FIELD OF INVENTION
The invention relates to a continuous casting process in the steel making chain.
More particularly, the invention relates to a system for real time monitoring and
control of a continuous casting process.
BACKGROUND OF INVENTION
Process monitoring and supervisory control systems are fairly widespread in
industrial processes, vehicular systems, building systems and the like. For an
industrial process, the importance of such a system and its performance and
reliability requirements stem from the criticality of the process that it is designed
to monitor and control.
Steelmaking is a major field of industrial activity with many component sub-
processes. One among the most critical of these is the continuous casting
process [1], where molten steel is uninterruptedly converted into solid slabs or
bars for further downstream processing. The criticality of this process flows from
two factors; firstly, it represents a bridge between the upstream processing of
molten steel and downstream solid processing and any disruption of this process
affects operations in both directions, secondly, the solidification process has
significant influence on the quality of the final product and minor drift in casting
parameters can have major impact on product quality even leading to
downgrading.
The continuous casting process essentially converts molten steel in ladles coming
from upstream production units into a continuous completely solidified strand
that is gas-cut into discrete slabs for downstream processing. Fig. 1 provides an
overview of the process. The molten steel is poured (1) from ladles into a
'tundish' which serves as a buffer between the batch process (ladles) and the
process of continuous casting. A Submerged Entry Nozzle (SEN) supplies (2)
steel from the tundish to the mold. The latter is of rectangular section and
around 0.9 meters in length; sectional dimension is approximately 0.25 x 1.5
meters. The inner walls of the mold are composed of copper plates to facilitate
heat transfer. Cooling water flows in pipes within mold walls to remove heat. The
steel flowing out from the SEN forms (3) a solid shell all around the interior of
the mold wall due to cooling, the shell thickens as the strand moves downwards.
The mold oscillates (4) along its longitudinal axis at frequencies of 1-5 Hz and
amplitude around 5 mm in a waveform that is slightly distant from sinusoidal
towards sawtooth [2]. The oscillation serves two purposes, first by creating
alternating relative motion between the mold wall and the formative shell
(moving monotonously downwards), the mutual sticking is prevented, second,
facilitating penetration of lubricant poured onto the molten steel surface near the
mold top into a thin (less than 1 mm) gap between the mold wall and formative
steel shell.
The emergent strand from the mold is composed of a solid shell enveloping
molten interior. The strand is further cooled in the secondary cooling stage over
a length of around 20 meters using a spray cooling. In this stage, the strand is
drawn over (5) rollers and bent from near-vertical orientation to the horizontal.
At the end it is completely solidified whence it is cut into discrete slabs.
One or more of the nozzles composing the SEN can get choked, creating a
directionally biased, disturbed flow within the mold leading to non-uniform shell
development and instability in the molten-steel level. The mass inflow rate from
SEN and the strand withdrawal rate based on roller speeds require a fine
balancing using a feedback level-controller, failing which, the steel level (called
"mold level") becomes unstable leading to poor shell quality and crack formation.
Uniformity in penetration of lubricant into the mold-shell gap is important, which
otherwise gets sticked, and culminate in tearing of formative shell including
'breakout' of the molten steel into the mold and onto machinery below. The
lubricant also serves as a heat-transfer medium between the strand and mold,
hence non-uniformity of penetration leads to non-uniform heat-transfer and
shell-thickness, which causes either quality problems or breakouts. If the
secondary cooling is too fast, an excessive thermal stress generates causing
cracks and clustering of inclusion materials, if the secondary cooling note is too
slow, a surface reheating occurs creating problems in the microstructure
including the properties of the steel. Thomas [3] provides an overview of the
possible problems in a continuous casting process.
Most of the steel plants worldwide use continuous parametric feedback to
provide information on individual parameters for the operators manning
continuous casting control 'cockpits' to take action to prevent any process drift.
Some attempts have been made by the steel industry solution providers, like for
e.g. Siemens-VAI (see [4], [5]), to integrate real-time data from different field
sources through process models and provide more synergized information on the
health of certain elements of casting to the operators. US Patent 6085183
describes a model-based real-time process control methodology for a generic
process where the model adapts to the drifting process conditions. US Patent
5854749 deals with quality assessment of steel product, but importantly, the
quality assessment is from the perspective of the customer's specifications, and
not the quality assessment in a running process, and thence its mapping into
quality of product. Japanese Patent 3856686 a method for prediction of quality
of process, which is substantially close to an online prediction. But the cited
invention only deals with the primary cooling or "casting" stage and not the
totality of casting and solidification. Overall, the continuous casting process being
quite complex, these systems cover only sub-processes of casting and as yet
there is no integrated monitoring and supervisory control system.
The problem is to convert the complexity of the continuous casting (CC, here
onwards) processes into an accurate system model that encapsulates all desired
functionalities while itself remaining free of complexity. Designing a system of
minimal complexity is essential from the perspective of the evolvability and
scalability of the system; refer Lehman's laws [7] of inverse relationship between
complexity and evolvability.
Considering the importance of the system performance and reliability flow from
the criticality of the CC process, a new process monitoring and control system
can be accepted and operationalized only if it satisfies stringent requirements on
these aspects. Designing a software modeling an extremely complex version of
the real world while satisfying high performance and reliability requirements is
the second major problem.
OBJECTS OF THE INVENTION
It is therefore an object of the invention to propose a Smartcast system for real
time monitoring and control of a continuous casting process, which provides the
operating stations with real time information in terms of quality indices of the
instantaneous condition of various sub-processes of primary and secondary
cooling extracted from the field data through processes of models, and
simultaneously provide the same information to various locations distributed
across the plant through a web connectivity.
Another objection of the invention is to propose a Smartcast system for real time
monitoring and control of a continuous casting process, which transforms the
real-time assessment of quality of process into an online assessment of quality of
the product during the production including corrections in the production
parameters if needed.
A further object of the invention is to propose a Smartcast system for real time
monitoring and control of a continuous casting process, which allows an
improvement of the quality of casting and solidification involving all the process
stages from SEN outflow to total freezing.
