U.S. patent application number 12/376760 was filed with the patent office on 2010-11-18 for process control of an industrial plant.
This patent application is currently assigned to Auckland UniServices Limited. Invention is credited to John J.J. Chen, Mark P. Taylor.
Application Number | 20100292825 12/376760 |
Document ID | / |
Family ID | 39033412 |
Filed Date | 2010-11-18 |
United States Patent
Application |
20100292825 |
Kind Code |
A1 |
Taylor; Mark P. ; et
al. |
November 18, 2010 |
PROCESS CONTROL OF AN INDUSTRIAL PLANT
Abstract
A system (10) for controlling an industrial plant (12) comprises
automatic control equipment (14) comprising a plurality of
measurement sensors (16) for sensing predetermined variables
associated with components of the industrial plant (12). The
sensors (16) generate measured data relating to operation of the
components of the industrial plant (12). A database (20) contains
operational data, including observational data, regarding the
industrial plant (12). A processor (18) is in communication with
the automatic control equipment (14) and the database (20) for
receiving the measured data from the sensors (16) of the automatic
control equipment (14) and the operational data from the database
(20). The processor (18) manipulates the measured and operational
data to provide an evolving description of a process condition of
each component over time, along with output information relating to
operational control of the industrial plant (12) and for updating
the database (20).
Inventors: |
Taylor; Mark P.; (Auckland,
NZ) ; Chen; John J.J.; (Auckland, NZ) |
Correspondence
Address: |
BENESCH, FRIEDLANDER, COPLAN & ARONOFF LLP;ATTN: IP DEPARTMENT DOCKET
CLERK
200 PUBLIC SQUARE, SUITE 2300
CLEVELAND
OH
44114-2378
US
|
Assignee: |
Auckland UniServices
Limited
Auckland
NZ
|
Family ID: |
39033412 |
Appl. No.: |
12/376760 |
Filed: |
August 6, 2007 |
PCT Filed: |
August 6, 2007 |
PCT NO: |
PCT/NZ2007/000211 |
371 Date: |
July 9, 2010 |
Current U.S.
Class: |
700/108 ; 706/12;
707/802; 707/E17.005 |
Current CPC
Class: |
G05B 2219/32009
20130101; G05B 23/0294 20130101; G05B 19/41835 20130101; G05B
2219/31455 20130101; C25C 3/06 20130101; G05B 11/01 20130101; C25C
3/20 20130101 |
Class at
Publication: |
700/108 ;
707/802; 706/12; 707/E17.005 |
International
Class: |
G06F 19/00 20060101
G06F019/00; G06F 17/30 20060101 G06F017/30; G06F 15/18 20060101
G06F015/18 |
Foreign Application Data
Date |
Code |
Application Number |
Aug 9, 2006 |
AU |
2006904359 |
Claims
1. A system for controlling an industrial plant, the system
comprising: automatic control equipment comprising a plurality of
measurement sensors for sensing predetermined variables associated
with components of the industrial plant, the sensors generating
measured data relating to operation of the components of the
industrial plant; a database containing operational data, including
observational data, regarding the industrial plant; and a processor
in communication with the automatic control equipment and the
database for receiving the measured data from the sensors of the
automatic control equipment and the operational data from the
database, the processor manipulating the measured and the
operational data to provide an evolving description of a process
condition of each component over time, along with output
information relating to operational control of the industrial plant
and for updating the database.
2. The system of claim 1 in which the automatic control equipment
constitutes a first system level, the processor and database
constitute a second system level with the system including a third
system level, being a management level.
3. The system of claim 2 in which the management level uses the
information output from the processor for effecting control of the
industrial plant.
4. The system of claim 2 in which the levels are configured to
achieve an improvement in a number of operating variables of the
plant.
5. The system of claim 4 which is operable within a range for each
variable as determined by variability within the process and which
acts to reduce variation within each variable and other key process
variables through identifying abnormal or systemic, damaging
patterns of variation which can be related to a single dominant
cause.
6. The system of claim 1 which includes a classifier module in
communication with the processor for classifying variations of
operating variables of the plant into one of a predetermined number
of classes of variations.
7. The system of claim 6 in which the classifier module classifies
variations in a process variable into one of three classes being:
common cause or natural variation, special cause variation or
structural variation.
8. The system of claim 1 which is operable to take into account
information about the total process condition including process
variable trajectory over a preceding period of time.
9. The system of claim 1 in which, when the plant is an aluminium
smelting plant, the automatic control equipment includes bath
superheat sensors, bath resistivity sensors, sensors for monitoring
and noting electrical current variation and characteristic
frequencies, cell off-gas temperature and flow rate sensors, and
other control inputs.
10. The system of claim 9 in which the observational data relates
to the operational state of the individual cells, the operational
state being formally monitored and integrated into individual cell
process conditions and including: anode condition including red
carbon, airburnt anodes, red stubs, spikes, cracked anodes; bath
condition including carbon dust, gap between bath and crust, bubble
generation and location of evolution of the bubbles in the cell,
bath level; metal level and the projected metal tap history; cover
condition (remaining thickness and height on the anode
connectors/stubs), crust damage, fume escape from superstructure;
alumina and bath spillage on electrical conductors (rods, beams,
bus bars); control action history over previous weeks including
aluminium fluoride addition, alumina addition, extra voltage,
excessive, unplanned anode beam movements, metals and bath
transfers, etc; cathode condition including cathode voltage drop
(CVD) history, collector bar current density, instability history,
anode changing observations, anode effect frequency, etc; shell
condition, including red plates, shell deformation and excessive
heat rising to the catwalk from a certain shell location; hooding
condition--gaps, damage, fitment, door and quarter shield sealing;
bus bar and flexible damage, collector bars cut; lack of duct gas
suction as observed through fume escape into the pot room; feeder
operation, feeder chutes, feeder holes blocked, alumina not
entering feeder holes; side wall ledge condition, silicon carbide
mass loss, history of silicon level in metal; excessive liquid bath
output from cells or from a pot room, indicating a change in heat
balance causing melting of ledge, crust or dissolution of bottom
sludge; iron level in metal which is an indicator of bath level and
anode condition; trace elements in the metal which is indicative of
trends in current efficiency over time; flame colour, including
blue flames, lazy yellow flames (sludge), bright yellow (sodium)
shooting flames which may indicate some anode to metal direct
contact in a cell; and general housekeeping around each cell.
11. The system of claim 10 in which each operational state is
monitored automatically by the sensors, using regular cell
observations or both by the sensors and by observation, information
obtained from the monitoring process being integrated with state
variable measurements to build a description of the process
condition of each individual cell and its evolution over time.
12. The system of claim 11 in which the processor and the database
are operable to check the process condition for each cell
individually with the database being updated periodically.
13. The system of claim 9 in which the processor includes a causal
framework for relating identified problems and cell process
conditions to specific causes.
14. The system of claim 13 in which the causal framework forms part
of a learning algorithm of the processor which is improved and
periodically updated over time using data from the database.
15. The system of claim 14 in which the management level employs
causal trees containing the learning algorithm to provide a growing
framework of decision support and, in the case of a smelting
operation, cell diagnosis over time.
16. The system of claim 13 in which the processor further uses a
complexity measure to assess predictability of the process outcomes
and the overall operation of the plant.
