if
only the above equation is implemented in the software. Apparently, it is very
difficult to write the software for implementing the above equation for a
microprocessor based system. However, the above equation can be modified such
that its soil ware implementation becomes easy. The modifications are discussed
in the following section.
Y
n
Yo
Wt'
K
p
KI
KD
en
V
n
8
DAC H Actuator
Memory
Serial
I/O ports
Timer/Counter
I Clock
L
MUX

I
I
t
I
I
I
I
I
I
I
I
l
Fig. 3.19 Direct digital control
The microprocessor performs the following tasks.
1. It reads the various process variables from different transmitters
through multiplexer and ADC.
2. It determines tho error for each control loop and executes control
strategy for each loop.
3. It outputs the correction value to control value through DAC.
3.4.2 DDC Structure
The DDC (Direct Digital Control) directly interfaces to the process for the
data acquisition and control purpose. That is, it has necessary hardware for
directly interfacing (opto-isolator, signal conditioner, ADC) and reading the data
from process. It should also have memory and arithmetic capability to execute
required P, P + I or P + I -v D control strategy. At the same time, the interface
to control valve should also be part of DDC. Figure shows the various
functional blocks of a direct digital control system. These functional blocks have
been described in number ofbooks on microprocessor. The multiplexer acts like
a switch under microprocessor control. It switches and presents at its output
the analog signal from a sensor/transmitter. The analog to digital converter
converts the analog signal to digital value.
computer as a CO(1troller
-
3.33
Thus, differential term, can be expressed as
Yes
Proportional P= K
p
x En
Error En=Set point - MeasuredValue
Reinitialise sampling Intervalcounter (ST)
No
Computer Control of r o e ~
Thus integral term can be represented as
3.32
and the differential term, KD . L1 e / L1 t at any instant t
n
is proportional to
the rate of change of the error.
en - e
n
- 1
KD----
L1t
Integral I = In+1n-1
where en is the current error and
StoreEn-1
ST = Sampling interval counter
Differential D= ~ x Ed
Fig. 3.20 Flow chart of PIO control
Trapezoidal rule for integral term
The integral term can be represented by using trapezoidal rule
(2)
n
en - en - 1 1
Yn =Kp . en + KD . L1 t + KI Lei L1 t + Yo
t=o
en - 1 is the previous error calculated at instant t
n
_ r-
Thus, with these modifications the three-mode controller equation will
become:
The integral and the differential control forces are dependent upon the
interval between the two consecutive terms. This interval is the inverse of the
rate at which the value of the controlled variable is measured i.e the sampling
rate. Hence the provision for defining the sampling rate should be made
available in the software.
n
"" ei +ei - 1
L..J 2 L1t
t=o
This will give better accuracy than previous term,
The flowchart for calculating PID control output based on the above
equation (Eq.2) is shown in Figure 3.20.
The 'sampling interval counter', the set-point, the proportional constant
IS), the integral constant KI and the derivative constant KD are defined by the
user.
The two modifications that can be performed on above equations are
Trapezoidal rule for integral term and interpolation technique for derivative
term.
based on rectangular rule.
n
t=o
(3)
3.34 Computer Control of
3.35
Interpolation technique for derivative term
The first difference in the derivative term, (en - en -1) is affected by noise,
and thus differentiation is sensitive to data error and noise. The noise can be
reduced by using analog or digital filters.
However the technique commonly used III interpolation method with
four-point control difference technique (Fig. 3.21).
Thus, with these two modifications the controller equation becomes
(7)
In position algorithm computer recalculates the full value of the valve
setting at each sampling interval. As shown in Fig 3.20 the analog signal is
sent to valve actuator through DAC.
Fig. 3.21 Interpolation technique for derivative term
________ n " n l
OV Full scalevoltagefor control valve .,
+Ve
In velocity algorithm, the computer calculate the required change in valve
position. The output is digital pulse train which can be directly used in case
valve is stepper motor driven. In case of other valves, stepper motor combined
Fixed Voltagecorresponding
point to newposition
-- _-----------1--Lj
I L'Changein valveposition
: I by velocityalgorithm
I '
,.... ---
I '
I '
I '
I :
, ,
, ,
I I
I I
I I . Movablepoint drivenby
: : stepper motor
Tho position algorithm has distinct property that it maintains its own
reference in Yo' However it has two drawbacks namely lack of bumples transfer
from manual to auto switching and set wind-up due to integral saturation in
test mode. These drawbacks are not present in velocity algorithm.
3.4.8.2 The Velocity Algorithm
In number of control loops, the final control element is stepper motor or
stepper motor driven valve. In such cases, the requirement at the computer
output will be a pulse train specifying the change in the valve position. Thus
output of position algorithm cannot be used, since it gives the new position of
the valve, in absolute term.
(4)
(5)
V' =V n + Vn - 1 + Vn - 2 +Vn - 3
4
v =.1 rV;1 - V'" + Vn 1 - VI' +V' - Vn - 2 + V' - V
n
.. 3 1
t 4 L 1.5 t 0.5 t 0.5 t 1.5 t J
Since set-point is constant,
Let Vn, V
n
r- V
n
2 and V
n
. 3 be the values of controlled variable at
current and their previous consecutive sampling intervals.
(6)
Fig. 3.22 Slide wire arrangement
3.36
At t n _ 1 i.e at previous instant, the valve position
where Y
n
is valve position at t
n
(13)
1 [en + en - 1 ]
+- .6.t
KI 2
n
Y
n
= .6. Y
n
+Y
n
- 1
=.6.Yn+[ .6.Y
n
- 1 +Y
n
-
2
]
=.6. v, + .6. Y
n
- 1 + [ L\ Y
n
- 2 + .6. Y
n
- 3 ]
The relationship between position and velocity algorithm is,
L\ Y
n
=Y
n
- Y
n
- 1
I.e
(8)
(9)
with slide wire arrangement as shown in Figure 3.22 can be used. The same
function can be performed by an integrating amplifier.
The Eq. (2) of position algorithm derived earlier is,
The change in valve position L\ Y
n
at t
n
will be,
i=o
KD . 1 (10)
=Kp . (en - en - 1) + L\ t (en - 2 en - 1 + en - 2 ) + KI en . .6. t
The integral term and derivative term can be modified by using
Trapezoidal rule and Interpolation technique similar to position algorithm.
I Integral overshoot
We shall discuss these and modify the algorithm accurately.
Logically also, the present valve position is equal to original position plus
sum of all the changes occurred so far.
The velocity algorithm at equation (13) exhibits two measure problems:
Controller drift
Controller drift