SUMMARY OF INVENTION
The continuous casting process in the steelmaking chain is a complex and critical
process with major impact on quality of the steel product and productivity of the
manufacturing chain. According to the invention, there is provided a SmartCast
system for real time monitoring and control of continuous casting process. This
system is enabled to provide in real time the quality of the process including the
underlying sub-processes, and further provides a real-time mapping of quality of
the process into quality of the product. To maximize the system reliability,
performance, scalability and evolvability in its representation of the real process
to higher levels of fidelity, a state of the art inter-modular architecture is
adapted, wherein the various system modules working in parallel are enabled to
inter-communicate only through database transactions between the module-
captive identical databases.
This invention does not deal with monitoring quality of individual processes and
their impact on product, but only with online control.
Thus, the invention discloses two major aspects, namely a system for monitoring
and control of the continuous casting process. This primarily focuses on the
process models to transform different sub-processes of the continuous casting
represented by respective field data into online indicators termed abnormality
indices for monitoring and control. In a second aspect, the invention provides a
highly reliable, scalable, evolvable system having a plurality of sub-systems
working in parallel based on an approach of coupling identical databases with
each sub-system and then interfacing between these sub-systems only through
transactions among the coupled databases.
The development of such an integrated monitoring and supervisory control
system is provided considering the totality of all the component sub-processes of
continuous casting. According to the data flow and process model hierarchy
adapted within the inventive Smartcast system, the field data for the major
component processes are captured in real time and then transformed through
models into degrees of (ab-) normality for the respective processes. These
models are further grouped into two major classes - for primary cooling or
"casting" and secondary cooling or "solidification". The component abnormality
indices are transformed through another level of models into the two major class
abnormalities, and then finally combined to yield a "strand abnormality index", all
in real time. While this step is performed as the monitoring and supervision
channel, another channel converts the component process abnormalities into
respective automated control actions.
The invention firstly configure the inventive system under a modular concept.
Two levels of modularity are followed - first, modularity of the major software
components. One module relates to the real-time software incorporating bulk of
the critical performance and reliability requirements, others relate to the control-
room HMI software, Web-HMI software, and software for simulation/playback
(simulation for designers and playback for operations). This is illustrated in fig. 3.
While the first three modules behave synchronously, the last one is
asynchronous from the rest. The next level of modularity relates to the object-
oriented structure of each of the software modules. One of the measures of
structural complexity of the software is the degree of coupling between the
modules (objects) - see e.g. Sangwan [8]. If modules are represented as nodes
and existence of coupling between two of them denoted as a connecting link,
then more complex the software, the more intertwining between the links. To
minimize complexity, care has been taken to minimize inter-modular coupling, as
explained in a later section.
The second major design aspect relates to the level of hierarchy between the
modules. All the modules are independent and work in parallel, in contrast to a
master slave relationship, where one master commands the activities of the rest
and controls the system as a whole. Parallel performance of independent
modules can be implemented either using inter-process communication via
message passing or message-oriented middleware, or using database oriented
architectures where DBMS-provided transaction processing and indexing facilities
can be used on a central database to communicate between processes, as in
Base One [9] or Lind [10]. The database-centric architecture leads to better
performance, reliability and scalability, see e.g. [11]. The Base One system
triggers activity for a module based on the status/updating of data in the
database. The present Smartcast system maintains identical (mirror) databases
in each module and cross-modular communication is ensured through DBMS

transactions. The activities of individual modules are triggered using localized
timers. This aspect of the design flows from the high-reliability-induced
"islanding" requirement of the real-time software module.
BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS
Figure 1 : overview of a continuous casting process.
Figure 2 : a process flow of a 'SmartCast' system of the invention.
Figure 3 : shows an architecture of a SmartCast system according to the
invention.
Figure 4 : shows a network block diagram of a SmartCast system of the
invention.
Figure 5 : a context diagram of all Real Time (RT) modules of a Smart cast
system
Figure 6 : a Generic state transition diagram for data receiving modules of smart
cast system.
Figure 7 : a Generic state transition diagram for data processing modules of
smart cast system
Figure 8 : a snapshot of the Operations visible on the Main Screen of a
SmartCast system.
Figures 9a & b : a flowchart of the main (driver) function of the SM1R module
and the corresponding data flow diagram (DFD) of the module, respectively.
Figures 10a & b : a flowchart of the main (driver) function of the SM2R module
and the corresponding data flow diagram (DFD) of the module, respectively.
Figures 11a & b: a flowchart of the main (driver) function of the SM3R module
and the corresponding data flow diagram (DFD) of the module, respectively.
Figures 12a & b : a flowchart of the main (driver) function of the SM4R module
and the corresponding data flow diagram (DFD) of the module, respectively.
Figures 13a & b : a flowchart of the main (driver) function of the SM5R module
and the corresponding data flow diagram (DFD) of the module, respectively.
Figures 14a & b : a flowchart of main (driver) function of the SM7R module and
the corresponding data flow diagram (DFD) of the module, respectively.
Figures 15a & b: a fowchart of main (driver) function of the SM8R module and
the corresponding data flow diagram (DFD) of the module, respectively.
Fig. 16 : an Overall flowchart of the SmartCast system of the invention which
constitutes a composition of the roles of the modules SM1R-SM8R.

DESCRIPTION OF THE PREFERRED EMBODIMENT
Fig. 1 shows an overview of a Continuous Casting process. The molten Steel
brought in a ladle is poured into the tundish (buffer to convert batch process to
continuous). The liquid metal passes from the tundish to a mold where the
primary cooling starts. A solid shell starts forming at the liquid boundary with a
copper mold, slowly thickening as the 'strand' moves downwards. The mold
oscillates along its vertical axis (fre ~ 4 Hz, amp ~ 1 cm) to prevent sticking and
lubricant penetration between the strand shell and the mold wall. The strand is
drawn out on rollers even as it is completely solidified by spray cooling in the
secondary cooling zone, and then finally cut into discrete slabs for downstream
batch processing.
The process and data flow chart of the inventive SmartCast system is shown in
fig. 2. This diagram shows each potential activity zone or hardware-software
combine as a rectangular box or nameplate. The boxes are of three types
depending on the possible intensity of technology development in that area.