17. A method of controlling an industrial plant, the method
comprising: monitoring operation of the industrial plant by a
plurality of sensors forming part of automatic control equipment;
transferring measured data from the sensors and observational data
relating to operation of the industrial plant to a processor;
accessing a database containing operational data including data
from the sensors and the observational data relating to operation
of the industrial plant, as periodically updated by the processor;
and generating evolving process condition descriptions of each
monitored component of the industrial plant and output information
relating to operation of the industrial plant.
18. The method of claim 17 which includes forming three system
levels, the automatic control equipment constituting a first system
level, the processor and database constituting a second system
level and a third system level being a management level.
19. The method of claim 18 which includes using the information
output from the processor in the management level for effecting
control of the industrial plant.
20. The method of claim 18 which includes configuring the levels to
achieve an improvement in a number of operating variables of the
plant rather than acting only to maintain the operating variables
at arbitrary target levels.
21. The method of claim 20 in which the plant is an aluminium
smelting plant and in which the method includes configuring the
levels to achieve improvements in a number of operational aspects
of the plant.
22. The method of claim 21 in which the operational aspects may
include feed control to achieve desired alumina dissolution; feed
control to reduce, and, if possible, eliminate, periods of sludge
accumulation; compositional control to maintain the mass of
aluminium fluoride at an approximately constant level in a bath in
each cell and reduce compositional and temperature variation over
time; energy balance control to maintain both sufficient superheat
and actual bath temperature for alumina dissolution; energy balance
control to inhibit periods of excessive superheat over time;
statistical and causal analysis to continuously reduce variations
across pot lines; and enterprise level management to assess actual
pot line capabilities cell by cell to organise and prioritise
improvement actions to improve capability over time and to optimise
the production of metals with specifications matching sales
orders.
23. The method of claim 21 which includes operating the plant
within a range for each variable as determined by variability
within the process and which acts to reduce variation within each
variable and other key process variables through identifying
abnormal or systemic, damaging patterns of variation which can be
related to a single dominant cause.
24. The method of claim 23 which includes correcting or minimising
identified causes as appropriate, reducing the range of each
process variable and improving process capability over time.
25. The method of claim 21 which includes classifying variations of
operating variables of the plant into one of a predetermined number
of classes of variations.
26. The method of claim 25 which includes classifying variations in
a process variable into one of three classes being: common cause or
natural variation, special cause variation or structural
variation.
27. The method of claim 21 which includes taking into account
information about the total process condition including process
variable trajectory over a preceding period of time.
28. The method of claim 21 in which the observational data relates
to the operational state of the individual cells, the method
including formally monitoring and integrating the operational state
into individual cell process conditions and the operational states
including: anode condition including red carbon, airburnt anodes,
red stubs, spikes, cracked anodes; bath condition including carbon
dust, gap between bath and crust, bubble generation and location of
evolution of the bubbles in the cell, bath level; metal level and
the projected metal tap history; cover condition (remaining
thickness and height on the anode connectors/stubs), crust damage,
fume escape from superstructure; alumina and bath spillage on
electrical conductors (rods, beams, bus bars); control action
history over previous weeks including aluminium fluoride addition,
alumina addition, extra voltage, excessive, unplanned anode beam
movements, metals and bath transfers, etc; cathode condition
including cathode voltage drop (CVD) history, collector bar current
density, instability history, anode changing observations, anode
effect frequency, etc; shell condition, including red plates, shell
deformation and excessive heat rising to the catwalk from a certain
shell location; hooding condition--gaps, damage, fitment, door and
quarter shield sealing; bus bar and flexible damage, collector bars
cut; lack of duct gas suction as observed through fume escape into
the pot room; feeder operation, feeder chutes, feeder holes
blocked, alumina not entering feeder holes; side wall ledge
condition, silicon carbide mass loss, history of silicon level in
metal; excessive liquid bath output from cells or from a pot room,
indicating a change in heat balance causing melting of ledge, crust
or dissolution of bottom sludge; iron level in metal which is an
indicator of bath level and anode condition; trace elements in the
metal which is indicative of trends in current efficiency over
time; flame colour, including blue flames, lazy yellow flames
(sludge), bright yellow (sodium) shooting flames which may indicate
some anode to metal direct contact in a cell; and general
housekeeping around each cell.
29. The method of claim 28 which includes monitoring each
operational state automatically by the sensors, using regular cell
observations or both by the sensors and by observation, information
obtained from the monitoring process being integrated with state
variable measurements to build a description of the cell process
condition of each individual cell and its evolution over time.
30. The method of claim 29 which includes operating the processor
and the database to check the process condition for each cell
individually and updating the database periodically.
31. The method of claim 21 which includes using a causal framework
to relate identified problems and cell process conditions to
specific causes.
32. The method of claim 31 which includes integrating the causal
framework into a learning algorithm of the processor which is
improved and updated over time using data from the database.
33. The method of claim 32 which includes employing causal trees
containing the learning algorithm to provide a growing framework of
decision support and, in the case of a smelting operation, cell
diagnosis over time.
34. The method of claim 21 which includes using a complexity
measure to assess predictability of the process outcomes and the
overall operation of the plant.
35. A system for controlling an industrial plant, the system
comprising: automatic control equipment comprising a plurality of
measurement sensors for sensing predetermined variables associated
with components of the industrial plant; a database containing
operational data, including observational data, regarding the
industrial plant; and a processor in communication with the
automatic control equipment and the database for receiving data
from the sensors of the automatic control equipment and from the
database, the processor using causal tree analysis comprising at
least one continually updated learning algorithm to provide a
framework of decision support and plant component diagnosis over
time.
36. A method of controlling an industrial plant, the method
comprising: monitoring operation of the industrial plant by a
plurality of sensors forming part of automatic control equipment;
transferring data from the sensors and observational data relating
to operation of the industrial plant to a processor; accessing a
database containing operational data, including the data from the
sensors and the observational data relating to operation of the
industrial plant, as periodically updated by the processor; and
using causal tree analysis comprising at least one continually
updated learning algorithm to provide a framework of decision
support and plant component diagnosis over time.
37. Automatic control equipment for a system for controlling an
industrial plant, the system comprising: a plurality of measurement
sensors for sensing predetermined variables associated with
components of the industrial plant; a signal processing module
responsive to the sensors and control input data; and a classifier
module in communication with the signal processing module for
classifying variations of operating variables of the plant, as
detected by the sensors, into one of a predetermined number of
classes of variations.
38. A method of operating an industrial plant, the method
comprising: monitoring operation of the industrial plant by a
plurality of sensors; processing data from the sensors and other
control inputs; and classifying variations of operating variables
of the plant, as detected by the sensors, into one of a
predetermined number of classes of variations.
39. A method of operating an industrial plant, the method
comprising monitoring operation of the industrial plant by a
plurality of sensors forming part of automatic control equipment;
transferring measured data from the sensors and observational data
relating to operation of the industrial plant to a processor;
accessing a database containing operational data, including the
data from the sensors and the observational data relating to
operation of the industrial plant, as periodically updated by the
processor, to provide mechanisms to assist in identification and
removal of causes of variations in the measured data; and combining
automatic control as carried out by the automatic control equipment
with said mechanisms to provide continuous improvement in the
operation of the plant.
40. A system for controlling an industrial plant, the system
comprising: automatic control equipment comprising a plurality of
measurement sensors for sensing predetermined variables associated
with components of the industrial plant; a database containing
operational data, including observational data, regarding the
industrial plant; and a processor in communication with the
automatic control equipment and the database for receiving data
from the sensors of the automatic control equipment and from the
database, the processor using a complexity measure to assess
predictability of the plant.
