FIELD OF THE INVENTION
The present invention relates to channeling prediction in blast furnace. More
particularly, the invention relates to an online method of acquiring and analysing
blast furnace data to pre-determine channeling occurrence.
BACKGROUND OF THE INVENTION
Nippon steel Technical Report No.94 of July 2006, describes a process for
visualization of shaft pressure variations and spatial changes caused by slipping
in the blast furnace, wherein stave temperature and shaft pressure data were
turned into images distributed in two dimensions. In addition, combining the two
dimensional distribution of secondarily processed data of changes in space and
time with the progress of operation data enables early detection of shaft
pressure fluctuations. It has been also found that there exists a relationship
between the cohesive zone root position, based on visual observations of
dimensional image of the change in stave temperature over time, and the
origins of shaft pressure fluctuations. The relationship constitutes an extracted
quantity of characteristic namely an independent ingredient from a two
dimensional image using an independent component analysis. This makes
determination of the spatial image and time series order easy by watching a
change in time series of independent ingredient. A "Large scale database Online
Modeling" based on the Just-in-Time modeling concept on blast furnace
operation is known, which has very complicated physical phenomena and strong

non-linear specific characteristics. The validity of the modeling method was.
confirmed by the study with blast furnace operation data, then the past similar
operation data have been searched and the prospective operation data have
been estimated very quickly and precisely.
It is also known that while taking into the blast furnace body shape into
consideration, the furnace outer profile can be projected onto two-dimensional
planes in vertical and horizontal directions to lay out the measured values
obtained by sensors on the two-dimensional planes in such a manner that they
precisely corresponded to the three-dimensional sensor installation position
information to prepare equal-value, contour and vector diagrams of the
measurement data.
Since many of the blast furnace sensors are installed at equal intervals, the prior
art teaches an equal-value curve retrieval algorithm applicable to any sensor
position. For any region devoid of sensors, a virtual grid appropriate to the
required space resolution was set and actual measurement data obtained in its
neighborhood was subjected to spatial interpolation using the actual three-
dimensional Eculidean distance between the virtual grid and its nearest sensor so
as to interpolate the value on the virtual grid.
An example of two-dimensional visualization of stave temperature as described in
prior art non-patent literature, is shown in Fig.1. In the figure, the horizontal axis

represents the furnace azimuthal angle, the vertical axis represents the furnace
height, and each asterisk (*) indicates' the position of sensor. By continuously
updating the visual information, it is possible to quantify and visualize non-steady
phenomena of stave temperature distribution in the furnace in an animated
form.
An example of a known two-dimensional visualization of shaft pressure is shown
in Fig. 2. In the figure, each arrow indicates a spatial variation vector of shaft
pressure, that is, a pressure drop. It can be seen that compared with numerical
data, the visual image greatly facilitates understanding of the change in shaft
pressure, such as the point of occurrence of a pressure fluctuation.
The shaft pressure sensitively reflects the changes in packing structure and gas
flow in the furnace. In order to make an in-depth analysis of the pressure
information, the spatial differential of shaft pressure, or the shaft pressure drop,
has been monitored. The newly developed system employs a visual image of the
spatial differential vector of shaft pressure, which is a generalized pressure drop,
in place of the pressure drop.
As secondary processing of the visualized image of the shaft pressure, the
authors defined the spatial differential vector of shaft pressure in a three-
dimensional space that takes into account the bottle furnace body characteristic

of the blast furnace and visualized it on two -' dimensional, planes projected in the
vertical and horizontal directions (Fig. 3).
Spatial variation can be defined for stave temperature as well.
According to another known method of estimation using a two-dimensional visual
image, there are two possible approaches - one using the stave temperature
distribution and the other using the shaft pressure distribution.
In the method using stave temperature distribution, the cohesive zone is
estimated from the time differential of stave temperature (Fig. 4). In the
cohesive zone, the gas permeability resistance is so large that the flow of gas
passing through the cohesive zone does not always become a plug flow and
hence, partial out-gassing occurs frequently. As a result, at the position
corresponding to the root of the cohesive zone, the stave temperature is
considered to change locally. Therefore, it can be assumed that the region in
which the stave temperature variation per unit time is conspicuously large is the
root of the cohesive zone.
In the method using shaft pressure distribution according to the prior art, the
cohesive zone is estimated (see Fig. 5) from the spatial differential vector of the
shaft pressure. Ordinarily, the pressure drop in the cohesive zone is about twice
that in the shaft. In the shaft, even when the permeability decreases due to a
restrained central flow, powder accumulation, etc., the gaseous flow