They are arranged and interconnected such that, firstly, the movement from left
to right is in increasing order of technology development intensity, second, the
upward flow is along the direction of evolution of the system, and third, there is
a visible demarcation between the primary and secondary cooling stages.
Field data acquisition is shown at the lowest level. The first level of models
converts these data into various parametric indicators. The thin line between

primary (left) and secondary cooling is shown. At the next level overall casting
abnormality indices are calculated and displayed. In between real time control
signals are extracted for automatic control of specific parameters.
Figure 2 further defines the different process-flow channels, or sub-systems, of
the SmartCast system. At this stage the word "channel" rather than module or
sub-system is being used because it relates directly to a mapping of the real
continuous casting process and its component sub-processes into the system.
The channels are the following:
(1) Flow Biasedness Channel
(2) Crack Intensity Channel
(3) Level Channel
(4) Friction Channel
(5) Oscillation Channel
(6) Metallurgical Length Channel
(7) Strand Temperature Profile Channel.
The first five channels belong to a primary cooling area, the latter two to a
secondary cooling area. The primary cooling channels then coalesce to form the
primary cooling main channel, and the secondary channels into the secondary
cooling main channels. The primary and secondary channels finally combine to
yield the total system. Associated with each channel, and at each of the three
levels, is an index of abnormality. This index varies from 1 to 100 representing

variation from best to worst possible conditions of the corresponding (sub)
process. The abnormality indices are synthesized from bottom to top. The
weights assigned to the various components of primary abnormality index and
secondary abnormality index to obtain these values, when considered in any
pairs of two, vary inversely as the degree of overlap in the fundamental sensory
data and physical mechanisms used to extract the abnormality indices of the two
composing that pair.
Each of these channels is described in more detail below:
(1) Flow Biasedness Channel:
One or more of the nozzles composing the SEN can get choked, which
creates a directionally biased, disturbed flow within the mold leading to a
non-uniform shell development and instability in molten-steel level. Under
ideal conditions the flow rate and pattern from the nozzles should be
symmetric. This channel converts the degree of flow asymmetry or
"biasedness" into an index of abnormality called Flow Biasedness Index
which varies from 0 under ideal conditions to 100 at worst possible
conditions. The field inputs are the mold temperatures sensed at narrow
and corner thermocouples at the top and next-to-top layers. Under ideal
conditions, the temperatures at diagonally opposite corners (of the mold

rectangular section) at the same layer should be very close; as the
asymmetry increases, these temperatures start diverging. A
transformation function is established to convert these temperature
deviations into the stated flow biasedness index.
(2) Crack Intensity Channel:
When either a longitudinal or a transverse crack passes a thermocouple
embedded in the mold, the temperature time-history curve shows some
typical patterns termed as "crack signatures". These signatures allow easy
transformation into a crack intensity index. The preliminary assessment on
crack intensities is performed in the underlying breakout detection system
[12, 13] which analyzes these signatures from the perspective of
possibility of breakouts - and immediately reduces the speed of casting
automatically when the possibility of breakout is seen to be high. This
added aspect of the SmartCast system is shown in the middle of fig. 2, in
the zone contained under "real time control signals". The SmartCast
system is further enabled to convert these crack intensities from the
breakout-perspective to the quality-perspective, i.e. how much does an
identified crack affect the quality of the product. The Crack Intensity index
takes values from zero, under conditions of no crack detection, to 100
when a crack is seen to be sufficiently intense to lead to unambiguous
downgrading of a consequent slab.

(3) Level Channel:
The mass inflow rate from SEN and the strand withdrawal rate based on
roller speeds have to be finely balanced using a feedback level-controller,
failing which the steel level (called "mold level") becomes unstable leading
to poor shell quality and crack formation. Apart from these gross factors
causing mold level unbalance, there can be other factors like the pattern
of outflow from the SEN which can induce localized level fluctuations that
may convert to low-frequency standing waves at narrow ends, or the
periodic "quashing" of the solid shell encapsulating liquid in the upper
segments of secondary cooling which manifest again as low-frequency
fluctuations of the liquid level. Two sets of field inputs, first, from the top
layer of mold thermocouples, and second, from the mold-level sensor, are
integrated to yield the level abnormality index. This index is a function of
the rate of change of temperature from individual thermocouples
(providing local effect), from the totality of all thermocouples (providing
gross or global effect), and the rate of change of level as seen by the
sensor (which observes the mid-wide-face zone). Under ideal conditions
the index is zero, rising to 100 under worst possible conditions of level
instability.
(4) Friction Channel:
Uniformity in penetration of the lubricant into the mold-shell gap is

important, which otherwise get sickened and culminate in tearing of
formative shell and 'breakout' of the molten steel into the mold and onto
the machinery below; the lubricant also serves as a heat-transfer medium
between the strand and mold, hence non-uniformity of penetration leads
to non-uniform heat-transfer and shell-thickness, which causes either
quality problems or breakouts. Under identical powder and oscillation
conditions, degree of disuniformity in the lubricant penetration is directly
proportional to intensity of friction between the copper plates of the molf
and the moving strand. There can be no other measure of uniformity of
lubricant penetration. Furthermore, less the friction - due to better
uniformity of lubricant film - the smoother will be the strand surface
finish. Thus all these aspects, i.e. lubricant uniformity - level of friction -
level of smoothness - are interrelated (with cause-to-effect flowing from
left to right) when the oscillation (in a sense, the casting speed) and
powder are constant (i.e. under intransient casting conditions). The field
inputs for friction abnormality are hydraulic oscillator piston and rod
pressures and concurrent mold displacements, along with mold level,
width, oscillation stroke, frequency and waveform. These data are
converted into friction values. The friction levels are scaled against
corresponding smoothness levels obtained upon image processing of
strand surface interalia the CCD-camera images to evaluate best and
worst possible friction conditions, and the running friction value is
converted into an abnormality index that spans the range from zero (best
conditions) to 100 (worst).