The velocity algorithm should always include integral term, otherwise it
will give rise to controller drift. To explain this let us substitute the following
equation in velocity algorithm.
(11)
ei+ei-l 1
J
Integral term = KI.Li 2 - .Li 2 .6. t
t=o i=o
_ [ en + en - 1 ]
- KI 2 .6. t
Derivative term
= [(en + 3en - 1 - 3en - 2 - en - 3 ) - ( en - 1 + 3e
n
- 2 - 3e
n
- 3 - en - 4 ) ]
where S =set point and
V
n
=value of controlled variable at t
n
By substituting modified integral and differential terms in (10), we get,
KD
= 6 .6. t [ en + 2en - 1 - 6en - 2 + 2en - 3 + en - 4 ]
(12)
computer as a Controller
3.39
The' controller .response may thus oscillate. The oscillation problem
explained above can be solved by disregarding the sign of proportional term and
assigning if the same sign as integral term. Thus,
From the above, it is clear that only integral term has set-point and thus
this term will force controlled variable to come to set-point. If integral term is
not present in velocity algorithm, then controller drift may be caused. The
proportional term of velocity algorithm may give rise to oscillations. Let us
consider the velocity algorithm with only proportional and integral term.
Proportional term (sign of integral term), K
p
(en - en 1)'
Vminbla
Value
+-----
Integral overshoot
This modification while solving one problem, creates another problem of
integral overshoot and consequently integral oscillations. When proportional
term is forced to have the same sign as integral term, the value of controlled
variable will reach the set-point at a faster rate and overshoot it.
11 t [ en + en - 1 ]
The integral term -.-
. . , KI 2
will register the change in the direction of error and oppose this along with
proportional term and give negative correction. The value of controlled variable
may overshoot the set-point in the opposite direction, giving rise to oscillation
(Fig : ~ 2 8
Output =+ Ve.
Proportional term = K
p
( en ~ en _ 1 ) = + Ve
11 t ~ en + en - 1 ]
Integral term = KI l 2 . =+Ve
When error IS increasing i.e when value of controlled variable is movmg
away from the set-point.
When error IS decreasing because of control action, i.e when value of
controlled variable is moving towards set-point.
Case II
CaseL
If integral term is not over balancing the proportional term then, output
11 Y
n
=negative.
TIme
Fig. 3.23 Case of Oscillation in velocity algorithm
The ideal solution to this problem will be to fix a band across the set-point.
When value of controlled variable is outside the band the proportional term
takes the sign of integral term. Thus controlled variable will reach the band
steeply. Whereas when the value of controlled variable is inside the band, the
proportional term takes the sign as calculated, i.e (en - en _ 1)' This gives
damping action inside the band while controlled variable reaches the set-point,
thus preventing integral" vershoot and oscillation (Fig. 3.24).
11 t [ en + en - 1 ] .
=- =+ Ve
KI 2
Integral term
Proportional term =K
p
[ en - en - 1 J=~ Ve
The negative correction may increase the error thus, giving positive
correction as per Case 1. The positive correction may decrease the error and
give negative correction.
It has been found that 7% full scale setpoint band gives good results on
simulation tests of systems with first and second order time constants. The
above modification to the velocity algorithm will yield the following equation
SP
Variable
Value
7%,+- _
lime
b. t [ en + en - 1] KD .
b.Yn =K1. 2+6b.t( en+2en-1-6en-2+2en-s+en-4)+
Proportional term
(set-point Band) SB = 0;07 of full scale
When I en - 1 I > SB
Fig. 3.24 Effect of set-point band in velocity algorithm
Readcontrol algorithm constants, steady-state algorithm
output to valve, samplingperiodandset-point
Give D/Acommand to set valvefor steady-state operation
No
Yes
GiveAIDcommand to read measured variable
Current error =set point- measured variable
Presentcontrolleroutput = controlleroutput one sample
periodbeforepresent+ (P +1+0) correctiontenns
Yn=Yn+ KP(en-9"..1) + 2
AKl t
(en KD
6At
(en+2e"..1-6en4 +28n-a +61>-4)
Set past error =presenterror
Give O/Acommand to forward controlleroutput to
Fig. 3.25 Program flow chart for velocity "'gorithm
Proportional term = (Sign of integral term) K
p
I en - en - 1 I
When len - 1 I < SB
Proportional term =KP I en - en -1 I
The flowchart for velocity algorithm is shown in Figure. 3.25. The ADC
output can be directly used for calculation if set-point and SB are expressed in
terms of fraction of ADC output values.
The range of ADC output is taken as full scale value. This avoids time
consuming portion of conversion to Engineering Unit III software. If
proportionality constant KP integral constant (1/ K1) and derivative constant
KD can be expressed as fraction of ADC values, then in addition it will save
computer time in calculation.
3.4.3.3 Position Vs. Velocity Algorithm
Reference position
The major advantage of position algorithm is the reference position. In its
equation itself, the reference position of control valve is maintained as Yo' This
is however not there in velocity algorithm. Thus whenever there is disruption
due to shut down, communication error/failure or any other reason, the median
position of valve will be known in position algorithm and eventually valve will
catch up without any synchronization problem. In velocity algorithm, external
device like stepper motor or integrating amplifier should store the last position
of valve.
Reset windup
When the DDC control loop IS being tested, then sensor inputs are
connected to see whether the DDC accepts the signals. But control valves in
3.42 computer Control of Process
3.43
Common memory
Bus
Microprocessor Microprocessor Microprocessor Microprocessor
+ + + +
Local memory Local memory Local memory Local memory
--------- -------- -------- --------
MUX+ADC MUX+ADC MUX+ADC MUX+ADC
[]} - - - Transmitter
3.5 ARTIFICIAL INTELLIGENCE AND EXPERT SYSTEMS
have only to specify the pertinent input variables which establish the setting
of each final output element. A programmer could easily put these variables
together (as described above) to get the desired control action.
Fig. 3.26 Multimicrocomputer based DDC structure
Introduction
Process control and monitoring is becoming increasingly automated. In
today's environment large process control systems and distributed control systems
are being automated using computers. Such automation is absolutely essential for
fast and complex processes like power generation, transmission and distribution
systems, petrochemical processing, metal industries, discrete manufacturing, etc.
These manufacturing and process environments are exceedingly complex with
