FIELD OF THE INVENTION
The present invention relates to a Support System for taking right operational
decision at right time at the shop floor of a manufacturing plant for the
Operational Performance Improvement. More particularly, the present invention
relates to an on-line rule-based support system to improve thermal state of blast
furnaces in steel plants.
BACKGROUND OF THE INVENTION
Iron making is one of the major cost drivers of an integrated steel plant. Safe
operational practice and a long campaign life is required to be maintained. In
addition, increase in productivity is a target. Thus, to reduce the hot metal
production costs, it is necessary to integrate the intelligent decision making
parameters with the blast furnace operation control system. Right decision at
right time for right solution is the need for efficient blast furnace operation.
Success of any manufacturing organisation largely depends on the the
operational excellence. Thus, right decision at right time to implement right
solution is an inescapable need for the operating plants. This is normally
achieved through the following steps:-
The business objective, strategy, business performance measures and business
rules of the value chain are aligned to the operational objective, strategy,
operational performance measures and operating rules of the manufacturing
units.
These operational decisions or operating rules and performance indicators are
designed by the managers of the operational units along with experienced

operators and process experts, from the business rules and business
performance indicators.
These operating rules are implemented by the shift operators at the shop floors
to monitor and control operational processes.
This is done offline. Once the operating rules are ready, shift operators are
trained by the plant managers and experienced operators, so that
implementation of these rules happen consistently across all shifts for corrective
actions. On the other hand, operators must choose right rules at right time for
better process control. Corrective actions taken by the operators across the
shifts are monitored and reviewed by the plant managers on daily basis. This
process is known as Shop Floor Daily Management. Thus shop floor daily
management depends on the decision making capability of the operators.
Decision making capability of the plant operators is directly linked to their
learning and experience curves. Hence decisions taken by the operators across
the shifts are not consistent. Also, at shop floor, decision making takes place at
high frequency by the operators across the shifts, which is mostly repetitive in
nature, whereas managerial decision making is less repetitive and more
unstructured. Thus the managerial learning and decision making process has to
be aligned to the learning and decision making process of the shop floor
operators for effective daily management.
There have been many offline initiatives like TQM, Six sigma etc, to develop
managerial procedures, to improve the alignment. Manufacturing units have also
implemented real time rule-based expert systems to control their processes
online based on experts knowledge. Though these type of expert systems are
capable of detecting process disturbances and providing suggestions to the

operators for corrective actions, but they have not yet mastered the art of
manufacturing performance improvement of a plant because the rules are
designed only by the technical experts for plant problem solving. Thus, these
independent initiatives are not able to bring right decision at right time for right
solution for effective operational performance improvement and management.
At present a large number of blast furnaces has known Expert systems. The rules
in the known Expert System are embedded in the software code, which can only
be changed or altered by the system manufacturer. Thus maintainability of such
system is known quite problematic. A modified generation module, which is
modified offline partially enables the system maintenance based on trial and
error rule. To make the knowledge acquisition scientific, it is necessary to
integrate the case and effect to generate operating rules and key parameters for
process control.
OBJECTS OF THE INVENTION
It is therefore an object of the invention to propose an on-line rule-based
support system to improve thermal state of blast furnaces in steel plants.
Another object of the invention is to propose an on-line rule-based support
system to improve thermal state of blast furnaces in steel plants, which is
enabled to continuously analyse operation of the blast furnace.
A still another an on-line rule-based support system to improve thermal state of
blast furnaces in steel plants, which integrates key performance index (KPI)
including standard operating rules (SOP) of the steel plant.

