MINE 432
Robotics and Industrial Automation
Lecture 14
Real-Time Supervisory Control
Definition
A real-time control system is an integrated computer system that responds to interrupting external
events to provide accurate, fast control or alarm action. Typical response times are between ~ 10
msec to 2 - 3 seconds depending on the purpose and application.
Intelligent Control is computer modeling techniques that employ elements that mimic the humanthought-
process. Response times depend on the purpose and application:
Direct real-time: 10 msec to 2-3 seconds
Pseudo real-time: 2-3 seconds to 1-2 minutes
Supervisory Control is often run in "pseudo" real-time" since the module used is usually adjusting
set-points for local control loops that are operating in "true" real-time.
In Intelligent Supervisory Control, as we have seen in the earlier chapters, the elements that are
used might consist of one or more of the following items:
l rule-based modeling (expert systems)
l fuzzy logic inferencing
l artificial neural network modeling
l genetic algorithm optimization
l ability to explain and justify
l ability to adapt or learn from experience
l management of temporal-reasoning
l
A typical rule in a Supervisory System using a rule-based approach is as follows:
Rule Name: water_valve_high
IF tank level is definitely “high“
AND pump speed is “maximum“
THEN valve position change is "closed alot"
DEFUZZIFY (valve position)
FIND (pulp flowrate * )
WAIT (“water_valve_high“, 120 )
ELSE valve position change is not "closed alot"
Combining these rules with the power of Fuzzy Inferencing using fuzzy sets such as:
Fig. 14.1. Typical Fuzzy Sets covering the full universe of discourse of a variable.
provides a way to develop Fuzzy Associative Maps of two or more input variables into a third
output variable so that rules are applied partially depending on the degree of belief in the concept
in question.
Intelligent User Interfaces
In addition to the Inference Engine and Knowledge Base, a good real-time supervisory system
will have the capability to locate "intelligence" directly within the User Interface:
Some of the items that might be found in these interfaces include:
l Process mimic diagrams with various I/O icons such as slider bars, dials, charts, etc.
l Trend diagrams of data
l Window for viewing and logging messages from the expert system
l Explanation and Justification capabilities for novice operators
l Messages can be filtered into classes depending on the needs of the workstation
Integration of Components into a Real-Time Supervisory System
User Interfaces
Multiple
Inference
Engine
InterNet
Bridge
Process
Bridge
Blackboard
Knowledge
Base
Artificial
Neural Network
Genetic
Algorithm
Fig. 14.2. Components that make up a Real Time Supervisory Control System.
There is now a trend toward the integration of intelligent supervisory modules with existing
SCADA software packages such as Intellution's The FIX, Factory-Link, Real-Flex and others.
The interface in these systems will look the same as a normal system without intelligence with the
exception of a window for displaying messages from the Expert System. These messages might
include alarm situations, process state analysis messages, explanations for control actions taken,
scheduled events, etc.
Comdale Technologies' SmartWorX Suite is an example of this new approach to Intelligent Real-
Time Supervisory Control.
Fig. 14.2 shows how the modules using the various components of CI can be integrated with
other software: user interfaces, SCADA tools, network drivers, etc. Future systems will also
allow connections to the Internet opening the plant to control from very remote locations.
Operating Systems
Two operating system alternatives are useful to analyse.
l UNIX-based
l Windows-NT
First, UNIX-based systems can be selected but with the exception of QNX, this will mean using
UNIX-based hardware systems such as SUN or IRIS (expensive hardware). QNX, on the other
hand, is an operating system specifically designed for use in process control applications in a PC
environment. UNIX-based as well, this O/S is designed for very rapid message-transfer and is the
state-of-the-art for numerous real-time SCADA applications.
However, Windows-NT is poised to become the dominant OS in the near future and for easy
data-transfer and compatibility between various software modules, WinNT is likely to be the
preferred choice in the future especially since hardware continues to become faster and faster.
WinNT will likely never be as fast as QNX but within 2 years, hardware will make it as fast as
QNX today. Currently WinNT is about 4 times as slow as QNX.
Intelligent Manufacturing Systems
Another important area to note for future expansion of CI and AI technologies is in the area of
plant-wide manufacturing systems. One very important approach is the development of agent
software which can be used to model a complex set of interacting processes by dividing-up the
process into "holons". A "holon" is a word derived from Holography meaning the "parts that
make up the whole". Holons can be combined together to make other larger holons. The
interaction between holons is a complex optimization and scheduling problem that demands the
use of "heuristic" methods such as are available in AI and CI techniques.