considerably diverges and becomes uniform. In the cohesive zone root, by
contrast, the gas hardly diverges laterally because of a large permeability
resistance. This is considered to cause an out-gassing toward the furnace top
and an abnormally large pressure drop. In other words, the pressure rise due to
out-gassing plays the role of a sensor that reveals the cohesive zone. Thus, there
is the possibility that the root.of the cohesive zone can be determined from the
position at which the pressure rises abnormally.
From the viewpoint of reducing the environmental pollution and reserve of fossil
fuel, using a lesser amount of reducing agents. Techniques for improving the
rate of reducing gas utilization by enhancing one reduction and coke reaction are
also known, which inter alia decrease the need for calorific power by minimizing
Si in pig iron. In order to actually decrease the input rate of the reducing agents
however, it is essential to continually operate under stable conditions that are
near the critical thermal condition. The phenomena occurring during non-
stationary operation to ensure stable operation, are as under:
Upon the blast Furnace operation, ascending gases passing through the packed
bed are sometimes disconnected and radically escaped from the furnace without
any heat exchange and reaction to the burden. This phenomenon is known as
Channeling. When it happens the temperature of the top furnace rises and gas
purifying facilitates can be damaged irrevocably. Because of appreciable losses of

high temperature reducing gas from the cohesive zone and the proper
preparation of the iron ore in the other regions, there is a serious deterioration
in the blast furnace performance- production rate, coke, oke rate, and reduction
of hot metal temperature, quality and dust losses.
It is further known that various non-stationary phenomena occur during blast-
furnace operations, including routine fluctuations due to raw material component
variations, daily thermal fluctuations, and burden descending fluctuations, with
large-scale fluctuations resulting in operations disorder. These phenomena
usually last for several minutes to several hours, varying from case to case, and
can be caused by abnormal heat transfer, reaction, gas flow, or solid flow.
Specifically, actual abnormal phenomena can be in the form of a sharp drop in
hot metal temperature, fluctuations in furnace top gas component composition, a
slip or drop or drop or hanging of burden, shaft pressure fluctuations, or
abnormal gas flow such as fluidization, or channeling.
These phenomena are often interrelated, and many analyses have been
performed in the past. Various physical models have been developed for
monitoring and control, of these non-stationary phenomena and three-
dimensional non-stationary physical models in particular, have been developed in
the recent years in an attempt to quantitatively evaluate and analyze non-
stationary phenomena (dynamic behaviors and characteristics).

Quantitative analysis and control techniques were also developed by applying Al.
Despite the progress of such physical models and Al, and the development in the
varieties of sensors and probes, however, the grasp and prediction of non-
stationary phenomena in the actual blast furnaces are largely due to the
experience and skills of on-side operators.
A great deal of improvement in computing capacity has been combined with the
spread of low cost hardware and database systems capable of storing a large
volume of digital data. This has allowed the enhancement and prevalence of
digital image processing technologies, and enabled sampling to be performed in
a very short time. It has also enabled mass-store blast furnace operation data
for. a long period to be achieved, and allowed reduction of blast furnace
operation data to be processed into image information.
JP 2002194405 describes a method and a device to correctly monitor the
operational condition of a blast furnace, and to predict any operational
abnormality such as channeling. According to the invention, the distribution
condition of the measurement data collected by a plurality of sensors installed on
a blast furnace facility is arranged on the two-dimensional plane or three-
dimensional space reflecting the installation positions of each sensor on the last
furnace facility, the pattern formed by the measurement data using isopleths or
the like, and the pattern or the characteristics information of the pattern is
operated by the image processing.; The operational condition of the blast

furnace is monitored by comparing the obtained pattern or characteristics
information of the pattern with the preset pattern or characteristics information
of the pattern, the pattern or characteristics information of the pattern is
updated corresponding to the temporal transition of the measurement data, and
any operational abnormality of the blast furnace such as channeling is predicted
by comparing the temporal transition with the transition condition of the present
pattern or characteristics information of the pattern.
OBJECTS OF THE INVENTION
It Is therefore an object of the invention to propose an online method of
acquiring and analyzing blast furnace data to pre-determine channeling
occurrence which eliminates the disadvantages of prior art.
Another object of the invention is to propose an online method of acquiring and
analyzing blast furnace data to pre-determine channeling occurrence which blast
furnace parameters namely, stave temperature, shaft pressure, probe
temperature above burden, exit gas composition, top pressure, production rate,
and coke rate are collected from a data acquisition system through a plurality of
sensors.