41. A method of controlling an industrial plant, the method
comprising: monitoring operation of the industrial plant by a
plurality of sensors forming part of automatic control equipment;
transferring data from the sensors and observational data relating
to operation of the industrial plant to a processor; accessing a
database containing operational data, including the data from the
sensors and the observational data relating to operation of the
industrial plant, as periodically updated by the processor; and
using a complexity measure to assess predictability of the plant.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] The present application claims priority from Australian
Provisional Patent Application No 2006904359 filed on 9 Aug. 2006,
the contents of which are incorporated herein by reference.
FIELD OF THE INVENTION
[0002] This invention relates to process control of an industrial
plant. More particularly, the invention relates to a system for,
and a method of, controlling an industrial plant particularly, but
not necessarily exclusively, a smelting plant.
BACKGROUND TO THE INVENTION
[0003] The present demand for an increasingly rapid financial
return in industrial plants such as smelting operations has driven
operating parameters beyond their current performance limits. This
has resulted in reduction in the lives of operating components of
the plant, reduced operating efficiencies and reduction in product
quality. The ever present need to reduce carbon and/or other
greenhouse gas emissions is adding additional pressure to the
situation. In the case of smelting operations, the control systems
that are in use were implemented in the early 1980s whereas
productivity, raw materials supply, energy price and environmental
issues associated with the industry have intensified considerably
since that time. Furthermore, the flexibility of pot line
electricity usage is an increasingly important issue for smelters
because of country and continental electricity grids and variation
in availability and price which connection to such grids can
impose.
[0004] Generally, control of processes has evolved in different
ways depending on the type of system under consideration. The
desire to maintain a process and its operating conditions at the
optimum operating parameters for which it was designed, or
subsequently retrofitted for the purpose of increased production
and minimal capital investment, is a common requirement since these
parameters determine the quality of the product and the efficiency
and cost of the process. In an attempt to maintain operation at
such optimum parameters, control systems have involved some form of
compensatory control loop or feedback loop in order to maintain
steady operating conditions for the industrial plant.
[0005] Thus, using a smelting operation as an example once again, a
normal control strategy has fixed or specified operating targets
for the key process variables associated with the smelting
operation. These key variables are adjusted in a compensatory
fashion using other control inputs. A problem with this approach is
that this may produce greater variation over time and compound the
initial causes of the variation. In fact, the initial causes of the
variation may not be addressed at all due to the reliance on
manipulation of control inputs not necessarily related to the
cause, allowing the causes of the variation to remain embedded in
the process and increase in number over time.
[0006] Further, in order to reduce complexity, assessment of the
process condition in smelting cells has been characterised by a
limited set of measurements performed, at different intervals, on
each cell. The last data point for each routinely measured variable
is usually the one used in assessment of cell state.
[0007] With the above arrangements, inadequate information is
provided to enable comprehensive operational or automatic control
of the smelting operation to be effected.
SUMMARY OF THE INVENTION
[0008] Throughout this specification the word "comprise", or
variations such as "comprises" or "comprising", will be understood
to imply the inclusion of a stated element, integer or step, or
group of elements, integers or steps, but not the exclusion of any
other element, integer or step, or group of elements, integers or
steps.
[0009] According to a first aspect of the invention, there is
provided a system for controlling an industrial plant, the system
comprising:
[0010] automatic control equipment comprising a plurality of
measurement sensors for sensing predetermined variables associated
with components of the industrial plant, the sensors generating
measured data relating to operation of the components of the
industrial plant;
[0011] a database containing operational data, including
observational data, regarding the industrial plant; and
[0012] a processor in communication with the automatic control
equipment and the database for receiving the measured data from the
sensors of the automatic control equipment and the operational data
from the database, the processor manipulating the measured and the
operational data to provide an evolving description of a process
condition of each component over time, along with output
information relating to operational control of the industrial plant
and for updating the database.
[0013] The automatic control equipment may constitute a first
system level, the processor and database constitute a second system
level with the system including a third system level, being a
management level. The management level may use the information
output from the processor for effecting control of the industrial
plant. The levels may be configured to achieve an improvement in a
number of operating variables of the plant.
[0014] The system and method are intended particularly, but not
necessarily exclusively, for use in an aluminium smelting
operation. For ease of explanation, the invention will be described
below with reference to this application. Those skilled in the art
will, however, appreciate that the system and method are suitable
for use in other applications. In particular, the method described
here is generally applicable to any complex industrial process
involving elements or sub-processes which interact in a non-linear
and/or unpredictable way and in which the state of the industrial
process has low observability for reasons of sensing or other
difficulties, and low controllability because of the interactive
nature of inputs and outputs and the varying and unpredictable time
scales of their response to a control input.
[0015] A non-exhaustive list of industrial processes with the
characteristics referred to above include: alumina refineries where
a multiplicity of interacting caustic liquor circuits exist, each
with a different dissolved sodium aluminate concentration and
degree of super-saturation, and some streams with precipitating
aluminium trihydroxide as well; steel plants where the iron ore
thermal reduction step, iron to steel making furnaces and
continuous casting processes are closely linked through steel
temperature, composition and heat transfer from the transporting
and holding vessels; steel or aluminium rolling and
annealing/coating lines where coil gauge and width profiles are
hard to measure a priori but have a profound effect on the heat
transfer to the strip as it is annealed and the correct velocity of
the strip through the annealing furnace for metallurgical
quality.
[0016] In an aluminium smelting operation, the smelter contains a
plurality of individual cells in which the smelting of aluminium
oxide, or alumina, occurs. The cells of the smelting operation are
arranged in lines, commonly referred to as pot lines. As indicated
above, the system levels are provided to achieve improvements in a
number of operational aspects of the smelting plant, more
particularly, feed control to achieve good alumina dissolution;
feed control to detect when good dissolution is not occurring and
to correct this to inhibit periods of sludge accumulation;
compositional control to maintain the mass of aluminium fluoride at
an approximately constant level in a bath in each cell and to allow
reduced compositional and temperature variation over time as
alumina feed control and energy balance control are improved;
energy balance control to maintain both sufficient superheat and
actual bath temperature for alumina dissolution; energy balance
control to inhibit periods of excessive superheat over time;
statistical and causal analysis to continuously reduce variations
across pot lines; and enterprise level management to assess actual
pot line capabilities cell by cell to organise and prioritise
improvement actions to improve smelter capability over time.
[0017] The system may be operable within a range for each variable,
as determined by variability within the process, and may act to
reduce variation within each variable and other key process
variables through identifying abnormal or systemic, damaging
patterns of variation which can be related to a single dominant
cause.
[0018] The system may include a classifier module in communication
with the processor for classifying variations of operating
variables of the plant into one of a predetermined number of
classes of variations. The classifier module may classify
variations in a process variable into one of three classes being:
common cause or natural variation, special cause variation or
structural variation.
[0019] Common cause or natural variation may occur where no
dominant cause is acting and a mix of causes results in a basically
random pattern of variation. This class of variation may not be
responded to automatically but may be the subject of process
investigations if certain circumstances are present such as if the
magnitude of the variation is still high or if there are safety
implications.
[0020] Special cause variation may be one where a statistically
significant, rarely encountered pattern of variation indicates that
a dominant cause is influencing the process at any one time and
that this cause is not part of the way the process is normally run.