The key issues here are:
l Agent-based modeling technologies
l Elements of a complex process are considered to be holons
l A holonic is simply an individual discrete element of a whole
l Modeling methodology can be directly applied to control and operate a plant
l Heuristic methods can be very helpful in providing "optimum" solutions
The diagram below shows the NASREM modeling environment being proposed to set up
interacting models of very complex tasks. The different levels in this environment are analogous
to the control level hierarchy shown later with the bottom level used for direct servo-control and
the top level dedicated to long-term analysis across the corporation.
Intelligent Manufacturing System Modeling
maps
object lists
state
variables
objective
functions
program
files
User
Interfaces
M1
M2
M3
M4
M5
S1
S2
S3
S4
S5
E1
E2
E3
E4
E5 servo
dynamic
path
planning
task
control
scheduling
operational
actions
Sensor
Processing
World
Modeling
Task
Decomposition
detect
and
integrate
model
evaluation
plan
and
execute
Time
Scale
milliseconds
seconds
minutes
hours
days to years
operations
Multiple
Fig. 14.3. NASA/NIST Standard Reference Modeling Environment System.
Other Issues
The following issues must be considered in developing an Intelligent Supervisory System:
l time to develop a system
l system maintenance
l networking capabilities
l system security
l intelligent alarming
- eliminate alarm ‘showers‘
- persistence, inhibition & nuisance
- institute an immediate action on entering or exiting an alarm state
A good tool should be capable of providing incremental development with rapid prototyping.
SmartWorX for example, has a module called Expert Optimizer which assists a developer in
translating the operating "facts" and "rules" into the code required by the system.
System maintenance is an important point since these systems are never finished -- they continue
to grow and expand as new information comes to the attention of the developers or as the plant
and ore situations change. Maintenance should be easily done by other personnel than the original
developers without requiring extensive training.
A good tool will also have excellent networking capability to allow the incorporation of
redundancy in the network. Mirrored modules can be made available on the network so that loss
of one work-station can be handled by the software. The back-up modules can be used instead to
continue control of the process.
Security is an important feature of an intelligent supervision. Security must deal with on- and offline
development issues as well as access to output screens and input set-points. Data can be seen
by multiple users but only certain variables can be changes. Passwords can be used to establish
priority levels for multiple users to protect the integrity of the overall system.
Finally intelligent alarm systems are important for all plants. The context by which a plant is
currently running needs to be known to provide correct advice and warnings. A plant may be
starting-up, shutting-down, coming-out of an alarm situation, etc. Each of these contexts must be
taken into account in determining how alarms are passed to the User Interface.
"Smart" Alarm systems can deal with alarm "showers" wherein a large influx of alarm signals is
really not very useful -- only one underlying alarm is really important. These nuisance alarms need
to be filtered out of the User Interface. Persistent alarms are also important issues to deal with.
The time interval between consecutive messages can help but more can be done such as
troubleshooting through a well-developed knowledge base to source-out the cause of the
persistence. Actions ca be programmed to take place upon either entering or exiting an alarm
state. Intelligence can be incorporated into an overall alarm strategy to improve the environment
for the plant operators as they work with the system.
Temporal Reasoning
A good development tool must be able to reason about time. Functions need to be available to
determine certain trends in data -- these include rates of change and time-averaging.
Scheduling of events are also important -- these include setting up actions to occur at some
particular time, some particular time interval, after or before a particular occurrence, etc.
The ability to recognize certain data patterns or shapes can be very important in diagnosing
certain problems. These might include concave-up or down, sudden drops or rises in data, spikeevents
(either up or down), multiple sequential changes, etc.
High Speed Intelligent Monitoring
Intelligent real-time systems must make high-level decisions and diagnose unexpected events.
They acquire data automatically, apply heuristic methods to interpret sensor readings and feed
advice out to the process or up to the user via a friendly man-machine interface. Although, data
are acquired directly, decision-making can be slow for effective action. Data must be filtered
before passing to the knowledge base to ensure efficient processing.
In analyzing control of an industrial process, a hierarchy can be delineated (see below):
Level 0 - Process Instrumentation,
Level 1 - Direct control,
Level 2 - Supervisory control,
Level 3 - Plant-Wide control
Level 4 - Enterprise
ENTERPRISE
PLANT WIDE
SUPERVISORY CONTROL
DIRECT CONTROL
PROCESS INSTRUMENTATION
PROCESS
Fig. 14.4. Control System Hierarchy.