(5) Oscillation Index:
Mold oscillation impacts the casting in two ways - directly by preventing
attachment between the mold face and strand (two bodies moving
oppositely with varying relative acceleration tend not to stick), and
indirectly by significantly facilitating the penetration of lubricant film down
into the mold-strand gap. In fact, apart from the mold powder
characteristics itself, the only other issue that influences is the lubricant
penetration - and hence all the consequential good and bad (in absence)
effects discussed is the nature of oscillation. Oscillation is also
accompanied by two undesirable effects - formation of transverse-crack-
like oscillation marks and a peak-friction corresponding to the time in the
cycle when the mold is moving upwards at the highest speed. Thus, one
would like to maximize the good effects and minimize the bad ones - a
multi-objective optimization problem. To generate an oscillation
abnormality index, first an "ideal oscillation schedule", i.e. ideal variation
of the oscillation stroke, frequency and wave-form with casting speed is
generated [2]. This is done using evolutionary algorithms driven by
objective functions which seek to maximize a mathematically defined
lubrication index and simultaneously minimize the negative strip time
(with constraints) and a peak friction factor. The running oscillation
stroke, frequency and wave-form are taken as inputs and converted to
running values of lubrication index, negative-strip time and peak-friction-
factor. The difference between these running indices and the ideal indices

generated from the ideal oscillation parameters using evolutionary
algorithms, are then mapped into a real time oscillation abnormality index,
varying from zero when the indices match exactly, to 100 when they
maximally deviate.
(6) Metallurgical Length Channel
The total length from the meniscus to final solidification point - the
metallurgical length - is a function of the strand secondary cooling pattern
along with the pressures exerted by the rollers. Rapid cooling leading to a
short metallurgical length is undesirable due to the resultant high thermal
stresses which couple with the bending stresses to exacerbate cracks and
other deformities. Also, the internal deposition of inclusions occurs in
clumps rather than in a continuous and diffused fashion. Thus it is
preferable to have a longer metallurgical length and regulate roller
pressures particularly in the lower segments to bring about smoother
solidification in the final phases. For a given grade, there is a targeted
metallurgical length based on material characteristics and the deviation
between running and the targeted value is mapped into an abnormality
index - zero when deviation is zero and 100 when the deviation
maximizes. The running metallurgical length is evaluated from computed
strand temperatures that take various field inputs like casting speed,
material composition, superheat, heat removed in primary cooling, width
and spray cooling water flow rates. The strand temperatures are

themselves evaluated in the underlying dynamic secondary cooling control
(DSCC) system which modulates water-flow rates in the cooling sprays
located in different secondary cooling zones - this again belongs to the
aspect of SmartCast related to "real time control signals" illustrated in the
middle of fig. 2.
(7) Strand Temperature Profile Channel:
For a given grade, there will be desired strand temperature profile which
is partly related to the desired metallurgical length but more importantly
to the minimization of thermal and bending stresses and formation of
grain structure and mechanical properties, the appropriate growth of shell
in the upper zones to prevent breakouts, uniform cooling in lower zones
around region of final solidification and also prevention of surface
reheating. The deviation between running profile and desired profile is
transformed into the strand temperature abnormality index as in any of
the above cases. The running profile itself is obtained from the underlying
DSCC system which numerically solves the unsteady heat transfer
equation in real time.
The second aspect of the invention, that is the system design and
configuration is based on principles of modularity under stringent
conditions of performance, reliability, scalability and evolvability.
The defining specifications of the designed system consist of the
following:

Conditions on reliability, performance, scalability and evolvability:
i. The hardware platform on which the system sits shall not crash or behave
in any unpredictable manner.
ii. The operating system and other auxiliary software shall not behave in any
unpredictable fashion.
iii. The operating system and other auxiliary software shall not load the CPU
differently at different points of time.
iv. The hardware elements for transferring input data to the system from
other neighboring systems shall not fail or behave in any unpredictable
manner.
v. The hardware and software design and implementation for transferring
input data to the system from other systems shall guarantee that the time
difference between arrivals of consecutive input streams shall be strictly
invariant with time.
vi. The hardware on which the database is loaded shall not crash or behave
in any unpredictable manner.
vii. The database shall be incorruptible and elapsed time for identical
operations shall be strictly invariant with time.
viii. The elapsed time for one complete processing cycle of the software shall
be invariant with time.
ix. The elapsed time for writing every output stream to the database, an
activity performed after completion of every single computational cycle,
shall remain invariant with time.
x. The total elapsed time for performing one computational cycle and writing
one output stream to the database, denoted as the 'operational cycle
time', shall remain invariant with time and shall be less than 75% of a
preset time At.
xi. The preset time At shall preferably be 1 second, and this selection can be
violated only after all other options for reducing the operational cycle time
have been exhausted.
xii. The system, including HMI and online database storage facilities, shall be
islanded from the company LAN with only the provision of directed point-
to-point access using appropriate switches and routers to prevent virus
attacks and any other unpredictable and/or inadvertent
behavior/interference.
xiii. The hardware platform on which the operations HMI is running shall not
behave in an unpredictable manner.
xiv. The hardware platform on which remote HMIs are running shall not
behave in an unpredictable manner.
xv. The system, particularly the software, shall be based on modular
principles to facilitate easy production and implementation of newer
versions consequent to modification and/or upgradation of specific
sections.
xvi. The system shall be configured to allow for scalability of functionality,
particularly for real-time control of certain casting parameters like mold
level and oscillation conditions.
xvii. The system shall be configured such that that any malfunctioning part can
be replaced with low component downtime while the rest of the system
external to the defective component remains available for normal
functioning.
The definition of the Control-Room Human Machine Interface important screens
are provided hereunder:
> The HMI displays both instantaneous digital values as well as time-
history plots of derived indices and casting parameters.
> The HMI consists of four screens, namely, Operations Screen, Strand
solidification viewing Screen, Reduction & temperature trends Screen
and Parametric trends Screen.