This class of variation may be signalled or alarmed by the
automatic control equipment for investigation by operations staff.
The operations staff may use the processor to determine the cause
and, ideally, where possible, eliminate or correct the cause.
[0021] Structural variation may be one where non-random variation
occurs often or routinely through the action of physical and
chemical laws and the way the process is operated. Corrective
automatic control actions may be possible if undesirable structural
variations are detected by the sensors of the automatic control
equipment. This may require identifying, or "finger printing", the
structural variation and observing corresponding changes in process
condition over time.
[0022] In assessing the cell state, the present system may be
operable to take into account information about the total process
condition including process variable trajectory over a preceding
period of time. In addition, when the plant is an aluminium
smelting plant, the automatic control equipment may include bath
superheat sensors, bath resistivity sensors, sensors for monitoring
and noting electrical current variation and characteristic
frequencies, cell off-gas temperature and flow rate sensors, and
other control inputs such as the number of alumina shots fed to the
cells in different feeding modes, for example, underfeed or
overfeed modes and the degree of reduction in cell electrical
resistance which occurs when such feeding modes are executed.
[0023] The observational data may relate to the operational state
of the individual cells, the operational state being formally
monitored and integrated into individual cell process conditions
and including:- [0024] anode condition including red carbon,
airburnt anodes, red stubs, spikes, cracked anodes; [0025] bath
condition including carbon dust, gap between bath and crust, bubble
generation and location of evolution of the bubbles in the cell,
bath level; [0026] metal level and the projected metal tap history;
[0027] cover condition (remaining thickness and height on the anode
connectors/stubs), crust damage, fume escape from superstructure;
[0028] alumina and bath spillage on electrical conductors (rods,
beams, bus bars); [0029] control action history over previous weeks
including aluminium fluoride addition, alumina addition, extra
voltage, excessive, unplanned anode beam movements, metals and bath
transfers, etc; [0030] cathode condition including cathode voltage
drop (CVD) history, collector bar current density, instability
history, anode changing observations, anode effect frequency, etc;
[0031] shell condition, including red plates, shell deformation and
excessive heat rising to the catwalk from a certain shell location;
[0032] hooding condition--gaps, damage, fitment, door and quarter
shield sealing; [0033] bus bar and flexible damage, collector bars
cut; [0034] lack of duct gas suction as observed through fume
escape into the pot room; [0035] feeder operation, feeder chutes,
feeder holes blocked, alumina not entering feeder holes; [0036]
side wall ledge condition, silicon carbide mass loss, history of
silicon level in metal; [0037] excessive liquid bath output from
cells or from a pot room, indicating a change in heat balance
causing melting of ledge, crust or dissolution of bottom sludge;
[0038] iron level in metal which is an indicator of bath level and
anode condition; [0039] trace elements in the metal which is
indicative of trends in current efficiency over time; [0040] flame
colour, including blue flames, lazy yellow flames (sludge), bright
yellow (sodium) shooting flames which may indicate some anode to
metal direct contact in a cell; and [0041] general housekeeping
around each cell.
[0042] Each operational state may be monitored automatically by the
sensors, using regular cell observations or both by the sensors and
by observation, information obtained from the monitoring process
being integrated with state variable measurements to build a
description of the cell process condition of each individual cell
and its evolution over time. It will be appreciated that in any
industrial process there will be a set of equivalent observational
data representative of the operational state of the process.
[0043] The processor and the database may be operable to check the
process condition for each cell individually with the database
being updated periodically. For example, the cell process condition
may be updated at the commencement or termination of each
shift.
[0044] The processor may include a causal framework for relating
identified problems and cell process conditions to specific causes.
The causal framework may form part of a learning algorithm of the
processor which is improved and updated over time using data from
the database, including feedback from staff about the validity of
the causes identified and the effectiveness of corrective actions
applied. This may also include conflicts which are observed and
documented between the observational data and decisions and the
numerical state information and automated decisions at level 1 of
the control system. These updates may be subject to monthly review
by management before becoming part of the knowledge base in the
control system. Thus, the management level may employ causal trees
containing the learning algorithm to provide a growing framework of
decision support and, in the case of a smelting operation, cell
diagnosis over time.
[0045] The database may have information associated with each cell
and may contain process variable identifiers or "fingerprints"
associated with specific problems, process events and/or cell
process conditions.
[0046] The processor may further use a complexity measure to assess
predictability of the process outcomes and the overall operation of
the plant.
[0047] According to a second aspect of the invention, there is
provided a method of controlling an industrial plant, the method
comprising:
[0048] monitoring operation of the industrial plant by a plurality
of sensors forming part of automatic control equipment;
[0049] transferring measured data from the sensors and
observational (qualitative) data relating to operation of the
industrial plant to a processor;
[0050] accessing a database containing operational data including
data from the sensors and the observational data relating to
operation of the industrial plant, as periodically updated by the
processor; and
[0051] generating evolving process condition descriptions of each
monitored component of the industrial plant and output information
relating to operation of the industrial plant.
[0052] Thus, the method may include employing new formal control
objectives based on the long term reduction in variability of the
process, and on integrating human observation and decision making
into the computational organisation of sensed information in
traditional control systems.
[0053] The method may include forming three system levels, the
automatic control equipment constituting a first system level, the
processor and database constituting a second system level and a
third system level being a management level. The method may include
using the information output from the processor in the management
level for effecting control of the industrial plant. The method may
include configuring the levels to achieve an improvement in a
number of operating variables of the plant.
[0054] The plant may be an aluminium smelting plant and the method
may include configuring the levels to achieve improvements in a
number of operational aspects of the plant. These operational
aspects are generally not considered as part of the process
condition of the industrial plant from a control viewpoint. More
particularly, the operational aspects may include feed control to
achieve desired alumina dissolution; feed control to reduce, and,
if possible, eliminate, periods of sludge accumulation;
compositional control to maintain the mass of aluminium fluoride at
an approximately constant level in a bath in each cell and reduce
compositional and temperature variation over time; energy balance
control to maintain both sufficient superheat and actual bath
temperature for alumina dissolution; energy balance control to
inhibit periods of excessive superheat over time; statistical and
causal analysis to continuously reduce variations across pot lines;
and enterprise level management to assess actual pot line
capabilities cell by cell to organise and prioritise improvement
actions to improve capability over time and to optimise the
production of metals with specifications matching sales orders.
[0055] The method may include operating the plant within a range
for each variable as determined by variability within the process
and which acts to reduce variation within each variable and other
key process variables through identifying abnormal or systemic,
damaging patterns of variation which can be related to a single
dominant cause. Thus, the method may include correcting or
minimising identified causes as appropriate, reducing the range of
each process variable and improving process capability over
time.
[0056] The method may include classifying variations of operating
variables of the plant into one of a predetermined number of
classes of variations. In particular, the method may include
classifying variations in a process variable into one of three
classes being: common cause or natural variation, special cause
variation or structural variation.