At the lowest level, instruments sense, monitor and manipulate process variables. Devices are
connected to units such as single-loop controllers, PLCs or DCSs, which apply a combination of
sequential or continuous-time logic. Collecting, presenting, and managing sensor data requires
numerical methods. For a system to respond to external events fast, carrying out several activities
simultaneously, the system must operate within a multitasking environment that can efficiently
handle priority interrupts and multiple module actions.
Neither of the two upper control levels need to operate in real- or "pseudo" real-time but there is
much interest today, in applying Artificial Intelligence (AI) in process control. There are many
examples of non-real-time AI applications at level 3 and 4 for diagnosis and advice [1,2,3], but
most interest in real-time AI is at level 2. The question addressed here is:
"Can AI techniques be introduced into level 1 to provide support
for symbolic supervisory control?"
Computational Intelligence
AI is the branch of computer science dealing with symbolic, non-algorithmic problem-solving.
"von-Neumann" - type computers are designed for fast, accurate number-crunching. So how can
AI move into an intensive numerical computing environment?
The answer is found in a newly evolving paradigm -
Computational Intelligence (CI).
Coined by Bezdek in 1993, CI consists of "low level” knowledge in the style of the mind [4]. CI
consists of "primitive" concepts, in the AI sense, supporting the beginning of symbolic knowledge.
These "elements" are inputs to an AI structure that processes the symbols heuristically.
"Primitives" are basic numerical operations: addition, subtraction, multiplication, division, and
comparison that make up any complex structure to output "elements". Current hardware handles
such structures efficiently and so, CI can provide support for AI methods. CI comprises Fuzzy
Logic, Artificial Neural Networks and Genetic Algorithms.
This definition seems limiting, as CI should not rely only on pure arithmetic. Rules of thumb can
aid the search for symbolic input for the high-intelligence level. This can increase performance
speed, but usually creates error in the output. To bring AI into the lowest levels of the control
hierarchy, CI modules must create "elements" very fast. By introducing AI approaches into a CI
module, we cause certain error, but gain on speed. The key trade-off in real-time is always:
accuracy versus processing speed.
Considering that error derives from using heuristics in CI, it is proposed that feedback from AI to
CI can detect such error and provide interpretation. AI can test symbolic output from cooperating
sensors and recognize, tune out or reduce the error (see below).
Artificial Intelligence
Symbolic Intelligence Error Adaptation
Computational Intelligence
Process Instrumentation
Process
Fig. 14.5, Error Detection in AI module to correct the CI Module.
To assist AI in making rapid decisions intelligently, CI needs the following items:
• IF/THEN rules (inferences and relationships)
• prior knowledge (to direct the CI process)
• symbolic "elements" (output from CI module)
The approach resembles the hierarchy of human intelligence depicted below:
.
Fuzzy Logic
Genetic Algorithms
Neural Networks
Biological Intelligence
Symbolic Intelligence
Computational Intelligence
Biochemical Processing
Artificial Intelligence
Symbolic Intelligence
FL
GA
NN
Electronic Processing
Biological versus Machine Intelligence.
Fig. 14.6. Comparison of Biological Process of Information with Electronic Processing.
Biological Intelligence consists of manipulating symbols supported by low-level numerical
processing to generate belief in a particular symbol. This organization is mirrored in the
arrangement of AI with CI to form the basis for rapid problem-analysis. Like humans, computers
work better when tasks are divided between symbolic and numerical analysis.
Within Biological Intelligence, the ability of autistic savants to carry out rapid and accurate data
calculations, musical recall, etc., are examples of how the human brain can perform unusually
accurate real-time computation. Whether output is intelligent or not is determined by those who
interact with such exceptional people. Perhaps they use fuzzy sets with very broad support
characteristics (see below). While savants may be able to tell the day of the week for a particular
date, they rarely understand the significance of the date in question.
Degree
of Belief
Normal Fuzzy Sets "Autistic" Fuzzy Sets
Low Medium High
0
100
Universe of Discourse
Low
Medium
High
100
0
Universe of Discourse
Degree
of Belief
Fig. 14.7. Comparison of "Normal" fuzzy sets with "Autistic" fuzzy sets.
Application of CI in Real-Time
CI can be applied for many real-time tasks:
- high-speed data acquisition and filtering
- time series analysis and pattern recognition
- direct control and smart monitoring
- real-time supervisory control
A Real-Time Supervisory Application
An intelligent SCADA system for continuous casting of steel billets has been developed which
relies on the use of a CI module for real-time data processing.
Continuous casting, involves pouring molten steel into a water-cooled copper mould. The semisolid
strand is pulled from the mould by rotating pinch-rolls. Casting speed is linked to mould
metal level through a standard PID loop. Level changes are reflected by changes in casting speed.