? The Operations Screen consists of four graphs and the rest
digital columnar displays. The parameter strand abnormality
index is shown in Graph A, primary cooling abnormality index in
Graph B, secondary cooling abnormality index in Graph C, and
casting speed and mold level are shown in Graph D. These
parameters are also shown in graphical form. Whereas graphical
form shows evolutionary data, digital form indicates the current
numerical value.
? Furthermore the Operations Screen digitally displays the
following parameters:
¦ Strand Abnormality Index
¦ Primary Cooling Abnormality Index
¦ Secondary Cooling Abnormality Index
¦ Oscillation Abnormality Index.
¦ Level Abnormality Index.
¦ Friction Abnormality Index.
¦ Flow Biasedness Index
¦ Crack Intensity Index.
¦ Metallurgical Length Abnormality Index.
¦ Secondary Temperature Abnormality Index.
¦ Crack Breakoutability
¦ Sticker Breakoutability
¦ Maximum Breakoutability
¦ Superheat.
¦ Casting speed.
¦ %C.
¦ Mould Level.
¦ Friction Value.
¦ Width
¦ Last Alarm Reason & Location.
? Graphs B & D have the facility to be merged by user choice.
When these are merged, a drop-down menu allows the user
to select choice of max 4 variables from the above list of 14,
which are displayed in different colors with same column
colors for y- axes legends, two on either side.
? Graph A does not have the facility to be modified from
default conditions. The values for red and blue lines for
downgrading and non-conditioning are selected by HMI
from a table which takes %C as one of the independent
parameters. Double-clicking on either line can also modify
the red and blue lines and then inserting a value in a pop-
up.
? The user is provided the facility to modify the time-
historization value from default -1500 to any value between
-500 to -1500 by double clicking on the x-axis of any
graph. All four vertically aligned graphs necessarily show
same x-axis range.
? When the strand abnormality index crosses the red line, a
red warning "RUNNING UNDER DOWNGRADING
CONDITIONS" gets displayed prominently on the screen.
There may be more conditions where such value-based
alarms are raised; hence the program section where these
are inserted should be easily accessible to, and modifiable
by, the developers.
? The Strand Solidification Viewing Screen shows the strand
thermograph from mold exit to the final solidification point
by providing a view of the solidifying strand from a viewing
point at 45 deg. angle between the front and side. This also
shows the strand internal temperature distribution on any
section selected by the cursor.
? The Reduction & Temperature Trends Screen contains four
graphs of which three show historical trends, from t = -
1500 to 0, of, first, the metallurgical length, second, the
calculated temperatures at two surface control points in the
upper & lower solidification zones, and third, the
temperatures from pyrometers at the two locations. The
metallurgical length graph shows a desired metallurgical
length line (or curve) in blue (based on the steel grade).
The temperature graph also shows desired temperature
lines (based on steel grade) in green/blue for the
lower/upper locations respectively. The fourth graph shows
the temperature profile of the strand.
? The Parametric Trends Screen displays nine historical trend
graphs, from t = -1500 to 0, for nine parameters shown
digitally in Screen A (Operations Screen). These are
parameter numbers 1 to 8, and parameter 14. There are
three columns with three graphs each. This screen enables
the operator to view the condition of all these parameters in
a historized fashion to analyze causes of any abnormalities
observed in the derived parameters displayed in Screen A.
The t = 0 position of these graphs denotes current instant
(real time). In all these cases the operator can select a time
range between -500 to -1500 seconds.
A fifth screen is also designed through which many of the tunable parameters of
SmartCast can be adjusted. More importantly, there is an additional facility for
flashing instructions to operators when certain combinations of abnormality
indices become large. These instructions are stored in two tables - Commenting
Information Table Primary or CITP, and Commenting Information Table
Secondary or CITS. This screen provides access to these two tables through
passwords.
A Web-based HMI facility is created which satisfies all specifications as in the
baseline control-room HMI, except for the facilities to modify tunable features. At
most five remote users can access Web-HMI concurrently across LAN.
An offline simulation facility is also created, firstly to analyze casting conditions
and derived quality indices at any time in the past up to a year & secondly to
analyze and verify effects of modifying any of the internal relationships that are
used to generate online quality indices. At most three remote users can access
playback software.
It is seen that there are major contradictions in the specifications that tend to
pull the system in different directions. It is apparent that the process functional
complexity that SmartCast is trying to represent, and the performance and
reliability conditions are pulling in opposite directions. As mentioned above
Smartcast takes data from different sources each having its own time cycle/scan
rate, and has to process them at its own time periods. There is a multiplicity of
models at few levels of processing. There are automated control outputs. There
is a CR-HMI, a Web-HMI and simulation activity to be supported. The number of
web-browsers that are activated concurrently is undefined. Yet the probability of
failure (crashing/freezing/ junk-outputs) of the system should be near-zero. Its
operational cycle time should remain invariant. It should remain easily evolvable-
supporting newer modules & functionality, and scalable.
The plurality of activities as shown in the form of numbered boxes are as under:
1. Acquisition of mold wall temperatures from embedded thermocouples.
Analog temperatures received in BDS server, where it is converted to
Digital and sent to SES server via serial link. Received in Module SM1R,
see Context diagram. Update frequency is 1 Hertz.
2. Mold level data received in module SM1R from field sensor via BDS server
@1 Hz.
3. Mold Oscillator (Hydraulic) piston and rod pressures received in module
SM2R from pressure transducers via field PLC @ 50 Hz.
4. Mold displacement received from Linear Variable Differential Transformer
(LVDT) at module SM2R via field PLC @ 50 Hz.
5. Roller Loads calculated from motor drive currents received from DSCC @
0.2 Hz frequency in module SM3R.
6. Strand temperatures obtained from DSCC server @ 0.2 Hz in module
SM3R.
7. Calculation of flow biasedness index in module SM5R.
8. Calculation of Thermal breakoutability indices (for sticker, crack and thin
shell) performed in BDS server and received in module SM2R, then sent to
module SM8R for further integration.
9. Calculation of Level Abnormality index in module SM7R.
10. Calculation of Friction Abnormality index in module SM5R.
11. Calculation of Oscillation Abnormality index in module SM4R.
12. Calculation of Metallurgical Length Abnormality index performed in DSCC
and received in module SM3R, then sent to module SM8R for further
integration.