volumes of data and more opportunities to go wrong than to go right. The
management of complex processes is a difficult task requiring a wide range of
knowledge and expertise. The challenges in building intelligent control applications
Jar process control and management problems are vast and require the subtle
combination of many technologies and methodologies.
In intelligent control, the contrci algorithms are developed by emulating
certain characteristics intelligent biological systems. An intelligent control
technique that holds tremendous promise combines the methodologies of
artificial intelligence (AI).
Thus velocity algorithm with corrections incorporated IS best suited lor
DDC control loops.
A possible solution to above problem may be to disregard integral term
in test mode. However, then integral mode,cannot be tested along with other
modes by connecting the system inputs.
process are not connected, so that process is not disrupted. In position algorithm,
the integral term increases to its limit (saturate or windup) because when the
process is not connected, process reactions are same, i.e. sensors read more or
less the same value and for computer error is persisting. In this case when the
DDC system is finally connected, the correction drives control valve to an
extreme, causing process disruption. This problem does not arise in velocity
algorithm where the integral term is given by
3.4.4 Multimicrocontroller based DOC
This loop description can be a great help to the instrumcntation-d.o.ign
engineers, if each kind of loop component is tabulated separately according to
the plant instrumentation numbering system. The instrument enginfeT would
A parallel microcomputer based DDC ar chitecture, suited for parallel
multi-variable cascade loops is shown in Figure 3.26. The common memory
stores all the variables and is used for transferring parameters between various
systems.
Auto manual switching
The DDC control loop with position algorithm causes bumps :' Lthe time
of manual to auto switching. At the time of switching, the algorithm must be
initiated to zero error situation. Thus output must be equal to Yo. i.e median
position of valve. Depending on the process dynamics, operator selects the
median position of the valve and corresponding Yo value. The DDC algorithm
has no knowledge of this precise Yo value, and thus switching causes bumps.
. Velocity algorithm is not susceptible to this problem since Yo does not exist in
the algorithm.
.. Artificial neutral networks (ANN)
Expert systems
The inference engine
The knowledge base
The interface between the human and the expert system.
The inference engine gathers the information needed from the knowledge
base, from associated databases, or from the user. It guides the search process
in accordance with a programmed strategy; uses rules of logic to draw inferences
or conclusions for the process involved; and presents the conclusions (where
warranted) with explanations or bases for the conclusions.
The knowledge base consists of information stored in retrievable form in
the computer, usually in the form of rules or frames. the correctness and
3.5.1.1 Expert systems
Expert systems are computer software systems that. mimic the tasks
routinely carried out by experts. Expert systems mimic human judgement using
computer programs by applying substantial knowledge in specific areas or
expertise to solve finite, well-defined problems. These computer programs
contain human expertise (called heuristic knowledge) obtained either directly
from human experts or indirectly from books, publications, codes, standards or
databases as well as general and specialised knowledge that pertains to the
specific situation. Expert systems have the ability to reason using formal logic,
to seek information from variety of sources including databases and the user,
plant operator and to interact with conventional programs to carry out a variety
of tasks including sophisticated computations.
Usually, the domain of expertise of these systems is restricted to a
well-defined area. Unlike humans, these systems lack common sense n ~ hence
cannot extrapolate their knowledge much beyond their narrow domain of
expertise. They encode the relevant knowledge about the problem domain in a
form, which is suitable for manipulation. The structure of the expert system
separates the knowledge froni the mechanism for manipulating the knowledge.
This has significant advantages because it makes the system easier to maintain
and update. It is this separation of the knowledge base from the reasoning.
process that makes expert systems different from other software package.
Figure 3.27 shows the typical structure of an expert system.
The principal components of an expert system are:
3.44
3.5.1 Artificial Intelligence
" Natural language processing
Artificial Intelligence (AI) IS a field of computer science concerned with
the development and deployment of machines that mimic human behaviour. It
attempts to emulate certain mental process of human by using computer models.
AI can be viewed as a collection of tools, concepts, techniques and methodologies
that allow certain problems to be addressed that cannot easily be modelled in
a closed-form analytic solution. It can also be viewed as a way of increasing the
design centre of an existing problem where solution is not really satisfactory.
Artificial intelligence can be divided into the following major categories:
Many techniques are used to make the building of intelligent reasoning
systems feasible. in realistic time frames with proper software engineering
practices. AI is not a magic, and it is these techniques along with software
environments that translate what seems to be magic into actual step-and-repeat
methodological approaches.
There has been some exploratory work on the role of AI in process control.
Several tool kits have been developed to assist in this process. As an emerging
technology, it was experimented with AI to understand how it could help and
support process control. However, the?e experiments have largely indicated that,
by and large the techniques were not as effective as existing PID controllers.
Based on these, it was believed that AI should only be used in monitoring of
process control systems, and also in assisting working as quickly as possible.
Although, AI provides many powerful means for data abstraction and easy
modification of instructions, they still do not address the depth of control needed.
Fuzzy logic
3.47
User
Real - time expert system '
engine
interface
I
Knowledge base Database
ineer Messages to
Communication
operator
links
I Operator 1
I
Distributed control system
interface I