SUMMARY OF THE INVENTION
Accordingly, there is provided A key-performance indicators (KPI) based on-line
support system to improve thermal state performance of a blast furnace,
comprising a processor with memory devices incorporating data relating to the
desired Key-performance of the blast furnace; a plurality of sensor devices
positioned at various locations of the blast furnace to capture real-time
operational parameters of the blast furnace; an interface engine allowing transfer
of captured data to the processor including communication between a knowledge
engineering module; a display device incorporated in operators panel exhibiting
key parameters including cause and effect process indicators; the fuzzy logic
based knowledge engineering module generating heuristic operation control rules
using a standard operating practice (SOP), and monitoring and controlling the
key performance indicators; and a decision analysis and reporting module for
storing of data relating to decisions taken by the operators on daily basis to
overcome operating constraints which includes KPI trend graph and decision rule
graph wherein the thermal state (A) is the summation of parameters X1, Y1 and
Z1, wherein X1 = X* 0.6, X being (((HMT (n)*0.75) + (HMT(n-l)*0.25-
1490)/1490*100)+100))) wherein Y1 = Y*0.1, y being (((C (n)*0.75) + (C(n-
1)*.25)-4.4)/4.4*100)+100))), wherein Z1 = Z* 0.3, Z being =(((Si (n)*0.75) +
(Si(n-1)*.25)-0.9)/0.9*100)+100))), wherein n = last cast, ( n-1) = previous to
last cast, the weight factors considered for each of the parameters, HMT = 60%,
Si = 30%, C = 10%, n=75%, (n-1) = 25%, and wherein the target values are
HMT = 1490°C, C = 4.4%, and Si = 0.9%.
According to the invention, the KPI and SOP of the blast furnace are aligned to a
generic business model of the iron and steel supply chain. The knowledge
engineering part of the technology is designed with an cause and effect tool

(fish-bone diagram). Fuzzy heuristic operation control rules are generated using
the KPI and SOP of blast furnace. This technology has been demonstrated
successfully in one of the existing blast furnaces, for online decision making,
operation control and daily management of the thermal state, by the operators
at the shop floor. This system works as a monitoring and control system to
indicate to the operators for corrective actions. But the operators are allowed to
take final decisions and feed their observations into the system. Observations
logged by experienced operators help in improving the daily SOP of blast
furnace. Decisions taken by the operators across the shifts are reviewed at
higher level.
Accordingly, there is provided an Operating Rule based Online Learning and
Decision Support System for the control and performance improvement of
Thermal State of the Blast Furnace. The system can be customized for online
control and performance improvement of any plant at the shop floor level. This
system operates as a Rule-based expert system analyzes the furnace operation
process continuously. The prior data is stored in the system knowledge based in
the form of operating rules and available 24 hours a day. Adaptation of an offline
dynamic production rule generation system, maintained by the operators through
off-line terminals for modifying and test running the knowledgebase, enables the
inventive system largely to be maintained, operated and controlled by the field
personnel. In case of any abnormalities, the expert system indicates a corrective
action, and translates rules from calculations and expert models. Nevertheless,
the operators have the option to take the last decision and execute.
According to the invention, the knowledge is represented using Key Performance
Index and Cause and Effect Diagrams (Fish-bone diagrams).

BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS
Figure 1 - shows the cause and Effect Diagram for Thermal State
Figure 2 - shows the Online Data Control Diagram for the Rule-based System
Figure 3 - shows the Flow Diagram of the interface Engine of the Rule-based
System
Figure 4 - Shows the Rule Implementation Process through the Dashboard
Figure 5a - Shows the Daily KPI (Thermal State) Trend Graph
Figure 5b - Shows the Daily Decision Trend Graph
Figure 5c - Shows the Daily Decision History
DETAILED DESCRIPTION OF THE INVENTION
Operating Rule-based Learning and Decision Support Technology for
Blast Furnace Operation
Blast furnace unit runs 24hrs in three shifts based on the decisions of the line
operators across three shifts. The target of the operating plant is to achieve the
performance target of the operational KPIs, based on the daily management
guidelines and the standard operating practice (SOP). This is developed by the
plant management team and shop floor managers together with the process
experts. Real-time process control at the shop floor is done, based on the SOP
rules. These rules are kept in offline documents and used by the operators for
real time process control. Since daily performance management process is
derived from the Business Performance Management process, hence
development of a operating rule-based-based Learning and Decision Support
Technology to assist the blast furnace operators to take decision in real-time
Process Control is an innovative effort. This technology has four primary
modules;