13. Calculation of Strand Temperature Abnormality index performed in DSCC
and received in module SM3R, then sent to module SM8R for further
integration.
14. Calculation of Primary Cooling Abnormality index by integration of
parametric indices performed in module SM8R.
15. Digital output from BDS server to speed control PLC when Thermal
Breakoutability indices reach high values.
16. Level Instability control signals to be sent from SES server to speed
control PLC and slide gate control PLC. To be implemented in next version
of system.
17. Oscillation parametric control signals to be sent from SES server to
oscillation control PLC when functionality scale-up is implemented in next
version.
18. Roller Pressure Control signals to be sent from SES server to roller gap
control PLC when functionality scale up is implemented in next version.
19. Spray flow control serial output sent directly from DSCC server to
secondary cooling water spray control PLC.
20. Calculation of Secondary Cooling Abnormality index by integration of
parametric indices performed in module SM8R.
21. Calculation of Online Strand Quality Index by integration of upstream
indices. Displayed in online HMI. Used continuously for casting monitoring
and control. Used for real time strand quality assessment. Storedin
Database.
All the above aspects are synthesized to create the design for smartcast system.
The first aspect of this design is modularity. The broad classes of functionalities
are apportioned into major software modules, as illustrated in Fig. 3 shows that
the most critical performance and reliability specifications apply on the real-time
processing (RT) software module. Inputs on different parameters are obtained
from three major sources. These are, first, a server running the Breakout
Detection System or BDS [13] which provides serial inputs in 1 sec cycles. Next,
analog inputs on oscillator back pressures and mold displacements which are
scanned at a rate of 20 milliseconds. Third, from a server running the Dynamic
Secondary Cooling Control or DSCC software which provides serial inputs related
to secondary cooling at 5 sec cycles. The fusion of data incoming at different
rates into the operational time cycle of smartcast is discussed little later. Threads
running on both the BDS and the DSCC servers to send out data packets across
LAN to the RT module are based on localized timers. These are designed to be
robust, i.e. network outages or failure of either source or destination software
will not disrupt their functioning, which will resume automatically after revival of
normal conditions.
Encapsulated in an oval shape with dashed-line boundary. All serial contacts with
outside world through configured Layer 3 switch. Analog & Digital I/O with
outside directly at server backplane.
The RT module processes the incoming data and passes these to the other three
modules - CR-HMI, Web-HMI and Simulation. It may be noted that the
communication is one way, i.e. the RT module does not accept any information
from the other three (see Fig. 3). This is consistent with specification viii above,
i.e. RT module cycle time shall be unaffected by processing or conditions in other
modules. Furthermore, it is seen that the other modules do not communicate
among themselves. This is consistent with the higher-level requirement of
complexity minimization, as well as with another on reliability of each software
module - dependency is limited to modules of higher reliability, i.e. in this case
the RT module. It is also seen that all the smartcast modules are "islanded" from
external conditions, i.e. except for the inputs received by RT module strictly
according to conditions imposed by the requirement v mentioned above; no
other external inputs are received by any of the modules.
The issue relating to the mode of communication between the RT module and
the other Smartcast modules is quite important. The first point of note is that all
modules are at same level of hierarchy, i.e. perform in parallel. Communication
through message passing brings dependencies which would be undesirable from
the perspective of requirement nos. viii, x and xi. Thus a database-centric
approach is desirable, but the question arises - where will the database (DB) be
stored? If in the server on which runs the RT software, then presence of, and
operations on - like other modules reading data - a large DB can impact the
same requirements. If on a machine outside the server running the RT module,
then writing data across LAN will import external dependencies, again affecting
the same requirements.
The Smartcast design runs each of the synchronous modules on independent
platforms, and loads each such platform with identically-structured databases, as
illustrated in Fig. 4. While the DB on RT server stores only 2 days' data, the DBs
supported on CR-HMI PC and Web-HMI WS store a years' data. The three DBs
are synchronized with the time cycle of RT module, i.e. 1 sec (requirement xi),
using the 'Linked-Server' tool in SQL-server DBMS (used in Smartcast; other
DBMS use analogous tools) to update the two larger DBs with data written into
the RT-server DB by the RT software, all within the time cycle. It may be noted
that the activities of the non-RT synchronous modules are not triggered by the
changing status of their local DBs, but by localized timers triggering at the same
frequency as the RT-cycle. This maintains independent parallelism in a
synchronous framework.
Usage of database transactions between identically structured databases as the
means of communication between major software modules operating
synchronously is a key design choice and the fulcrum of the second of the two
inventions described in this patent. Before committing to this approach a series
of performance and reliability tests were conducted by the design team, the
successful outcome of which confirmed its efficacy.
The sequence of operations within a time cycle is the following. First, the RT
module receives processes and then writes data into the small-sized DB loaded
on its own server. It also writes an 'A' (denoting active) in the last column of
each data record (row). The RT server runs, apart from the RT module, also a
link-checking thread which simply verifies - within some milliseconds - that the
connection with the other two platforms is intact. If intact, it calls a stored
procedure in SQL-server which uses the Linked server tool to send the active"
data to the other two DBs, and then converts the 'A' in last column to 'P'
(processed). This ends a normal cycle. When connection is broken, the checking
thread doesn't call the stored procedure and data is not sent from RT-module DB
to other DBs, but the RT module continues to operate as normal and write
processed data into its' own server DB. Thus RT operations are completely
isolated from all external disturbances- fully consistent with requirements viii &
ix, and potentially consistent with x & xi . On revival of links, the stored
procedure that is called again checks for "A" in all records in the last n minutes
(configurable) of RT-module DB and sends across the chunk of qualifying data.
Figure-4 also shows that the Software architecture of fig.3. is an overlay on the
networked relationship with the three major blocks of the system namely the
SmartCast kernel, BDS and DSCC. The interfaces with the field are clearly shown.
All interactions are synchronized in process real time.
Figure 4 further shows how Smart Cast's serial connectivity to the outside world
is through a single configurable layer-3 switch, consistent with requirement xii.