Sensor Control
data actions
I
Process
Experts need not be present for a consultation, it may be delivered to
remote locations where expertise may not be available.
Expert systems do not suffer from some of the' short comings of human
beings, e.g. tiredness or carelessness as the workload increases.
The techniques inherent in the methodology of expert systems
minimize the recollection of information by requesting only relevant
data from the user or appropriate database.
Expert knowledge is saved and readily available because the expert
system can become a repository for undocumented knowledge that
might otherwise be lost.
Fig. 3.29 Structure of areal-time expert system

Operator
-,...__.-...; --'-__
Process eng
A number of characteristics of expert systems are unique and generally
advantageous. Some of the advantages of implementing an expert system in
computer-aided process control are:
Expert system applications to process control can be divided into two
areas: control system design and real-time process control. In an expert system
for design, the goal is to capture the expertise of the design engineer to reduce
the design time and to produce an optional design. The user or operator will
interact with the program in an interview mode, supplying relevant data on the
specific problem and the expert system will suggest an appropriate control
configuration. In the real-time process control, the expert systems are being
applied for batch process control, distributed control, fault diagnosis, and
statistical process control.
Output, c
Output results
Explanation of results
Queries
Input data
Fig. 3.27 Structure of expert system
Expert controller
Expert system shells are currently available in the market, tailored to
meet the requirements of process control applications. It requires close
integration with the underlying distributed control system (DCS) for exchange
of sensor data and communication with the plant operator. Figure a.ao shows
the structure of a real-time expert system. Examples of real-time expert system
shells include G2 (from Gensysm Corporation), and Expert 90 (from Bailey
Controls),
Fig. 3.28 Expert control system (ECS)
completeness of the information within the knowledge base is the key obtaining
correct results or solutions using expert systems.
3.46
The interface between the human and the expert system must translate
user input into the computer language, and it should present conclusions and
. explanations to the user in a written or graphical form. It should also include
an editor to assist in adding to or changing the knowledge base.
Figure a.28 illustrates the expert control system (ECS). Real-time expert
systems are being used for process control applications. It requires additional
features such as the ability to include the time element in the reasoning process.
ANNs are large-scale parallel distributed processing, non linear dynamic
systems. ANNs also exhibit a surprising number of the human brain
characteristics. Neutral networks are built of many simple nodes, called neurons,
which are distributed. These nodes are arranged in layers or and often
connected to many nodes in other layers. A 'layer' is set of nodes weights
are actively manipulated, and serve as a buffer between input or output or
other layers. A slab is a set of nodes that may be different in terms of internal
specifications or connectivity but which share the same layer. A single layer
may consist of multiple slabs. Each node processes the input it receives via
3.49
Nuetrondendrites

The learning rules used to train the networks can generally be classified
as supervised or unsupervised. Supervised learning rules require the desired
network output to be available and they adapt the weights so that the output
error is reduced. On supervised learning is used to organize the network's
structure based only on the training inputs presented to the network. ANNs
learn from experience, generalize from previous example to new ones and
abstract essential characteristics from input containing irrelevant data. ANNs
have other features such as high dimensionality, adaptability, self-organization,
and variety of connections among neutrals. Therefore, ANNscan use
nonlinearity, learning and generalization capabilities for applications .to
advanced process control. Figure 3.31 shows a three-layer fully connected
neutral network.
Fig. 3.30 Neural net analog with human brain
Synapses Axon
Good fit for non linear models.
these connections and provides a continuous analog value to other processing
elements via its outgoing connections. As in biological systems, the strengths of
these connections can charge. It is important to know in a neutral network that
how the processing elements are connected, how the processing elements process
their information, and how the connection strengths are modified. Figure 3.30
shows the neutral net analogy with the brain.
Advantages of neutral network include:
Ability to adapt, generalize and extrapolate results.
Speed of execution in recall mode.
computer Control of
Expert systems have following disadvantages also that affect their use:
They must be kept up to date as the condition changes.
Results are very dependent on the correctness of the knowledge
incorporated into expert systems.
They do not benefit from experience except through updating of the
knowledge base (based on human experience).
Expert systems are unable to solve problems outside their domain of
expertise and in many cases are unable to detect the . limitations of
. their domain.
3.48
3.5.1.2 Neutral networks
Neural networks, like expert systems, robotics, natural language
processing, machine vision and fuzzy logic, are parts of an overall technology
umbrella of artificial intelligence (Al). In process control applications, the
difficulties that arise in the control of complex systems, are mainly due to system
complexity, and the presence of non linearities and uncertainties. Such systems
are characterized by poor models, high dimensionality of the decision space,
distributed sensors and decision makers, high noise levels, multiple sub-systems,
levels, time-scales and performance criteria, complex information patterns,
overwhelming amounts of data, and stringent performance requirements. Even
modern theory is not in a position to cope with these situations. To
address these problems in a systematic way, a number of methods have been
proposed that are known as an intelligent control theories or methodologies.
Amongst these methods, perhaps the most attractive approach is that of
artificial neutral networks (ANNs).
e(t)