Knowledge Engineering module;
Inference Engine,
Operators Panel, and
Decision Analysis and Report generating module.
Knowledge engineering module
This is an offline module, which includes knowledge acquisition, representation
and classification. The prime objective of the technology is to capture and
encode the tacit knowledge of the domain experts in the form of fuzzy decision
making rules, which is also known as knowledge engineering. These rules occur
in sequences and are expressed in the form of, if then ,
where, if conditions are true, then actions are executed. Innovative design has
been used in the knowledge engineering module, with a cause and effect tool
(fish-bone diagram). This gives a clear definition to the correlation between the
KPI (effect) and the process indicators (cause).
Once the KPI and the process indicators are represented in a fish-bone diagram,
these parameters are then normalised in a fuzzy domain of, very low, low,
medium, normal, high, very high. "If-then"; fuzzy heuristic operation control
rules are generated using a daily management SOP for monitoring and control of
the KPI. Among the measures taken are the adoption of a dynamic production
rule generation system by the operators through off-line terminals for modifying
and test running the data base. As a result of these efforts, the knowledge base
can be largely maintained, operated and controlled by the operators. Since the
system is implemented on-line, periodic study meetings are held with the
Furnace operational team to evaluate the performance and update the
knowledge base.

Inference Engine
Based on the online data of the KPIs and the process indicators, a dialogue is
conducted with the knowledge base about the problem to be solved. The
system then provides insights derived (or inferred) from the process knowledge-
based SOP rules. These insights are provided by an inference engine after
examining the knowledge base. When the production rules are examined by the
inference engine, actions are executed, if the information supplied by the user
satisfies the conditions in the rules. Two methods of inference often are used,
forward and backward chaining. The current technology is designed using the
forward chaining inference concept.
Operators' Panel
All KPIs and Cause and Effect process indicators are displayed on an Operators'
Panel. The Operators' panel gets populated automatically once the KPIs and
process indicators are represented in the fish-bone diagram. This panel is
connected online to a data base of Level-2 Automation system at the backend.
Level-2 system fetches real-time data from a Level-1 plant data acquisition
system. Data transmitted either directly from the level-2 data base or through
calculations for any of the parameters in the cause and effect diagrams. Thus,
the operators' panel exhibits the real-time state of the KPIs and the process
indicators. Once the real-time data gets populated in the panel, the inference
engine fetches corrective action from a rule-base, and displayed in the panel.
The Operators take decision based on the instructions transmitted by the system.

Decision Analysis and Reporting module
Decision analysis is an offline module. Daily decision taken by the operators
across the shifts are logged in the system and stored in the system database.
This data is used to generate daily report for review in the daily management
meeting by the plant managers and the operators. Decision Analysis report
contains the KPI trend graph and decision rule graph and SOP rules implemented
by the operators across three daily shifts. Since operators are allowed to take
final decisions and feed their observations into the system, observations logged
by experienced operators help in improving the daily SOP of blast furnace.
Decisions taken by the operators across the shifts are discussed in the daily
management meetings by the operation managers along with the shift operators.
Hence operators are able to learn from each others decision making process and
knowledge. Each time a SOP is implemented, the hit count against that SOP rule
increases. This helps the process experts, plant managers to learn about the
SOP of the plant for daily management. Hence this is not only a Decision
Support System but also a Learning System.
Online Thermal State Control of the Blast Furnace - (EXAMPLE!
It is the prime responsibility of the operators to keep the furnace thermally stable
during production. Thus, Thermal state is the most important KPI (Key
performance index) in the blast furnace operation. This is calculated from other
process indicators, i.e.; HMT (Hot Metal temperature) of the present cast and the
previous cast, HM Si (Hot Metal Silicon) of the present and previous cast, HM C
(Hot metal Carbon) of the present cast and the previous cast. The SOP rules
are generated to control the thermal state of the furnace by controlling the Coke
Rate, Coal Injection Rate and Humidity.

