FIELD OF INVENTION
The present invention relates to a Fuzzy logic based system and method of monitoring
level of liquid metal and slag inside the hearth of a blast furnace. More particularly, the
invention relates to the implementation of the fuzzy logic technology to solve the
complex problem of estimation of levels of liquids inside the blast furnace hearth.
BACKGROUND OF THE INVENTIONS AND PRIOR ART
The levels of liquid metal and slag in the hearth are important control parameters for
the stable operation of the blast furnace. The knowledge of liquid metal and slag:
inflows and outflows and the hearth dimensions are the pre requisites for level
calculations. Liquid metal and slag inflows are obtained through Mass balance or
oxygen balance calculations. Liquid metal and slag inflows are obtained through the
radars that sense the levels of liquid metal in the torpedo ladles and the drum torque of
the slag granulation plant respectively. However, all the above calculations are prone to
errors due to wear and tear of torpedo ladle linings, disturbances and drifts in drum
alignments and calibrations, disturbances in furnace resulting in wrong inflow
calculations (since the inflow calculations are essentially steady state calculations) and
soon.

Further the hearth dimensions change dynamically due to wear and tear there by
rendering the original hearth dimensions incorrect. Once these inputs are used for
calculations the noise is propagated to the estimated level in the form of integration
errors. Typically, Kalman Filtering techniques have been used to filter the noise from
measurement data. However such techniques require a reasonably good mathematical
model of the noise which is quite difficult to obtain in terms of accuracy. Although
pother techniques like measurement of electrical potential difference between the top
and the bottom levels of the hearth do give an estimate of the levels, they too are
prone to drift due to a number of factors. The propose method eliminates all the
aforesaid complications and provides an effective solution for eliminating the integration
errors or drifts in the estimated liquid levels. Conventional approaches to estimate the
levels of liquid metal and slag inside the hearth have suffered from disadvantages of
measurement noise and dynamically changing hearth conditions. As a result, the liquid
level estimates are quite erroneous. In order to mitigate the problem of measurement
noise, the approach is to somehow create an effective filter to filter out the
measurement noise. Although, several techniques for noise filtration exist in practice,
they are either inefficient or difficult to implement. In the proposed scheme, the need
for noise filtration is eliminated by using short term and medium term averages of the
differential flows as inputs to a Fuzzy Rule Based inference system. A system is
developed for the measurement of level of liquid metal and slag inside the hearth of
Blast Furnace using fuzzy logic approach. The Fuzzy system translates the inputs to

levels of hot metal and slag in standard units, using a predefined set of rules and
membership functions.
OBJECTS OF THE INVENTION
It is therefore an object of the invention to propose a Fuzzy logic based system and
method of monitoring level of liquid metal and slag inside the hearth of a blast furnace
which is capable of solving the complex problem of estimation of levels of liquid metal
and slag inside the blast furnace hearth.
Another object of the invention is to propose a Fuzzy logic based system and method of
monitoring level of liquid metal and slag inside the hearth of a blast furnace which is
able to provide an effective solution to avoid the propagated noise to the estimated
level in the form of integration errors or drifts in the estimated liquid levels.
BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS
Fig.1 - shows a schematic diagram of the system and method
Fig.2 - shows Fuzzy Logic flow diagram

DETAILED DESCRIPTION OF A PREFERRED EMBODIMENT OF THE
INVENTION
The blast furnace is a counter current reactor which is continuously charged with the
raw materials from its top and the liquid products are tapped from the hearth in the
bottom. Typically it takes around six to eight hours for the charged material to become
liquid metal. Thus a time period larger than eight hours is considered as a long time
interval. The proposed method is based on the hypothesis that in a very long time zone
the cumulative differential flows of liquid metal and slag into the blast furnace hearth
have to be zero or in other words, whatever material has flown in must have flown out.
However in the short and medium terms this may not be true, which is indicative of the
fact that the short term and medium term flows reflect the net accumulated liquid in
the hearth. In order to prove this hypothesis a study of the distribution of differential
flow of liquid metal was undertaken for various time intervals. It was found from tr.d
distribution that the spread of data rapidly diminished with increasing time intervals.
Based on this analysis of differential flow data the short and medium zone time periods
are decided. These flows give an "idea" of the current level of liquids in the hearth, but
not the exact value.
The System
A novel system which makes use of the fuzzy logic technology to solve the complex
problem of estimation of levels of liquids (liquid metal and slag) inside the blast furnace
hearth.