Next the internal structure of the RT module of Smartcast is analyzed for
consistency with functional requirements and the need for minimal complexity
and evolvability. Fig. 5 shows context Diagram of a SmartCast real time module
running on server. The data sources for the software are shown in blue boxes at
left top and the data destination in green drum at right bottom. A timer generate
a trigger every 20 milliseconds which activates all modules for their respective
actions. State transition diagram of the 'data receiving modules' SM-1R, 2R and
3R is shown in generic form in Fig. 6 and of 'data processing modules' SM-4R,
5R, 7R and 8R is shown in Fig. 7. Note that SM8R also functions as a 'data-
sending module'. There are seven objects, the first three - SM1R to SM3R -
essentially read data from the three external sources, BDS, Analog and DSCC.
The next four objects - SM4R to SM8R - incorporate process models. With
reference to Fig. 2, object SM4R models oscillation, SM5R models thermal and
friction conditions, and SM7R models mold level. Secondary cooling models, as
well as models integrating indicators from the sub-processes, are incorporated in
object SM8R. SM8R also writes data into the DB.
It may be seen that the input-reading objects do not communicate with each
other; they only receive data from external sources and forward it to respective
processing objects. The data processing objects also do not cross communicate -
they only receive data from input objects and send their outputs to SM8R. Thus
object coupling and communication inter-twining is avoided minimizing
complexity. New input or processing objects can be inserted without changing
the architecture, consistent with requirement xvi.
Figure-6 shows Generic State Transition Diagram for 'data receiving modules'.
The transition from one state to another (rectangular boxes represent states) is
shown by arrows where the green text (above a horizontal line next to the
arrow) gives conditions for the transition and the blue text below shows action
taken. Note that 'action value on counf denotes the count value at which the
specific action sequence (in this case reading data and transferring to
appropriate destination modules) is triggered, essentially representing the
frequency of that action sequence.
Figures-7 shows Generic State Transition Diagram for 'data processing modules'.
The 'action value[i] on counf denotes the count value at which the specific
action-sequence[i] (where the actions are different for different values of I, while
the data transfer occurs at only 1 sec. intervals) is triggered, essentially
representing the frequency of that action sequence.
Figures 6 and 7 establish that for the two generic types of objects - input
readers and processors. A single trigger activated at the highest frequency
operation is used to time the process, and the manner of handling data at
different frequencies is self-explanatory in the figures. It may be noted that all
code in the RT module is written in standard C++ without reference to OS-
specific libraries, in response to a requirement of inter-operability across
Windows, Linux and Unix OS. However, no such constraint is imposed on the
other modules of Smartcast; CR-HMI and Simulation modules are written in
C#.NET, while Web-HMI is in ASP.NFT.
Figure 8 shows a snapshot of the Operations Main Screen of SmartCast in actual
running condition.
Module SM1R in the overall context of SmartCast in the Context Diagram of Fig.
5, and its functional contribution to the total functionality of SmarCast is
explained in the above detailed description of the Process Flow of Fig.-2
Module SM2R in the overall context of SmartCast in the Context Diagram of Fig-
5, and its functional contribution to the total functionality of SmartCast is
explained in the above detailed description of the Process Flow of Fig.-2.
Module SM3R in the overall context of SmartCast in the Context Diagram of Fig.-
5, and its functional contribution to the total functionality of SmartCast is
explained in the above detailed description of the Process Flow of Fig.-2.
Modules SM4R in the overall context of SmartCast in the Context Diagram of
Fig.-5, and its functional contribution to the total functionality of SmartCast in
explained in the above detailed description of the Process Flow of Fig.-2.
Module SM5R in the overall context of SmartCast in the Context Diagram of Fig.-
5, and its functional contribution to the total functionality of SmartCast is
explained in the above detailed description of the Process Flow of Fig.2.
Module SM7R in the overall context of SmartCast in the Context Diagram of
Fig.5, and its functional contribution to the total functionality of SmartCast is
explained in the above detailed description of the Process Flow of Fig. 2.
Module SM8R in the overall context of SmartCast in the Context Diagram of
Fig.5, and its functional contribution to the total functionality of SmartCast is
explained in the above detailed description of the Process Flow of Fig.2.
SmartCast - monitoring and control system for continuous casting process
Patent Authors:
Arya K. Bhattacharya, P.S. Mitra, P.S. Srinivas, Abhik Roy Chowdhury, Utpal
Nandi, S. Sistla, (all Automation, Tata Steel), J.B. Singh, H.B. Upadhyaya (all LD2
& Slab Caster, Tata Steel)

Other References
1. World Steel University website, continuous casting link:
http://www.steeluniversity.org/content/html/eng/default.asp?catid=27&paaeid=
2081271519. Oct 23, 2008.
2. A. K. Bhattacharya, S. Debjani, A. RoyChoudhury and J. Das,
"Optimization of Continuous Casting Mould Oscillation Parameters in Steel
Manufacturing Process using Genetic Algorithms", Proceedings of IEEE Congress
on Evolutionary Computation, CEC2007, also
http://ieeexplore.ieee.orQ/xpl/freeabs all.isp?tp=&amumber=4424992&isnumbe
r=4424446. Oct 23, 2008.
3. B.G. Thomas, "Modelling of the Continuous Casting of Steel -Past, Present and
Future", Metallurgical and Materials Transactions B, Vol 33B, No. 6, Dec 2002,
pp. 795-812.
4. Siemens-VAI web page:
http://www.industry.siemens.com/metals/enA/AI/automation continuous castin
a MoldExpert.htm?PIdent=1001&SIdent=16938iTIdent=1699&OIdent=1811,
Oct 23, 2008.
5. S. Lamdorfer, R. Ramler, C. Federspiel and K. Lehner, "Testing High-Reliability
Software for Continuous Casting Steel Plants - Experiences and Lessons Learned
from Siemens VAI", 33rd EUROMICRO Conference on Software Engineering and
Advanced Applications, Aug. 2007, also in
http://ieeexplore.ieee.ora/xpl/freeabs all.jsp?tp=&arnumber=43010878usnumbe
r=4301046 , Oct 23, 2008.