[0057] The method may include taking into account information about
the total process condition including process variable trajectory
over a preceding period of time. The observational data may relate
to the operational state of the individual cells, the method
including formally monitoring and integrating the operational state
into individual cell process conditions and the operational states
including:- [0058] anode condition including red carbon, airburnt
anodes, red stubs, spikes, cracked anodes; [0059] bath condition
including carbon dust, gap between bath and crust, bubble
generation and location of evolution of the bubbles in the cell,
bath level; [0060] metal level and the projected metal tap history;
[0061] cover condition (remaining thickness and height on the anode
connectors/stubs), crust damage, fume escape from superstructure;
[0062] alumina and bath spillage on electrical conductors (rods,
beams, bus bars); [0063] control action history over previous weeks
including aluminium fluoride addition, alumina addition, extra
voltage, excessive, unplanned anode beam movements, metals and bath
transfers, etc; [0064] cathode condition including cathode voltage
drop (CVD) history, collector bar current density, instability
history, anode changing observations, anode effect frequency, etc;
[0065] shell condition, including red plates, shell deformation and
excessive heat rising to the catwalk from a certain shell location;
[0066] hooding condition--gaps, damage, fitment, door and quarter
shield sealing; [0067] bus bar and flexible damage, collector bars
cut; [0068] lack of duct gas suction as observed through fume
escape into the pot room; [0069] feeder operation, feeder chutes,
feeder holes blocked, alumina not entering feeder holes; [0070]
side wall ledge condition, silicon carbide mass loss, history of
silicon level in metal; [0071] excessive liquid bath output from
cells or from a pot room, indicating a change in heat balance
causing melting of ledge, crust or dissolution of bottom sludge;
[0072] iron level in metal which is an indicator of bath level and
anode condition; [0073] trace elements in the metal which is
indicative of trends in current efficiency over time; [0074] flame
colour, including blue flames, lazy yellow flames (sludge), bright
yellow (sodium) shooting flames which may indicate some anode to
metal direct contact in a cell; and [0075] general housekeeping
around each cell.
[0076] The method may further include monitoring each operational
state automatically by the sensors, using regular cell observations
or both by the sensors and by observation, information obtained
from the monitoring process being integrated with state variable
measurements to build a description of the cell process condition
of each individual cell and its evolution over time.
[0077] The method may includes operating the processor and the
database to check the process condition for each cell individually
and updating the database periodically. For example, the cell
process condition may be updated at the commencement or termination
of each shift.
[0078] In addition, the method may include using a causal framework
to relate identified problems and cell process conditions to
specific causes. The method may include integrating the causal
framework into a learning algorithm of the processor which is
improved and updated over time using data from the database. Thus,
the method may include employing causal trees containing the
learning algorithm to provide a growing framework of decision
support and, in the case of a smelting operation, cell diagnosis
over time.
[0079] The database may have information associated with each cell
and may contain process variable identifiers or "fingerprints"
associated with specific problems, process events and/or cell
process conditions.
[0080] The method may include using a complexity measure to assess
predictability of the process outcomes and the overall operation of
the plant.
[0081] According to a third aspect of the invention, there is
provided a system for controlling an industrial plant, the system
comprising:
[0082] automatic control equipment comprising a plurality of
measurement sensors for sensing predetermined variables associated
with components of the industrial plant;
[0083] a database containing operational data, including
observational data, regarding the industrial plant; and
[0084] a processor in communication with the automatic control
equipment and the database for receiving data from the sensors of
the automatic control equipment and from the database, the
processor using causal tree analysis comprising at least one
continually updated learning algorithm to provide a framework of
decision support and plant component diagnosis over time.
[0085] According to a fourth aspect of the invention, there is
provided a method of controlling an industrial plant, the method
comprising:
[0086] monitoring operation of the industrial plant by a plurality
of sensors forming part of automatic control equipment;
[0087] transferring data from the sensors and observational data
relating to operation of the industrial plant to a processor;
[0088] accessing a database containing operational data, including
the data from the sensors and the observational data relating to
operation of the industrial plant, as periodically updated by the
processor; and
[0089] using causal tree analysis comprising at least one
continually updated learning algorithm to provide a framework of
decision support and plant component diagnosis over time.
[0090] According to a fifth aspect of the invention, there is
provided automatic control equipment for a system for controlling
an industrial plant, the system comprising:
[0091] a plurality of measurement sensors for sensing predetermined
variables associated with components of the industrial plant;
[0092] a signal processing module responsive to the sensors and
control input data; and
[0093] a classifier module in communication with the signal
processing module, for classifying variations of operating
variables of the plant, as detected by the sensors, into one of a
predetermined number of classes of variations.
[0094] According to a sixth aspect of the invention, there is
provided a method of operating an industrial plant, the method
comprising:
[0095] monitoring operation of the industrial plant by a plurality
of sensors;
[0096] processing data from the sensors and other control inputs;
and
[0097] classifying variations of operating variables of the plant,
as detected by the sensors, into one of a predetermined number of
classes of variations.
[0098] According to a seventh aspect of the invention, there is
provided a method of operating an industrial plant, the method
comprising
[0099] monitoring operation of the industrial plant by a plurality
of sensors forming part of automatic control equipment;
[0100] transferring measured data from the sensors and
observational data relating to operation of the industrial plant to
a processor;
[0101] accessing a database containing operational data, including
the data from the sensors and the observational data relating to
operation of the industrial plant, as periodically updated by the
processor, to provide mechanisms to assist in identification and
removal of causes of variations in the measured data; and
[0102] combining automatic control as carried out by the automatic
control equipment with said mechanisms to provide continuous
improvement in the operation of the plant.
[0103] According to an eighth aspect of the invention, there is
provided a system for controlling an industrial plant, the system
comprising:
[0104] automatic control equipment comprising a plurality of
measurement sensors for sensing predetermined variables associated
with components of the industrial plant;
[0105] a database containing operational data, including
observational data, regarding the industrial plant; and
[0106] a processor in communication with the automatic control
equipment and the database for receiving data from the sensors of
the automatic control equipment and from the database, the
processor using a complexity measure to assess predictability of
the plant.
[0107] The use of the complexity measure may provide an early
warning of a trend to more chaotic or less reliable operation of
the plant (and/or the people in the plant) over time which will not
otherwise be detected by the more repetitive, regular operation of
control inputs and process outputs.
[0108] According to a ninth aspect of the invention, there is
provided a method of controlling an industrial plant, the method
comprising:
[0109] monitoring operation of the industrial plant by a plurality
of sensors forming part of automatic control equipment;
[0110] feeding data from the sensors and observational data
relating to operation of the industrial plant to a processor;
[0111] accessing a database containing operational data, including
the data from the sensors and the observational data relating to
operation of the industrial plant, as periodically updated by the
processor; and
[0112] using a complexity measure to assess predictability of the
plant.
[0113] The complexity measure may further alarm deteriorating
trends in the reliability of the plant or elements of it (for
example a particular part of a potline or a whole potline may start
to behave less reliably than others in the same smelter).
BRIEF DESCRIPTION OF THE DRAWINGS
[0114] Embodiments of the invention are now described by way of
example only with reference to the accompanying drawings in
which:
[0115] FIG. 1 shows a schematic block diagram of a system, in
accordance with an embodiment of the invention, for controlling an
industrial plant;
[0116] FIG. 2 shows a tabular representation of the system of FIG.
1;
[0117] FIGS. 3-6 show a graphic representation of an example of the
determination of state changes for a cell using a complexity
measure;
[0118] FIG. 7 shows a tabular representation of a first control
objective of the system;
[0119] FIG. 8 shows a tabular representation of a second control
objective of the system;
[0120] FIG. 9 shows a tabular representation of a third control
objective of the system;
[0121] FIG. 10 shows a tabular representation of a fourth and a
fifth control objective of the system; and
[0122] FIG. 11 shows a tabular representation of a sixth control
objective of the system.