I
I
I . Neural network prior
information
Fig. 3.33 Cascade neutral network controller
Neural network
training
+
Fig.3.32 Plant modelling using neutral network
r (t)
Plant
from the present state (e.g. the difference between the and the actual
output position). The operation is such that the same deviation is applied to
the input of the NN, which in turn provides a correction to the controller signal.
The NN is trained continually online over the space of small changes as the
controller is operating normally. Gradually, the NN will take over the control
action because it makes better choices of the control signal than the controller
does.
Another configuration suggested is shown in Figure 3.34, in which the
NN is used to make the performance of the system appear linear to the
controller. For example, if the plant is a robot, which is represented by a highly
nonlinear model, then the controller would implement the computed torque
algorithm to provide the desired acceleration signal to the NN. This will, in
turn, perform the inverse dynamic evaluation and determine the force needed
to yield the desired acceleration, 'This approach is different than feed-forward
architecture, and requires the NN to be pre-trained to learn the true dynamics
Input layer
(9 nodes)
Hidden layer
(4 nodes)
Output layer
(2 nodes)
Rlifluxflow
Manipulated variables
Steam flow
Distilate Bottoms Steam Feed Top Bottom Reflux .Pressure
flow . flow flow temp temp temp temp temp
Disturbance variables 4+-- Controlled variables --+-
Fig. 3.31 Three-layer fully connected neutral network
Feed
flow
+--
Ease of maintenance.
However, neutral networks have some disadvantages, such as:

Cannot alone handle constraints.
Cannot optimize.
Needs of lots of data.
Needs lots of CPU power in training (learning) session.
Are unpredictable for utilization in untrained areas.
Figures 3.32 and 3.33 illustrate the examples of neutral networks-based
plant modelling.
A neutral networks (NN) control system configuration is shown in
Figure 3.31, in which theNN is used to emulate the plant dynamics. It is a
feed-forward architecture in which the input to the controller is a deviation
3.52 computer Control of Process Q?mputer as a Controller
3.53
+
As an example. in conventional computer systems, we express the
temperature in degree centigrade, height in feet or centimetre and so on.
However in real life we may say it is cold, not so cold or very cold or person
may be tall, very tall etc. Fuzzy logic gives mathematical representation to this
imprecise knowledge.
The linguistic variables provide link between natural language and
representations which accommodate quantification. The linguistic variable. has
valuesconsistinf of words or sentences rather than numbers. As an example, height
may assume the-value of 'very short', 'short', 'medium', 'tall', 'very tall'. These values
which are themselves linguistic variables may in turn be given meaning.
Figure 3.35 shows the membership function for the set of young persons
Fuzzy logic addresses these problems by presenting a meaning to these
imprecise informations, so that these can be used effectively by computers. Fuzzy
logic can be viewed as an extension of multivalued logic system. In two-valued
logic system, a proposition is either true or false. In multi-valued logic system
a proposition may be true or false or have an intermediate truth value which
may be an element of infinite truth valued set. In fuzzy logic, the truth values
may range over the fuzzy subsets.
An important concept in fuzzy logic is grade of membership - A number
which describes the extent to which an element is in a set. It also describes the
truth value of a particular statement. The grade of membership of fuzzy set is
defined by membership function. The membership function provides a direct
linkage between fuzzy logic and fuzzy sets. It is defined by rnA (x). Where x is
a member of A and A is some proportional or predicate class.
c(t)
c(t)
Plant
(a) Feedforwardarchitechtul'El
(b) Linearization architechture
Fig. 3.34 Neural network control system
of the plant off-line. Because the NN makes the closed-loop system appear to
be linear to the controller, this configuration is called linearization architecture.
3.5.1.3 Fuzzy logic system
Introduction
Decision making in the real world is based on mostly vague and imprecise
data. There is a famous saying that "Life is the art of drawing sufficient
conclusions from insufficient data". Although faster computers are now available
for best results, the number crunching power alone does not suffice to achieve
the effective process control, if the data itself is vague and imprecise. Some of
the important questions that have haunted control engineers are:
oL-- ..::-.
50 100
years
Fig. 3.35 Membership function for set young persons
Grade
of
membership
0.5
1. Is it necessary to know a detailed mathematical model of the system III
order to control it effectively?
2. Is it possible to make the controller learn as to how to control the plan
effoctively?
3. How does a human controller act in which circumstances where the data
is vague and incomplete?
These considerations have led to great deal of investigations.
There are some successful implementations . but difficulties have been
experienced where processes operate over wide range of conditions.
3.54
Grade
of
membership
0.5
Fig. 3.36 Complement, Complement intersection and union operations on fuzzy sets
Set-point
Sensor
signals
Defuzzylier
Plant
Actuator
signals
3.55
mA (X), mB (X) = membership function for two fuzzy sets A and B
inAC(X) =1 - mA
mA u13 (X) =mA(X) U mw:(X) =membership function of union of two set Aand B
mA n13 (X) =mA n mB'(X) = membership function of intersection of two sets A
and B
--.- -- - _
that they determine the way observations of process variables are expressed in
fuzzy sets. Appropriate membership functions are assigned to each range of
process variable. This yields the fuzzy set values. As in crisp sets, the concepts
of complements, Intersections and Unions are defined for fuzzy sets and
membership functions of fuzzy sets satisfy these operations. Figure 3.36 shows
these operations on membership functions of two sets.
We shall now discuss the concept of fuzzy controller design and its
implementation. However the theory of fuzzy sets, fuzzy logic etc. has been
purposely not included.
Ii'uzzy controller
A number of industrial processes such as blast furnace, cement kiln etc.
are difficult to control due to their non linear behaviour and poor quality of
variable measures. In some modern plants with process control computer, plant
models have been used to calculate the required controller parameters thus
automating the higher level control function. 'I'he plant models are
approximation to real process and may require large amount of computer time.
Fig. 3.37 Plant _control using fuzzy controller
An - alternative - approach to the control of complex processes is to
investigate the control strategies followed by a human operator which are based
on intuition and experience. The fuzzy controllers are a step in this direction.
The human operator can effectively control the process based on the
control strategy even with imprecise data. As an example typical rule for
temperature control task executed by human operator would be: If temperature
is high and rising then increase the cooling. The high and rising are imprecise
magnitudes for temperature and its rate of change.
The fuzzy controller is designed to deal with such situations where
available sources of information are inaccurate, subjectively interpreted or
uncertain. The block diagram of fuzzy controller is shown in 3.37.
The main constituents of fuzzy controller are:
Fuzzyfier
Knowledge base
Decision strategy
Defuzzyfior.
Fuzzyfier
In fuzzy controller, in order to include the control strategy of human
operator the sensor readings which are non-fuzzy should be converted to fuzzy
from by a process 'called fuzzification. In order to fuzzify the information, the
knowledge from experts is obtained to classify the crisp information into
3.57
n
n
I. rnA (x) Xi
i=o
Centre of gravity
2. Mean of Maximum approach in which the arithmetical mean value of
all the maximum available in the resultant membership function is
used.
3. Centre of gravity approach which calculates the point of centre of
gravity of resultant membership function curve. The centre of gravity
of a membership function MA (x) is defined by:
1. Maximum height approach in which the maximum value of resultant
membership function is used.
Thedefuzzification involves finding the exact value of conclusion (control
signal) from its membership function. Following methods are usually used:
Defuzzifer
Interference strategy involves executing the decision rules and then
drawing conclusion. In real-life examples more than one rule may be applicable,
resulting in more than one conclusion. The membership functions of antecedent
(left hand portion) and conclusions (right hand portion) of rules are prepared
in advance. The membership function of antecedent is evaluated with respect
to membership function of observed value. This if for then evaluated with respect
to membership function of conclusions.
Interference strategy
- - . - . , - ~ - _ . ~ - -_..._- . . - - - - - - ~ ~ -
r -6 -5 -4 -3 -2 -1 -0 +0 +1 +2 +3 +4 +5 +6
- ---------
..__. _.,-
I---
0 0 0 0 0 0 0 0 0 0 0,1 0.4 0,8 1.0
._..._-- --_._-- _._--1----- -- '---
0 0 0 0 0 0 0 0 0 0,2 0,7 1,0 0,7 0,2
_..,-._,--"_.- -_.__.....
1----
0 0 0 0 0 0 0 0,3 0,8 1.0 0,5 0,1 0 0
.. _........._..... _ . ~ - _ . _ - _ . _ -
----"-----1----,
0 0 0 0 0 0 0 1.0 0,6 0,1 0 0 0 0
-- "'-'-"'-'--
_. __.
0 0 0 0 0,1 0,6 1,0 0 0 0 0 0 0 0
-------- -------- - ~ --
0 0 0,1 0,5 1.0 0,8 0,3 0 0 0 0 0 0 0
_._-- -----" -----
0,2 0,7 1.0 0,7 0,2 0 0 0 0 0 0 0 0 0
.__... _-.-
------
1.0 0,8 0.4 0,1 0 0 0 0 0 0 0 0 0 0
-"--- _._-----
Fig. 3.38 Look-up table to relate fuzzy subsets to quantised error values,
NO
NS
-_..- ._-_._--
NM
NS
different levels. For each measured variable, a matrix of values is prepared in
advance. Thovaluesof variables are then classified in to different levels from
very high to very low. The membership function for each measured variable is
also drawn. rl'he sensor reading and resulting error values can be classified in
to 8 levels - "Positive Big, Positive Medium, Positive Small, Positive Nil,
Negative Nil, Negative Big, Negative Medium and Negative Small". Figure 3.38
shows a lookup table to relate fuzzy subsets to quantised error values.
Ouanflsed erro
PB
...._..- .....-_..._----
PM
._,._.._---_._... . _ ~
PS
......... _ . , , - _ . ~ - .."--'-
PO
Knowledge Base
The knowledge base contains the set of decision rules through which fuzzy
information obtained from fuzzyfier may be utilised to obtain decision. The
decision rules are developed heuristically for a particular control task and are
implemented as set of fuzzy conditional statements of the following form.
"If E isPB or PM and of GE is NS then CD is NM".
These expressions define fuzzy relation between a set os observed
imprecise values of error (E) its rate of change (CE) and the appropriate change
in control input (CU). '!'he complete control strategy is defined in the above
form of set of fuzzy rules. The above form is called "conditional composition".
Other statements of the following form are also used.
1=0
Figure 3.37 (a) shows the fuzzy rule processing using membership value
and input value. The fuzzy control loop is shown in Figure 3.39.
where n is the number of fired. The centre of gravity. method is widely used
and found very useful.
~ - - - = . = , , - , = ~ ~ -
itive composition - A IS Tall and B IS short
-_.-.._---_.__..._------ ._------
lVe composition - A IS Tall or B is short
_.-._----
nal and conjunct - lfA IS Tall
Disjunct
Conditio
THEN Y is short
Conjunc
ELSE Y might be short
3.58
IFs of rule set
Thena of ruleset
Q9mputer as a Controller
3.6.2 Select measurements
3.59
.. ---]
I 1 I: 1
1 :": 1 . :
I:: :. I I
I IF of I rule I 1 I Then ot rule I :
_j i .. !6 i
1
1 1 1 I I :
I IF ofl rule k I I Then of rule k '
I I 1 1 I I
:: :: 1 ,
"I 1 1 1 I :
" I I: I
I I 1 I I I
1 1 I 1 1 1
I I 1 I I I
1 I 1 1 I I
I I I I
I 1 1 I
----------------. I
J
rInput
Output value
Fig. 3.39 Fuzzy-rule processing
3.6 DESIGN OF COMPUTER-IZED MULTILOOP CONTROLLER
3.6.1 Control objectives: Definition
Whatever our control objectives are, we need some means to monitor the
performance of the process. This is done by measuring the values of certain
processing variables.
What variables should we measure in order to monitor the
operational performance of a plant?
We should like to monitor directly the variables that represent our control
objectives. Such measurements are called primary measurements.
If the control objectives are not measurable quantities (ie) they belong
to the class of unmeasured outputs. Such cages we must measure other
variables which can be measured easily and reliably; Such supporting
measurements are called secondary measurements.
Then we develop mathematical relationships between the unmeasured
outputs andthe secondary measurements.
Unmeasured output .= f (secondary measurements)
The third class of measurements that we can make to monitor the
behaviour of a process includes direct measurement of the 'external
disturbances.
3.6.3 Select manipulated variables
What are the manipulated variables to be used to control a process?
Usually in a process we have a number of available input, variables which
can be adjusted freely. Which ones we select to use as manipulated variables
is a crucial question, as the choice will affect the quality of control actions.
What are the operational objectives that a control system is called upon
to achieve'?
3.6.4 Select the control configuration
A control configuration is the information structure that is used to connect
the variable measurements to the available manipulated variables.
.What is the best control configuration for a given process control
situation?
Ensuring the stability of the process.