Metal level varies appreciably when turbulent conditions exist from "ropy" stream conditions. The
machine oscillates to strip the solidifying shell from the mould wall. Displacement is usually
sinusoidal but when binding occurs, these signals are distorted.
The following variables are monitored:
- Metal level position and variation
- Casting speed
- Cooling water flow and temperature
- Mould temperature, displacement and loads
The sensors and devices comprise the following:
- casting speed / metal level sensors
- mould thermocouples (TC)
- mechanical sensors - LVDT
- single loop PID controllers
The objective was to create the "Intelligent" Mould to monitor and control the process of
continuous casting. ProcessVision, a SCADA development tool from Comdale Technologies, was
used to build the application. The multi-tasking nature of a real-time ProcessVision system is
depicted in below. Each module interacts with others in a true multi-tasking fashion with
appropriate interrupts and priority scheduling as required.
Process
Instrumentation
Point
Database
Database
Historical Database
Administrator Administrator
Administrator
Trend
Message
Process
View
View
Inference Expert
Engine Administrator
Alarm
Event
Scheduler
Knowledge
Base
Explainer
Engine
Fig. 14.8. Typical Configuration of ProcessVision.
High-level supervisory decision-making occurs in the inference engine, the knowledge base and
the explainer module. ProcessVision (PV) operates under QNX, a UNIX-based truly-distributed
real-time multitasking operating system running on a PC. To accomplish rapid data acquisition
within the SCADA, a Keithley-Metrabyte DAS-20 plug-in board was used.
Building the CI module
Our research group at UBC has been working to interpret patterns from sensor responses from
many field trials conducted over the past 20 years. Specific curve shapes from thermocouple time
responses (temperature peaks, drops, etc.) are related to specific billet defects. These correlations
make up the knowledge base which detects surface defects - bleeds/laps and depressions.
When a temperature drop and rise propagates down the mould, a defect is indicated. The drop to
base temperature ratio together with upset duration define the significance and extent of a defect.
These inputs are used to predict defects. A data acquisition rate of 20 Hz captures all important
features of these traces. The "pseudo" real-time Expert System (ES) is not fast enough to process
data intelligently at this rate, so application of a CI module is a necessity.
A multi-threading 'C' program was written for the CI module. The objective was to acquire data
without delay while processing data. The two main threads are distinguished in Fig. 14.8.
The first program, main task, runs in an infinitive loop to execute the following functions:
-collect sensor data and store in RAM-resident table
-receive tuning instructions from ES
-create data processing task
The processing task receives data from the main task and sequentially filters inputs from each
channel. Up to 5 functions per channel are applied:
Symbols
Data Table
Sensor Inputs
(Recorded in mV)
Data
Acquisition
Task
Data
Processing
Task
ProcessVision
Expert System
(Point dbase)
Tuning Instructions TI
TI
Instrumented
Copper Mould
Process Brain
Fig. 14.9. Structure of the "Intelligent" Mould.
average - calculate ave. over specified number of points; pass key-word-triplet to PV.
minmax - find minimum and maximum values; pass two key-word-triplets to PV.
storedata - place collected data (in volts) into a file.
compare - combine data from 2 channels to calculate negative strip time using inputs from
an LVDT and the metal level sensor, at 200 Hz.
valley - this is an example of shape recognition and feature extraction that finds a "valley"
shape in the data. The algorithm uses prior knowledge to direct the search, and a
"window" technique to locate the minimum, left and right maximums. The search
is set up by the TI signal from the AI module. The function can recognize 5
valleys in the data table and pass 20 key-word-triplets to PV (see Table below).
calibration - convert input volts to actual values.
The main task collects data while the processing task filters recorded inputs. When the processing
task finishes its routines, it “kills” itself. The main program recreates the processing task when it
completes another acquisition cycle. The multi-threading design of the driver provides continuous
data acquisition and processing of signals at up to 400 Hz.
Connection to PV uses the Comdale 3rd-Party Interface Library, communicating with the point
database using keyword triplets. The Table below shows typical TC signal output.
CI output for one sensor for one scan cycle
Keyword Triplet Value Degree of Belief
TC1.drop.@f 39.9 100
TC1.base.@f 132.7 100
TC1.span.@f 9.9 100
TC1.time.@f 8.6 _ 100 __
The system tracks 4 thermocouples (TC). The AI module uses Fuzzy Logic to interpret the
temperature drop and span interval. Together with the number of valleys from each TC, a degree
of belief (DoB) is established that a bleed/lap or depression has occurred.