DETAILED DESCRIPTION OF EXEMPLARY EMBODIMENTS
[0123] In FIG. 1 of the drawings reference numeral 10 generally
designates a system, in accordance with an embodiment of the
invention, for controlling an industrial plant. The plant is
designated generally by the reference numeral 12. The system 10
will be described with reference to its application to an aluminium
smelting operation but it will be appreciated that the system 10
could be used in any industrial plant requiring control. The
smelting operation or smelter 12 comprises a plurality of pot
lines. Each pot line is constituted by a plurality of cells in
which alumina is processed to form aluminium. Aluminium formed in
the cells is tapped off periodically for casting or further
processing downstream in the plant 12.
[0124] The purpose of the system 10 is to minimize energy
consumption and smelter emissions, maximize productivity and metal
purity and promote a safer, healthier working environment over time
by continuously reducing and removing variations in the process.
The control objectives of the system 10 include:-
[0125] 1. Feed control to achieve conditions for good alumina
dissolution for a high proportion of the operating time through
identifying early signs of poor dissolution conditions and acting
automatically if heat or composition related or through operational
decisions and intervention if not able to be corrected
automatically.
[0126] 2. Feed control to inhibit periods of sludge accumulation
which are usually characterised by low alumina input during an
underfeeding control mode and high alumina input during an
overfeeding control mode and through this objective and objective 1
above jointly to minimize anode effects on all of the cells.
[0127] 3. Compositional control which maintains the mass of
aluminium fluoride at an approximately constant level in the whole
cell but also provides signals and mechanisms to correct the causes
of aluminium fluoride mass variation within cells and pot lines,
thereby improving the stability of aluminium fluoride concentration
in the bath over time.
[0128] 4. Energy balance control which maintains both sufficient
superheat and actual bath temperature for alumina dissolution and
which inhibits large scale electrolyte freezing during normal cell
operations such as anode setting, alumina feeding, metal level
changes, or other additions to the cell.
[0129] 5. Energy balance control which minimises periods of
excessive superheat over time by signalling when causes of
excessive superheat are present and by giving decision support to
cause identification and elimination.
[0130] 6. Continuous reduction in variation across the pot lines by
categorising and finger printing sensor responses, integrating
operational information with cell data histories and connecting the
resulting cell process condition and identified process signals to
proven causal trees. These causal trees contain a learning
algorithm (as discussed below) and provide a growing framework of
decision support and cell diagnosis over months and years of
operation of the industrial process.
[0131] 7. Enterprise level management which assesses actual pot
line capability cell by cell and includes carbon plant and
casthouse capabilities to organise and prioritise improvement
actions over time. The system builds levels of planning information
for future years based on cell and pot line constraints and tested
solutions to result in maximized production, capture of higher
purity metal and satisfying sales order metal specifications
(through bath level, anode airburn and bath temperature variability
reduction and linkage to the casthouse metal batch and furnace
capture). It also matches anode capability and cell capability
across their respective populations to reduce anode/cell adverse
interactions over time.
[0132] The above objectives are achieved through a three step
control model of, firstly, observing the process, secondly,
understanding the variation and, thirdly, controlling the outcome.
Particular emphasis is placed on the multivariate nature and
non-linearity of the system 10. The system 10 is implemented within
an architecture which systemizes observation of total process
condition in the first control step, learning through a causal
framework associated with the second control step and a human
decision guidance module associated with the third control
step.
[0133] The system 10 includes automatic control equipment 14. The
automatic control equipment 14 comprises a plurality of sensors 16
for monitoring operating variables associated with the smelter 12.
The automatic control equipment 14 and the sensors 16 constitute a
first level of the system 10.
[0134] A second level of the system 10 comprises a processor 18
which communicates with the automatic control equipment 14. The
processor 18 is further in communication with a database 20 in
which operational data relating to the cells of the smelter 12 are
stored. In addition to the sensor and routine numerical
measurements, the database 20 contains data relating to problems
and process variable identifiers or "fingerprints" associated with
specific problems or events related to each cell of the smelter 12.
These data are derived from the sensors 16 as well as qualitative
observational data as detected by operations staff, as illustrated
schematically at 22, which is input into the database via the
processor 18.
[0135] A third level of the system 10 is a management level 24
which uses data from the database 20 to control and improve
operation of the smelter 12, as will be described in greater detail
below.
[0136] A tabular arrangement of the system 10 is shown in FIG. 2 of
the drawings. The automatic control equipment 14, as indicated
above, communicates with the sensors 16. These sensors 16, in turn,
feed to a plurality of monitoring modules. Hence, the automatic
control equipment 14 comprises an abnormal raw resistance monitor
26 for monitoring a rapidly acquired cell resistance signal and
detecting patterns and frequencies indicative of abnormal
operation. Together with alarming of cathode and anode
abnormalities, the resistance monitor 26 is used in combination
with a feed monitor module 28 which monitors the response of the
individual cell resistances of the smelter 12 to the feeding of
alumina at different rates corresponding to the different feed
control modes.
[0137] A temperature/liquidus control module 30 also communicates
with the feed monitor module 28 and with the raw resistance monitor
26. This module 30 monitors changes in temperature, liquidus
temperature and bath resistance per millimetre of anode beam
movement (if these sensors are active) and also includes an alumina
concentration dimension computed with reference to the feed monitor
module 28.
[0138] The automatic control equipment 14 further includes a signal
processing module 32 which receives signals from all of the sensors
16 and other control inputs. It summarises the essential character
(mean, range, trend, frequency) of these signals and controls the
supply of the resulting information to the processor 18 of the
system 10.
[0139] The operating variables of the smelter 12 are classified in
three classes, as will be described in greater detail below. To
enable this to occur, the automatic control equipment 14 includes a
classifier module 34. This classifier module 34 classifies
variations into one of the three classes. In addition, the
classifier module 34 classifies the cells of the pot line.
[0140] In level two of the system 10, the processor 18 includes a
cell process condition module 36 which communicates with the
database 20 for maintaining a history of state variables,
operational observations and non-conformances of both as well as
cell process complexity trends (which can show process state
changes) for all cells. The cell process condition is tracked
daily, weekly and monthly. Short and longer term aspects of the
cell process condition are determined and updated periodically, for
example, at the commencement or termination of each shift, weekly
and/or monthly.
[0141] The cell process complexity trends are used to determine
aspects of the cell state which are not evident from the normal
physical measurements of process variables. Predictability of the
system 10 is assessed using a complexity/information measure
referred to as T-entropy. Briefly, T-entropy is an algorithmic
technique which allows computation of the complexity of a finite
string of characters which are produced by symbolically
transforming information from visual and other analogue or digital
signals. (A complete treatment of the derivation of T-entropy can
be obtained from the following reference: Titchener, M. R.,
Gulliver, A., Nicolescu, R., Speidel, U. and Staiger, L. (2005)
Deterministic Complexity and Entropy. Fundamenta Informaticae,
64(1-4), 443-61.)
[0142] T-entropy is analogous to its thermodynamic equivalent which
is most commonly referred to as the `level of disorder` in a
(chemical) system. Similarly, T-entropy evaluates the level of
disorder in a finite, two-dimensional signal. T-entropy contains
information which is not provided in traditional signal processing,
where the repetitive, regular frequency characteristics of a signal
are determined. The non-repetitive, chaotic or non-linear elements
of signals in real world problems contain more information however.