SUMMARY OF THE INVENTION
According to the invention, the values of following parameters are collected.from
a Data Acquisition System of a Blast Furnace (BF) for this invention:
1. Stave rib thermocouple readings.
2. Shaft pressure readings
3. Above burden probe temperature
4. Exit gas composition
5. Production rate, Coke rate, Top pressure.
These data were collected per one minute basis during stable and unstable
operation of the furnace.
Channeling can be predicted by analyzing the collected data as under:
- converting into images distributed in two dimensions, the stave
temperature and shaft pressure data collected from a plurality of sensors
spatially located circumferentially and vertically in the blast furnace;
- viewing and quantifying the data relating to shaft pressure variations and
spatial charges of associated parameters caused by Channeling in the
blast furnace;
- deriving a relationship between a cohesive zone root position, and the
origins of shaft pressure fluctuations, based on a temporal evolution of
two-dimensional images of the stave temperature.

The relationship can.be understood better with.the help, of other representations
of data and these are as follows:
• The wall surface of a blast furnace is unfolded and projected onto a 2 D
plane where the ordinate and abscissa represent the height and
circumference respectively of the furnace. This particular image was
developed by making a contour plot of interpolated values taken from that
pressure sensors located across the shell of the BF. The arrow diagram is
presented in he same display which represents the spatial rate of change
of the pressure where the arrow size indicates the magnitude of difference
and the direction shows the change pattern.
• It is assumed that during instability, the time fluctuation of the stave
temperature is greater near the region where there is local and selective
gas passage than in the other regions. For this purpose, the time
differential ratio of stave temperature is plotted. The stave thermal
sensors do no directly measure the temperature inside the furnace but
measure that of the staves. In consideration of the influence of the heat
capacity of a furnace wall including the staves, a time differential ratio
based on a forgetting type moving average standard, and using an
exponential function as the weighing coefficient, is introduced.
• The variation of different parameters over time is represented by
displaying data form last six hours.

BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS
FIG. 1 - shows a two-dimensional image of stave temperature
collected in a prior art method.
FIG. 2 - shows a two-dimensional image of shaft temperature
obtained in a prior art method of Fig. 1.
FIG. 3 - shows a spatial distribution of spatial differential vector of
shaft pressure in a prior art method of Fig. 1.
FIG. 4 - shows time differential of stave temperature according to
prior art
FIG. 5 - shows estimation result of root of cohesive zone by time
differential rates of stave temperature according to prior art.
FIG. 6 - shows a schematic diagram depicting a process flow chart
for channel prediction in blast furnace according to the
present invention.
FIG."7 - shows locations of plurality of sensors disposed in the blast
furnace to acquire data relating to furnace parameters.

FIG. 8 - shows measurement of data plotted on the wall surface of
an unfolded and projected onto a 2D - plane according to
the invention.
FIG. 9(a) & 9(b) - respectively shows distribution of shaft pressure drop
and stave temperature according to the invention.
PETAIL PESCRIPTION OF THE INVENTION
Figure 6 shows that, the various temperature and pressure data from the
sensors being logged into the programmable logic controllers (PLCs) called level
1 automation. From these PLCs, the data is fetched in Level 2 servers through
the use of OPC and inturn the data is fetched into channeling prediction server
and this data is then processed and displayed in the display unit to the operator.
Figure 7 shows the relative position of various thermocouple and pressure probes
mounted on the blast furnace wall when the blast furnace is radially opened and
represented on a 2 dimensional surface. The elevation of Bl, B2, SI, S2, S3, S4
is indicated below. Elevation of zone represents the Height from the ground
and L for any zone shows the difference between upper Zone and corresponding
zone.


In figure 8, the wall surface of a blast furnace is unfolded and projected onto a
2 D plane where the ordinate and abscissa represent the height and
circumference respectively of the furnace. The measurement data of each sensor
is plotted on said plane at a position precisely corresponding to the position of
the sensor from which the data, came. A virtual grid reflecting required spatial
resolution is drawn covering the areas of the plane where no sensors are
provided, and the value of a grid point is calculated by spatial interpolation based
on the measurement data of the sensors located near the point and the actual
and. Euclidean distances from these sensors. The five rows in the first screen will
be displayed in the order of Shaft pressure, Spatial differential rates of shaft
pressure, time differential rates of shaft pressure, stave temperature, time
differential rate of stave temperature.