Results
The "Intelligent" Mould was implemented and tested at two Canadian mini mills in November
1994 and March 1995. The system was used for real-time monitoring. For the first time, negativestrip-
time could be followed on the screen in real-time. Correlations between metal-level
fluctuations, casting speed and billet defects were clearly evident.
The system has established a clear association of ropy streams and a turbulent meniscus with online
TC responses. These conditions lead to the formation of surface defects such as depressions,
laps or bleeds. Depression defects were predicted and a high degree of mapping between
predicted and actual defects was obtained. TC responses obtained during the March 1995 plant
trial are shown below. Two temperature drops were detected for the period shown. The degrees
of belief that a defect occurred were 98 and 97 respectively.
Fig. 14.10. Thermocouple traces from a continuous casting trial at Alta Steel - temperature drop
indicate a surface defect formation (lap/bleed or depression.
The system correctly ignored other "apparent" drops. The CI module can be setup to include or
exclude temperature depressions based on the desired degree of surface defect detection.
Correlation with surface measurements was virtually perfect. Examination of the billet cast during
this period proved the predictions correct: depressions were obvious on the billet surface. A
mathematical model to predict temperature drop and duration based on depression measurements
shows high correlation with the on-line detection system. The future plan for the "Smart Mould"
is to use these defect predictions to provide an on-line quality rating for each cast billet.
To create a solid foundation for a real-time Quality Control System, we examined a number of
billets for which detailed temperature trends were recorded. Surface defects were measured
according to their distribution, position, and surface depth. These served as input to a 2-
dimensional mould heat-transfer model. The depth and distribution of billet defects were
translated into upsets in mould heat flux, assuming that the deepest depression represented about
a 70% upset in heat extraction. Output from the model was obtained for several thermocouples
located around and down the mould. Model output was plotted concurrently with real sensor data
obtained during casting, as depicted below.
A clear correlation between these trends can be seen - one for the model temperature profile
based on depression measurements and the second for the direct measurements taken from the
TC. The modeled output has the same shape as the real thermocouple-time response.
Fig. 14.11. Comparison of temperature trend modeled from surface profile
measurements to sensor-based temperature trend.
The differences in the position of peak and drop values are a result of assuming a 70% heat-flux
upset in this analysis. Subsequent follow-up showed that 80% gave a better fit. The model
assumes a constant casting speed, although speed varies appreciably during casting. Fluctuating
casting speed also causes shifting of the thermocouple data. The analysis has established clear
correlation with existing defects on the billets, sensor data, and our mould heat-flux model
providing a basis for direct measurement of defects at the time of formation.
Conclusion
A new paradigm known as Computational Intelligence is evolving in which intelligent numerical
manipulation forms the underpinning of successful real-time AI. A parallel with Biological
Intelligence is evident from man's ability to process numbers rapidly when necessary or whenever
specialized tasks are required. The CI module acquires high-frequency data rapidly using AI
techniques that allow monitoring and control of a continuous casting process. Field trials have
confirmed the method can be applied to predict bleeds/laps and depressions. Our efforts are
continuing to develop fast-responding modules to give sufficient accuracy for high-level AI
processing. Effective real-time analysis of billet quality will lead to a useful performance criterion
to give on-line analysis of cast product quality. The ultimate goal of this system is to provide online
billet quality measurement. Both surface and internal defects are included in this system and
an adaptable billet quality index is derived. Finally, with respect to real-time intelligent control, the
following points are relevant:
• knowledge-based modeling is finding increased use within industrial plants
– on-line diagnosis and training
– real-time monitoring and control
• on-line learning will become more important in future generations of these technologies
• voice recognition and voice-activated messages will become more widely used
• operator replacement must be addressed as knowledge becomes a valuable commodity
References
1. S. Kumar, J. Meech, I. Samarasekera, K. Brimacombe, "Knowledge Engineering an Expert System for Quality
Problems in Continuous Billet Casting of Steel Billets", Iron & Steelmaker, 1993, 20(9), 29-36.
2. K. Otsuka et al,, "Expert System for Blast Furnace Operation", Sumitomo Search, 1992 , 50, 43-50.
3. R. Edwards & A. Mular, "An Expert System Supervisor of a Flotation Circuit", CIM Bulletin., 1992, 69-76.
4. J.C. Bezdek, "What is Computational Intelligence?", Computational Intelligence - Imitating Life, 1994, IEEE World
Congress on CI (WCCI), 1-12.
ليست هناك تعليقات:
إرسال تعليق