It is this complex, real world behaviour which is transduced
through the T-entropy computation.
[0143] Using the example of a pseudo-resistance trace from a
pre-bake operating cell with metal pad noise developing over a
period of four hours, FIG. 3 exemplifies the computation of
T-entropy. In FIG. 3, an entropy surface 52 and its maximum entropy
54 for pseudo-resistance trace 56 is illustrated.
[0144] In FIG. 4, the maximum entropy 54, an integral under the
curve, or raster, 58 and a three-dimensional hybrid trace 60 are
illustrated. Another "view" of this (in a z-y plane) is shown in
FIG. 5 of the drawings. This shows areas of density in the hybrid
trace 60. The two areas of density are shown at 62 and 64 with the
transition between them shown at 66. The three clusters 62-66
indicate three distinct states of the process with state changes
associated with physical changes inside the cell which it is not
possible to detect routinely by other sensors and methods.
[0145] The three clusters or states 62-66 identified in FIG. 5 are
plotted against time as a trace 68 together with the T-entropy
output 52 and the pseudo-resistance output 56 as shown in FIG. 6.
This state information 68 would not have been obtainable from the
pseudo-resistance trace 56 alone and provides information
concerning changes in the complexity of the cell behaviour. The
information provided by the trace 68 is integrated into the cell
process condition module 36 and is made available to the
operational staff of the plant to enable analysis and remedial
action, if necessary, to be undertaken.
[0146] The processor 18 makes use of a learning algorithm 38 for
causal tree analysis. The learning algorithm comprises a causal
framework 39 (FIG. 11) for the pot line of the smelter 12 and
relates identified problems and cell process conditions to specific
causes. The processor 18 therefore communicates with the database
20 and under management authorisation the new causal/cell condition
links and corrective actions are added to the framework so that the
causal framework is improved and updated over time to render the
learning algorithm 38 more applicable to prevailing circumstances.
Operational staff at the plant provide feedback concerning the
success of causal analysis and recommendations into the processor
18 to integrate the practical aspect of plant operation and to
improve future control system actions.
[0147] The management level 24 includes an assessment module 40.
The assessment module 40 assesses pot line efficiency and
production capability on a cell by cell basis, taking note of the
incidence and severity of variations occurring in two of the
classes, i.e. special cause variation and structural cause
variation.
[0148] Additionally, the management level 24 comprises an analysis
module 42 for effecting planning options analysis based on
potential capability improvement and calculated risk over a period
of time.
[0149] As indicated above, the classification module 34 classifies
variations in each variable into one of three classes. These three
classes are:-
[0150] common cause or natural variation, special cause variation
and structural cause variation.
[0151] Common cause or natural variation is a variation where no
dominant cause is acting and the mix of causes results in a
basically random pattern of variation. This class of variation is
not responded to automatically but may be the subject of process
investigations if the magnitude of the variation is too high or has
safety implications.
[0152] Special cause variation is one where a statistically
significant, rarely encountered pattern of variation indicates that
a dominant cause is influencing the process at that particular time
and that this cause is not part of the way the process is normally
run. This class of variation is detected by the sensors 16 of the
automatic control equipment 14 and/or through the systematic
observations of staff at the cells during routine daily operations
and is investigated by the operations staff 22. The staff 22 use
the causal tree analysis of the processor 18 to determine and, if
possible, eliminate the cause.
[0153] Structural variation occurs where non-random variation takes
place often or routinely through the action of physical and
chemical laws and the way the process is operated. Automatic
corrective action is possible if undesirable structural variation
is detected in the cell sensors 16. This requires access to the
database to determine established connections between the cell
sensor responses, causes of variation and control objectives of the
system 10. This is achieved through finger printing the structural
variations and observing corresponding changes in the process
condition over time.
[0154] Insofar as the first step of observing the process is
concerned, the control strategy of the system 10 does not rely
solely on fixed or specified operating targets for the key process
variables such as bath temperature, bath composition, and cell
voltage. Rather, the control strategy operates over a range of
process variables determined by variability within the process
itself to produce a target cell process condition related to the
desired process outcomes (for example energy efficiency, metal
purity, anode effects, cell life, cost of production and safety).
The control strategy acts to reduce variation in the range of these
key process variables through identifying abnormal or systemic,
damaging patterns of variations which can be related to a single
dominant cause at a given point in time. The identified causes can
then be corrected or eliminated as appropriate reducing the range
of each process variable and improving the process capability over
time. Thus, the system uses, in addition to the existing cell
sensors 16, new cell sensors such as bath superheat sensors, bath
resistivity sensors, sensors for monitoring anode current variation
at characteristic frequencies, cell off-gas temperature and flow
rate sensors and other control inputs.
[0155] In addition, observational data as detected by the operation
staff 22 include monitoring the operational state of the cells by
monitoring of the following: [0156] anode condition including red
carbon, airburnt anodes, red stubs, spikes, cracked anodes; [0157]
bath condition including carbon dust, gap between bath and crust,
bubble generation and location of evolution of the bubbles in the
cell, bath level; [0158] metal level and the projected metal tap
history; [0159] cover condition (remaining thickness and height on
the anode connectors/stubs), crust damage, fume escape from
superstructure; [0160] alumina and bath spillage on electrical
conductors (rods, beams, bus bars); [0161] control action history
over previous weeks including aluminium fluoride addition, alumina
addition, extra voltage, excessive, unplanned anode beam movements,
metals and bath transfers, etc; [0162] cathode condition including
cathode voltage drop (CVD) history, collector bar current density,
instability history, anode changing observations, anode effect
frequency, etc; [0163] shell condition, including redylates, shell
deformation and excessive heat rising to the catwalk from a certain
shell location; [0164] hooding condition--gaps, damage, fitment,
door and quarter shield sealing; [0165] bus bar and flexible
damage, collector bars cut; [0166] lack of duct gas suction as
observed through fume escape into the pot room; [0167] feeder
operation, feeder chutes, feeder holes blocked, alumina not
entering feeder holes; [0168] side wall ledge condition, silicon
carbide mass loss, history of silicon level in metal; [0169]
excessive liquid bath output from cells or from a pot room,
indicating a change in heat balance causing melting of ledge, crust
or dissolution of bottom sludge; [0170] iron level in metal which
is an indicator of bath level and anode condition; [0171] trace
elements in the metal which is indicative of trends in current
efficiency over time; [0172] flame colour, including blue flames,
lazy yellow flames (sludge), bright yellow (sodium) shooting flames
which may indicate some anode to metal direct contact in a cell;
and [0173] general housekeeping around each cell.
[0174] The cell process condition elements monitored above can be
monitored either by the operations staff 22 or by the sensors 16.
This information is integrated with state variable measurements to
build a description of the total process condition of each
individual cell and its evolution over time.
[0175] The cell process condition for each cell is tracked
individually by the processor 18 and is updated periodically, for
example, at the commencement or termination of each shift. The
process condition description is also used in the automatic control
equipment 14 and may be used for operational decisions during any
shift for individual cells or for the pot line. The process
condition is also used for process engineering investigations to
correct individual cells manifesting long term problems and,
finally, is used by the management level 24 which uses pot line
condition for judging the capability of the smelter 12 to alter its
operating settings, for example, production rate or energy
usage.