6. Tata Steel Automation Division website:
http://www.automationtatasteel.com/, Oct 23, 2008.
7. M.M. Lehman, J.F. Ramil, P.D. Wernick, D.E. Perry and W.M. Turski, "Metrics
and laws of software evolution-the nineties view", Proceedings of Fourth
International Software Metrics Symposium, Nov. 1997, also in
http://ieeexDlore.ieee.org/xpl/freeabs all.isp?arnumber=637156, Oct 23, 2008.
8. R.S. Sangwan, LP. Lin and CJ. Neill, "Structural Complexity in Architecture-
Centric Software Evolution", IEEE Computer Magazine, Oct 2008, pp. 96-99.
9. Base One International Corporation publication:
httD://www.boic.com/dbgrid.htm, Oct 23, 2008.
10. P. Lind and M. Alm, "A Database-Centric Virtual Chemistry System", J. of
Chemical Information Modelling, Vol. 46, No. 3, Mar 2006, pp. 1034-1039.
11. Wikipedia page on Database-Centric Architecture:
http://en.wikipedia.org/wiki/Database-centric architecture, Oct 23, 2008.
12. A. K. Bhattacharya, K. Chithra, S.S.V.S. Jatla and P.S.Srinivas, "Fuzzy
diagnostics system for breakout prevention in continuous casting of steel",
Proceedings of Fifth World Congress on Intelligent Control and Automation,
Hangzhou, China, June 2004, Vol. 4. pp. 3141-3145, also
http://ieeexplore.ieee.org/xpl/freeabs all.jsp?isnumber=29581
&amumber=13431008&count=2168index=66. Oct 23, 2008.
13. A. K. Bhattacharya, P.S. Srinivas, K. Chithra, S.S.V.S. Jatla and J. Das,
"Recognition of Fault Signature Patterns Using Fuzzy Logic for Prevention of
Breakdowns in Steel Continuous Casting Process", Lecture Notes in Computer
Science, Springer-Verlag, 3776, Dec. 2005, pp. 318-324, also
http://www.springerlink.com/content/p187n3722w2885v7/?p=f7974a892990423
681ff52029dddldd6&pi=3 , Oct 23, 2008.
WE CLAIM
1. Within the continuous casting process of the steel manufacturing chain
where molten steel is passed through a water-cooled near-vertically-
aligned lubricated mold of rectangular cross-section about a meter long to
emerge in the form of a continuous strand consisting of a solidified shell
encapsulating molten material,
this strand being further cooled to complete specification using water
sprays even as the orientation is changed to horizontal using rollers (see
Fig. 1) before being finally cut into discrete slabs,
wherein the above two steps are referred to as primary cooling or casting
stage, and secondary cooling or solidification stage, of the totality of the
continuous casting process,
the criticality of this process in the steel making chain flowing first from its
ability to potentially disrupt both upstream and downstream processes of
the chain under conditions of its own disruption and second to impact the
quality of the final product due to its own process quality deviation and
drift,
there is a great need for the development and establishment of a system
for monitoring in real time the quality of the process and to control the
process to return to ideal conditions either by direct intervention or
through operator instruction whenever it tends to deviate from these
conditions,
herein such a system has been designed, developed and implemented.
2. The said system in claim (1) additionally monitors in real time the quality
of the product from the perspective of the quality of the process insofar as
the latter impacts on the former which is the truest and pristine indication
of product quality independent of the commercial perspective of targeted
customer's priorities.
3. The quality of the process monitored in real time by the said system in
claim (1) is apportioned into real time indications of quality of component
sub-processes and any drifts in the former are appropriately reflected as
drifts in one or more of these component sub-processes and thence easily
translated into corrective control action either by means direct or
operator-suggestive.
4. The said system in claim (1) captures field data from multiple sensors
operating at different scan rates and fusses this data as input to a single
processing engine operating at a consistent frequency with high reliability
unaffected by the vagaries of the industrial process.
5. The said system in claim (1) monitors the process in real time not only at
the baseline level of the operator sitting in the process control room, but
also at the plant managers' level using web-based HMI (Human-Machine
Interface) accessible across the reach of plant-wide LAN (Local Area
Network).
6. The said system in claim (1) is composed of a server on which the main
processing engine runs at a frequency compatible with the continuous
casting process times to qualify for real time monitoring, a workstation to
support real-time web-HMI as well as offline even-based playback, and a
computer to support control-room HMI.
the three major software modules sitting on the three platforms, namely,
real time processing engine, web-HMI and control-room HMI, constituting
the three major sub-systems of the said system, operate parallel to each
other in a synchronous matter.
wherein the three sub-systems communicate among each other only
through DBMS (database management systems) database transactions
between identically-structured database each coupled with one of the sub-
systems and in no other manner whatsoever
the said database transactions originate only at the real-time server
database and are received at the other two database and no
communication occurs in the reverse direction
this manner of architecture ensuring the high performance, consistency
and scalability of the real-time monitoring sub-system and the reliability
and evolvability of the system in totality.

Within the continuous casting process of the steel manufacturing chain where
molten steel is passed through a water-cooled near-vertically-aligned lubricated
mold of rectangular cross-section about a meter long to emerge in the form of a
continuous strand consisting of a solidified shell encapsulating molten material,
this strand being further cooled to complete specification using water sprays
even as the orientation is changed to horizontal using rollers (see Fig. 1) before
being finally cut into discrete slabs, wherein the above two steps are referred to
as primary cooling or casting stage, and secondary cooling or solidification stage,
of the totality of the continuous casting process, the criticality of this process in
the steel making chain flowing first from its ability to potentially disrupt both
upstream and downstream processes of the chain under conditions of its own
disruption and second to impact the quality of the final product due to its own
process quality deviation and drift, there is a great need for the development and
establishment of a system for monitoring in real time the quality of the process
and to control the process to return to ideal conditions either by direct
intervention or through operator instruction whenever it tends to deviate from
these conditions, herein such a system has been designed, developed and
implemented.