Supporting the influence of external.
Optimising the economic performance of a plant or
A combination of the above.
Depending on how many controlled outputs and manipulated inputs.

We have in a process, we can distinguish the control configurations as
either single input, single output (SISO) or MIMO control systems.
(4)
DistnJrbances
t 1
Flnal control
I:'
y,
element
Process
Y2
Multlploxor
1-----1
I I
_--'---'LLi--' P:}-Hi11-----1'--__-'
'-----' I I
0-....:..... :---
1 1 : -----'
___________________________ : d: dlscratErUme signal
Computer-process Interface c. continuoussignal
Fig. 3.39 Computerized multiloop controller
Primary
Controller
elgorithm
Typewriter
termlnal
2. Subprogramfor
loop 2
1. Subprogramfor
loop1
3. Coordination
subprogram
Compute,
Control program
Set
, Point
I
I
I
I
I
I
I
I
I
3.6.6 Multiple-loop control
A digital computer can be used to control simultaneously several outputs,
not only one as discussed above. We still need.an interface between the computer
and the process, but it is now some what different. 'rhus,

(1)
(2)
(3)
\
Gsecondary = GC . HOp,!I
For the primary loop the overall open-loop TF is
[
Gc' JI Gp,!I '. J
Gprimary=GcI 1 G JIG II GpI
, + c' p'
The stability of the secondary loop is determined by the roots of its
characteristic equation
(ii) The loop that measures disturbance uses the output of the primary
controller on its set point and is called the secondary or slave loop.
3.6.5 Design the controller
In every control configuration, the controller is the active element that
receives the information from the measurements and takes appropriate
control actions to adjust the values of the manipulated variables, based
on control law.
Design of multiloop controller:
Multiloop control systems involve
(i) More than one measurement and one manipulation (or)
(ii) One measurement and more than one manipulated variables
(eg) Cascade control
If has two control loops using two different measurements, but sharing a
common manipulated variable.
(i) . The loop that measures controlled variable is the dominant or primary
or master control loop and uses a set point supplied by the operator while
Computerized multiloop controller
The closed loop response of the primary loop is influenced by thedynamics
of the secondary loop, whose open-loop transfer function is equal to
The characteristic equation whose roots determine the stability of the
primary loop is,
Fig. 3.40 Digital computer used in two distinct control loops
3.62
computer as a Controller 3.63
Problem solving
known algorithms
Conventional EPD systems AI=__ =----j1
.. --- in the field of Problem
solves
usmg well problem solving by
-reasomng
4. State the requirement of CIMPLICITY HMI Plant Edition?
CTMPLICITY HMI Plant Edition provides an extraordinary selection of
features that. enables to configure comprehensive and robust project. It a
quick tour provides links to the related subject in the documentation.
5. What is a DDC?
The nDC (Direct Digital Control) directly interface to the process for data
acquisition and control purposes.
6. List the tasks performed by IlP in DDC?
It reads the various process variables from different transmitters
through multiplexer and ADC
It determines the error each control loop and executes control
strategy for each loop.
It outputs the correction value of control value through DAC.
7. Mention the two algrothims used in DDC software?
There are two algorithms for programming a three mode PID control loop
Position algorithm
Velocity algorithm
8.. Define AI?
Artificial intelligence has been defined as the branch of computer science
which deals with the.' software and hardware techniques to solve symbolic
problems as against 'number crunching' problems solved by EDP machines.
9. What is heuristics?
The learned facts, experience and institution are three basic inputs which
enable decision making collectively these arc as heuristics. In AI systems
the decisions are taken on the basis of heuristics.
10. Detail on the difference in working of conventional EDP systems
and AI systems.
ControlledVariable
Control Equation
Measuring Varible
'------------1Meesuring Element 14'-_-'--_---.,. _
TWO MARKS QUESTIONS AND ANSWERS
1. Instead of using one AID convertor for every measured variable, we
employ a single AID converter which server all measured variables
sequentially through a multiplexes.
2. A multiplexer can also be used to obtain several outputs from a single
D/A converter.
8.. The control program is now composed of several sub programs. Each
one used to control a different loop. Furthermore, the control program
should be able to coordinate the execution of the various subprograms
so that each loop and all together function properly.
Fig.
1. Draw the general block representation of a computer control
system?
2. Define SCADA?
Supervisory Control And Data Acquisition (SCADA) system is computerised
hardware and software based project that provides a single integrated view of
all control and information resources, enable engineers, supervisors, managers,
operators to view and interact with the working of an cntiro operation through
graphical representation of their product process.
3. What is CIMPLICITY?
CIMPLICIT'Y is a SCADA package. It is based on a client-server
architecture consisting of servers and viewers. Servers are responsible for the
collection and distribution of data. Viewers connect into servers and have full
access to the collected data for viewing andcontrol actions.
3.64 computer Control of Process
11. Mention the various categories under the AI system.
Al system can be divided into three major categories
Natural language system
Perception system for vision, speech and touch
Expert or knowledge based systems
12. List the various types of information presented, to expert system.
The information presented to expert system may be factual, incomplete,
judgmental, speculative, experimental, uncertain, fuzzy, intuitive. Expert system
should derive conclusions based on such incomplete information, just like real
life human expert.
3.65