[0176] As indicated above, understanding the variation in the
operating variables comprises classifying these variations in one
of the three classes.
[0177] The third step of the control system 10 is achieved by
altering the traditional function of each level of the system 10 so
that the new control objectives set out above are met. Insofar as
the first level is concerned, the control system 10 seeks to
achieve a systematic reduction in individual cell variations
through corrective control of variables such as alumina feed, bath
composition and energy input and relies on the integration of
operator observations of cell condition and their subsequent well
informed decisions and actions to remove causes of these
variations.
[0178] The second level of the system 10 seeks to achieve pot line
variation reduction through removal of causes of the variations. It
further relies on pot line management which emphasises decision
making using the database framework of variation and causes which
is continually updated with evidence accumulating over time and is
systematically linked to the operational observations and practical
decision making so that contradictions between theoretical control
decisions and direct observation are constantly being sought and
resolved. It also facilitates individual cell process condition
description and tracking using T-entropy trends to identify hidden
state and state change information. This level also uses human
decision guidance linking physical cell condition stimuli to
detection and decision making.
[0179] The management level 24 of the system 10 is used to effect
pot line capability assessment based on cell state, metal purity
distribution data from the second level and quantified improvement
potential. The cell distribution data is linked to improvement
strategies such as reducing poorly performing cells or moving the
entire cell distribution, as well as metal marketing and financial
planning and control.
[0180] FIG. 7 shows a tabular implementation of the first control
objective for achieving good alumina dissolution. The module
letters correspond to the module labels in FIG. 2 of the
drawings.
[0181] Similarly, FIG. 8 shows a tabular representation of the
second control objective of elimination of periods of sludge
accumulation without incurring anode effects. As is the case with
the first control objective, the second control objective relies on
operational observations triggered automatically by sensing and
level 1 control logic to indicate specific observations (concerning
the alumina feeders primarily) required to achieve the control
objectives.
[0182] FIG. 9 shows a tabular representation of the third control
objective of compositional control based on achieving near constant
mass of aluminium fluoride in each cell and its improvement over
time. In this control objective, there is, once again, a
requirement for observational data and also operator input for
adjustment of the cell or line, particularly in the case where the
variation is identified as being special cause. The identification
of adverse structural variation such as thermal and compositional
cycling allows these to be related to the systemic causes embedded
in the control system and the smelter 12 itself through the
learning algorithm 38 at level 2 in the control system.
[0183] FIG. 10 is a combination of control objectives 4 and 5 to
achieve energy balance control to maintain changes in temperature
within a range which can be withstood by the cell without damage to
the process.
[0184] In the case of control objective 6, this control objective
is met substantially at level 2 of the system 10 and is shown
diagrammatically in FIG. 11. Because the system design is now
specifically aimed at improvement and not only control, the
architecture of the level 2 system differs from previous
supervisory systems.
[0185] Better understanding of what constitutes the cell process
condition now enables a single screen view of the state of each
cell of the smelter 12 and incorporates both state variable
measurement and operational state attributes as well as the
respective histories. One embodiment of this view is shown in the
first part of FIG. 11. Each variable or attribute is described by a
colour being red (R), orange (O), blue (B), or green (G)
representing not only the last observation but also the stability
of the observations over a specified operating period and within
the stable, multi-dimensional control volume for selected groups of
variables. Red and orange status indicates abnormal status
conditions requiring attention and potential abnormality
respectively.
[0186] Taking the example of "Alumina kg/d" the stability of the
uni-variate measurement will be judged by the statistical stability
of the cusum of the "Alumina Daily addition. The capability of the
cell with respect to "Alumina Feeding" will be judged by the
flatness of the Cusum Chart. In other words: "Is the cell
consumption of alumina matched to the metal production rate?"
However, this variable is also combined into multivariate views of
the whole cell process condition because of its interaction with
the thermal arid compositional balance. In this example of alumina
feeding, kg/d of alumina fed during an underfeed mode and kg/d of
alumina fed during an overfeed mode can be analysed as a bi-variate
surface, leading to a state descriptor for feeding, as one element
of the overall cell process condition.
[0187] The database 20 contains the normal comprehensive numerical
information over time, but with new classes of discontinuous, cell
specific information as shown in FIG. 11. This "event driven" data
is stored in time stamped flat files and is used along with process
variable fingerprints stored in the database to establish likely
causes within the causal framework 39.
[0188] The causal framework 39 is largely automated in its data
queries and logic processing. It is designed to respond to
management requirements in two ways by, firstly, providing causes
and corrective actions for individual problems through request at
any time. These requests can also be automated at a start of a
shift through the cell process condition module 44, if
required.
[0189] The causal framework 39, secondly, provides timed (daily,
weekly, monthly) review reports to people within the organisation.
These reports are configurable and summarise problems requested,
those resolved and those with adverse consequences stemming from
the advice provided, learning opportunities formulated (for
authorisation) and conflicts between causal logic and observations
(for resolution). This is provided on a human decision guidance
module 46.
[0190] The causal framework 39 drives improvements in control and
in performance on the pot line by use of the enhanced database 20
and process condition descriptions to solve single cell and
systemic pot line problems.
[0191] The presentation of summary data on the number of problems
outstanding on the number of cells in various states of control is
a stimulus for management attention and is facilitated by having a
continuous tracker 48 of both cell process condition and identified
cell problems. The tracker 48 aids in operational implementation of
cell action plans as shown at 50.
[0192] The tracker 48 plays an integral part of the management
process embedded in level 2 of the system 10. Decisions are based
on the scientifically formulated and evidentially confirmed causal
framework 39, the diagnosed cell process condition and the computed
trend in the complexity or chaotic nature of the cell condition
using T-entropy.
[0193] The control objective 7 relies on achievement over time of
the first six control objectives. It also requires that the
measured and predicted future capability of the pot line is
formally integrated into financial management and planning
processes for the smelter 12. This is achieved by the modules 40
and 42 of the management level 24. The actual design of the modules
40 and 42 will depend on the enterprise level system which is in
use at the smelter 12.
[0194] It is therefore an advantage of the invention that an
improved system 10 is provided which enables more accurate control
of a smelter 12 to be achieved over a period of time by the use of
observational data, a causal framework 39 and automatic control
equipment 14 which is more integrated with the formal control
objectives and with the observations of the staff. With the new
system, reduction in variation in individual cells through
integrated automatic and operational control decisions can be
achieved over a period of time resulting, in the long run, in
improved operating efficiencies of the smelter 12.
[0195] A further advantage of the system 10 is that it achieves
integration of energy, composition, alumina feed and operational
controls with smelter improvement plans to minimise energy
consumption and smelter emissions and to maximise production of
metal of the highest possible purity/value over time. Still
further, it facilitates a holistic assessment of the process
condition of each individual cell, the process condition of each
cell being maintained and updated over time.
[0196] It will be appreciated by persons skilled in the art that
numerous variations and/or modifications may be made to the
invention as shown in the specific embodiments without departing
from the spirit or scope of the invention as broadly described. The
present embodiments are, therefore, to be considered in all
respects as illustrative and not restrictive. In particular, while
the system and method have been described with reference to its
application in an aluminium smelting plant, that has been done for
ease of explanation only. The system and method are equally
applicable in any industrial process where a set of equivalent
observational data representative of the operational state of the
process can be employed in improving the operation of the
process.
* * * * *