Thermal State Operating Rules of the Blast Furnace
Thermal State Calculation : Thermal state is calculated with the online data
and weight factors from the formula below. This formula is a part of the
backend code. Rules for the online control of the thermal state is given in the
table above.
Thermal State : A = X1 + Y1 + Z1
where
X1 = X* 0.6, Y1 = Y* 0.1, Z1 = Z * 0.3
and
X =(((HMT (n)*0.75) + (HMT(n-1)*0.25)-1490)/1490*100)+100)))
Y =(((C (n)*0.75) + (C(n-l)*.25)-4.4)/4.4*100)+100)))
Z =(((Si (n)*0.75) + (Si(n-l)*.25)-0.9)/0.9*100)+100)))
n = last cast, ( n-1) = previous to last cast,
the weight factors taken for each of these parameters are follows
HMT = 60%, Si = 30%, C = 10%, n = 75% , (n-1) = 25%
The target values are
HMT = 1490 Degree Centigrade, C = 4.4%, Si = 0.9%
These parameters are represented by the cause and effect diagrams as shown in
figure 1 using the fish-bone diagram generation tool of the system. Thermal
State, which is the KPI, is represented as the effect at the tip of Fish-bone,
whereas HMT, HM Si, HM C, Coal Injection Rate, Coke Rate and humidity are in
the tail part, representing the cause parameters.

Once the KPI is entered and the cause and effect diagrams are completed, fuzzy
normalisation table is created to map the KPIs and process indicators values to
the fuzzy domain. The rule generation table is also automatically created for the
entry of the Thermal State Control rules. Brain storming workshops are
conducted with the furnace operators and managers to generate and populate
the thermal state control rules from the SOP. Any change in the business
process gets reflected directly in the daily management SOP, KPI, and the
process indicators.
Thus the system works in the real time in a control mode in the blast furnace
shop floor. The operators are allowed to take final decisions and feed their
observations into the system. Observations logged by the experienced operators
help in improving the daily SOP of blast furnace. Decisions taken by the
operators across the shifts are reviewed at higher level.
Table 1: Cause and Effect Diagram for Thermal State
Decision Analysis


Benefits of Operating Rule-based Leaning and Decision Support
Technology at F BF
• Maintaining productivity and quality despite being at the end of its
campaign life.
• Standardized the action taken by shift operators
• Improvement in process capability (Cpk) for hot metal Silicon control.
• Improvement in standard deviation of hot metal silicon.
• Lower fuel rate
Following additional benefits are also envisaged from the system: System to
make Tacit knowledge explicit, Training support system, Audit support system,
Corporate memory system for tacit knowledge, Act as Knowledge engineering
platform in brain storming process, Keep the team on focus during brain
storming, Help to Standardise SOP. Expert system analyzes process continuously
to detect process disturbances.

WE CLAIM
1. A key-performance indicators (KPI) based on-line support system to
improve thermal state performance of a blast furnace, comprising:
- a processor with memory devices incorporating data relating to the
desired Key-performance of the blast furnace;
- a plurality of sensor devices positioned at various locations of the
blast furnace to capture real-time operational parameters of the
blast furnace;
- an interface engine allowing transfer of captured data to the
processor including communication between a knowledge
engineering module;
- a display device incorporated in operators panel exhibiting key
parameters including cause and effect process indicators;
- the fuzzy logic based knowledge engineering module generating
heuristic operation control rules using a standard operating practice
(SOP), and monitoring and controlling the key performance
indicators; and
- a decision analysis and reporting module for storing of data relating
to decisions taken by the operators on daily basis to overcome
operating constraints which includes KPI trend graph and decision
rule graph

wherein the thermal state (A) is the summation of parameters X1, Y1
and Z1, wherein X1 = X* 0.6, X being (((HMT (n)*0.75) + (HMT(n-
1)*0.25-1490)/1490*100)+100))) wherein Y1 = Y*0.1, y being (((C
(n)*0.75) + (C(n-1)*.25)-4.4)/4.4*100)+100))), wherein Z1 = Z* 0.3,
Z being =(((Si (n)*0.75) + (Si(n-1)*.25)-0.9)/0.9*100)+100))),
wherein n = last cast, ( n-1 ) = previous to last cast, the weight
factors considered for each of the parameters, HMT = 60%, Si = 30%,
C = 10%, n=75%, (n-1) = 25%, and wherein the target values are
HMT = 1490°C, C = 4.4%, and Si = 0.9%.