omputer as a Controller
Frames
Frames are templates or modules for holding clusters of related knowledge
about a particular narrow subjects.
(a) If { this happens} then do { that} or
(b) If { this is the case } then { that is true } or
(c) If { this is the case} then do { that}.
15. List down the methodologies used in knowledge representation.
The knowledge representation methodologies include
Semantic networks
In semantic network nodes and arcs form graphical notation to represent
objects (achons, events) and relationship among them.
Rules
Rules may be in one of the three forms.
AI systems
-search
-heuriatics
Solutions in the form of
-opinions
-conclusions
-logic solutions
Newknowledge
specific numerical-solutions
new data (result)
I relyon
Reasoning

Learnedfacts Experience
13. ' What is a domain knowledge?
Tho'domain knowledge is stored III the knowledge base of system..., The"
domain knowledge contains:
- fads
14. Draw the flow diagram of Human Decision making Process?
HumanDecisions
Ireiyon
judgement
16. Differentiate the two reasoning mechanisms.
The two main reasoning mechanisms include
Backward chaining ---7 stars with a goal (i.e.) conclusion which is most portable
and then tries to go backward and match the left hand side of rule i.e. conditions.
Forward chaining ---7 is a top down approach. It proceeds in the direction
until the goal, i.e. conclusion is reached.
17. List the goals for an ideal expert controller.
An ideal expert controller should be able to satisfy the following goals:
1. 'Ability to control a large class of processes which may be time varying,
non linear, with variety of disturbance.
2. Requirement of minimum prior knowledge about the process.
3. Ability to improve its performance with time, as it acquires more knowledge.
4. Ability to monitor the performance of the system and detect problems
with sensors, achitors, and other components.
, \
- Relation between facts
- Heuristics
3.66
4.1 EVOLUTION OFPLC
CHAPTER - 4
Programmable Logic Controller
The PLC was originally designed and developed by a group of engineers
of general .motors corporation in 1968 to eliminate costly scrapping of
assembly-line-relays during model change overs of cars. These PLC's had to be
easily programmed and reprogrammed, preferably in plant, easily maintained
and repaired smaller than its relay equivalent, and cost-competitive with the
solid-state and relay panels then in use. This provoked great interest from
engineers of all disciplines using the PLC for industrial control. A microprocessor
based PLO was introduced in 1977 by Allen-Bradley corporation in the USA,
using an Intel 8080 microprocessor with circuitary to handle bit logic
instructions at high speed.
The early PLC's were designed only for logic based sequencing operations
(ON/OFF signals). Today there are hundreds of different PLC models in the
market. They differ in their memory size (from 256 bytes to several kilobytes
and megabytes) and input / output capacity (from few lines to thousands). The
difference also lies in the features they offer. The smallest PLC's serve just as
relay replacers with added timer and counter capabilities. This is an extremely
basic controller that is amazingly in expensive. This small dedicated controller
is enclosed in a single mounted hardened case. It provides reliable control to a
stand alone section of a process.
The modern medium-sized PLC's perform all the relay replacement
functions expected of it but also adds many other functions such as counting,
timing and complex mathematical applications, to its repertoire. Most
medium-sized PLCs perform P ~ feed forward and other control functions as
well. In addition medium-sized and largescale PLCs now have data high way
capabilities and they function well in distributed control system (DCS)
. environment.
Output
Plant
MUX
REVIEW QUESTIONS
1. List the functional requirements of a distributed process control system.
2. Sketch the structure of a Direct Digital Control system and explain the
functions of each block.
8. Enumerate the advantages of digital computer control of processes.
4. Write short notes on AI and expert system.
5. Write short notes of the following: (i) SCADA; (ii) Direct Digital Control
6. Write short notes on: (i) AI control system; (ii) Expert control system
20. Draw the structure of DDe?
18. Differentiate between analog controller and digital controller.
Analog controller though are faster than digital controllers, the later IS
preferred because the changes in the parameter values are possible -in digital
controller while not in analog controller.
19. Draw the block representation of expert controller?

Comp PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Comp PDF

Uploaded by

Copyright:

Available Formats

CHAPTER - 3

Fig. 3.29 Structure of areal-time expert system

However, neutral networks have some disadvantages, such as:

Ensuring the stability of the process.

Depending on how many controlled outputs and manipulated inputs.

You might also like