figure 8 further highlights that the shaft pressure sharply rose at about 14:42
hours and the slip occurred a about 15:08 hours on the same day,.and it is
further noticed from the shaft pressure distribution that the furnace bottom
pressure sharply rose between seventeen thirty and nineteen hours in the 0-
degree direction. The figure 8 also indicate that although the furnace bottom
pressure rise subsided after the slip, the circumferential pressure distribution
imbalance remains, as the cause of the pressure variation is not completely
removed. The stave temperature rose denoting a relationship with the shaft
pressure rise. Also, it is seen that the shaft pressure distribution significantly
varied around seventeen thirty hours. It can therefore be stated that a bubbling
zone occurred in a certain region within the furnace at about seventeen thirty
hours, and it materially affected the stock condition. When the gas component
data variations are analyzed.
From such a point of view, a CO increase after 5 o'clock, which is a sign of
bubbling. Close examination of Fig. 9(a), (9b) reveals two points. Firstly, a zone
where the spatial differential rates of shaft pressure is abnormally high (the
pressure gradient is larger than in the neighboring area), and secondly, a zone
where the spatial differential rates of shaft pressure is abnormally low (the
pressure gradient is smaller than in the neighboring area). At the same time, the
burden decent stagnates. The indications are that the cancellation of the burden
descent stagnation, that is, the occurrence of the slip or drop, causes shaft
fluctuations to cancel the pressure gradient.

Accordingly, the fluctuations in space is a furnace in its height and
circumferential directions in the course of time is predicted by viewing the two-
dimensionally save temperatures, pressure distribution, and their analyzed
values. Furthermore, the shaft pressure variations, the visualization permits
faster and surer detection of these than using furnace top sensors only.
The Channeling can be controlled by controlling Coke consumption (Rate) and
controlling burden particle size. This can be largely accomplished by screening
the Coke and other blast furnace feed stock. The annular distribution of the
burden components is avoided through alteration of charging sequence of the
Ore and Coke.
The methods of representation of data provided have proven to be an effective
means of predicting non-stationary behavior in the blast fur4naces in general
and channeling in particular.

We Claim:
1. An online method of acquiring and analyzing blast furnace data to pre- .
determine channeling occurrence comprising the steps of:
- converting into images distributed in two dimensions, the stave
temperature and shaft pressure data collected from a plurality of sensors
spatially located circumferentially and vertically in the blast furnace;
- viewing and quantifying the data relating to shaft pressure variations and
spatial changes of associated parameters caused by Channeling in the
blast furnace;
- deriving a relationship between a cohesive zone root position, and the
origins of shaft pressure fluctuations, based on a temporal evolution of
two-dimensional images of the stave temperature.
2. The method as claimed in claim 1, wherein the step of deriving the
relationship between a cohesive zone root position and the origin of
shaft pressure fluctuations comprises:

- generating an 2D-image of the wall surface to the blast furnace where the
ordinate and abscissa represents the height and circumference of the
furnace, the contour plot developed constitutes interpolated pressure
sensor values;
- generating an arrow diagram on said 2D-image which represents the
spatial rate of change of the pressure where the arrow size indicates the
magnitude of difference and the direction exhibiting change pattern;
- plotting a time differential ratio of the stave temperature based on a
moving average standard and an exponential function being introduced
as a weightage co-efficient; and
- displaying variation in the values of associated furnace parameters for
temporal evaluation of said image.
3. An online method of acquiring and analyzing blast furnace data to pre-
determine channeling occurrence substantially as herein described and
illustrated with reference to the accompanying drawings.

ABSTRACT

During the blast furnace operation, ascending gases passing through the
packed bed are sometimes disconnected and radically escape from the
furnace without any heat exchange and reaction to the burden. This
phenomenon is known as Channeling. When it happens the ascending gas
remains largely unutilized leading to malfunctioning of the furnace operation,
which is manifested in terms of poor blast furnace performances like poor
production rate, coke rate, decrease in hot metal temperature and quality and
dust losses. According to the present invention, a Visual information
Technique for Channeling Prediction prior to its occurrence in a blast furnace
has been developed based on analysis of the data collected from Steel plants.
Stave temperature and shaft pressure data have been collected by a plurality
of sensors spatially located circumferentially and vertically in the blast
furnace, the acquired data converted into images distributed in two
dimensions. Shaft pressure variations and spatial changes of associated
parameters caused by Channeling in the blast furnace have been quantified.
A relationship between a cohesive zone root position, and the origins of shaft
pressure fluctuations, has been derived based on the temporal evolution of
two-dimensional images of the stave temperature. Finally, the two
dimensional distribution of secondarily processed image data representing
changes in space and time have been combined with the progress of
operation data, for an early detection of shaft pressure fluctuations and
Channeling.