The system comprises of the following components,
a. Sensors installed in the blast furnace to measure the following parameters:
i. Blast volume
ii. Blast humidity
iii. Oxygen enrichment
iv. Steam injection
v. Height of liquid in the torpedo ladle through the radar sensor
vi. Slag outflow rate
b. A server in which the sensory data is collected from the plant data base in real
time through an interface program which communicates with the plant data base
using the OPC communication protocol.
c. A program hosted in the server which fetches the sensor inputs from the
database server and processes them to compute the following parameters,
i. Liquid Metal Inflow
ii. Liquid Metal Outflow
iii. Slag Inflow
iv. Slag Outflow
d. The fuzzy logic toolbox of the Matlab™ residing in the server which takes the
differential flow of liquid metal and slag from the parameters as computed in Y
and translates the differential flows to height using a set of predefined rules and
memberships.

e. A Matlab program which stores the outputs of the Fuzzy Logic Toolbox in the
database.
f. The Human Machine Interface (HMI) program which fetches the stored results
from the database and displays to the end user (clients) in the required format.
The complete schematic of the system is shown in Figure I.
The differential flows of liquid metal and slag are calculated by the program residing in
the server using the sensor outputs available in the database as mentioned in the
previous section. The program stores the short term and the medium term cumulative
differential flows in the data base. The fuzzy logic tool box fetches this data from the
database and estimates the levels of liquids in the hearth through a set of prefixed
rules, membership functions and inference methods. The fuzzy logic methodology is
explained in the subsequent section.
The Fuzzy Logic Methodology
Fuzzy logic is a popular technology which is commonly used in the industry to handle
typical problems, where it is not possible to obtain even a reasonably accurate
mathematical model of the plant under consideration. It is essentially an artificial
intelligence technique which tries to mimic the human way of thinking by establishing a
mathematical representation to determine the degree of belongingness of a particular
parameter to a particular class or category. This is achieved by creating membership
functions for a parameter using linguistic variables. For example the variable
temperature can have its membership in any arbitrary membership functions like hot,

very hot, cold, very cold and so on. Figure II shows the flow diagram of the fuzzy
inference system. Each of its components is explained in Fig. II.
Fuzzification
In any fuzzy inference system all the input and the output variables of the system under
consideration have to be first fuzzified before they could be used. Fuzzification means
definition of the input and the output variable membership functions. In this case the
variables are
System Inputs: Short Term Cumulative Differential Flow (ST), Medium term
Cumulative Differential Flow (MT) for liquid metal and slag in cubic meters per unit
time.
System Output: Liquid Metal Level in meters, Slag Level in meters
The differential flows have been categorized into the following five membership
functions,
Negative Large (NL), Negative Small (NS), Zero (Z) Positive Small (PS) and Positive
Large (PL).
The output which is the height or level (HT) of liquid in the hearth is categorized into
the following five membership functions.
Very Low (VL), Low (L), Medium (M), High (H) and Very High (VH)
Thus the inputs to the fuzzification block are the input variables themselves and the
output is the belongingness or degree of association of each of the inputs to all the
defined memberships.

Implication
The heart of the fuzzy inference engine is the set of rules which decide the relationship
between input variables and output variables. A typical rule looks like
If ST is NL and MT is NL then HT is VL [weight 0.9]
A table consisting of all the possible combinations of the input memberships called the
Fuzzy Associative Map (FAM) is formed giving the corresponding output membership
and the weights for each rule. Formulation of the rules is purely through the experience
and the process knowledge of the designer.
These degrees of memberships fire one or more rules defined in the FAM. The net
effect of the inputs for each rule is evaluated suing the standard function of "AND"
(min()). Once the net effect of input for each rule is obtained the effect is passed on to
the output (implication) using the "max" method. Thus the input to the implication
block is a single number which gives the combined impact of the system inputs and the
output of this block is an output membership function which has been reshaped by the
strength of the inputs and the weight of each rule. Implication is done for every rule
which is fired.
Aggregation
For any given input data set {ST, MT}, one or more rules are fired and each rule after
implication returns a reshaped output membership function. Aggregation is the process
of obtaining the combined output membership function due to all the rules which are
fixed. Typical methods used for aggregating the fuzzy sets due to each rule are the
sum, probor or max. Here the max method has been considered for aggregation. Thus

the inputs to the aggregation block are the output fuzzy sets corresponding to each rule
and the output is a combined single output fuzzy set or function.
Defuzzification
The final aim of this system is to obtain a measure of the liquid metal and slag heights.
As mentioned earlier the output membership function (height of liquid) has been
defined over a range of numbers from zero to the maximum height of the blast furnace
hearth. The final aggregated fuzzy set or function is defined over this range only. This
function has to be converted to the corresponding height in meters. Several well
defined methods are available to covert the fuzzy set to the corresponding crisp output
value like the centroid, bisector, middle of maximum (the average of the maximum
value of the output set), largest of maximum, and smallest of maximum. Here the
centroid method has been used for de-fuzzification.
Thus the input to the de-fuzzification block is the aggregated fuzzy output set and the
output of this block is the crisp output value, which in this case is the height of the
liquid metal and slag inside the furnace hearth.
Conclusion
The application of fuzzy logic is liquid level estimation of the hearth has eliminated the
need for complicated filtering schemes, knowledge of the exact hearth dimensions, the
need to incorporate new hardware in the hearth [2][3] and hence the integration error
associated with the estimated liquid levels in the hearth. This system translates the
"idea" of the levels to actual levels of the liquids in the hearth through a set of rules,
membership functions and inference methods. This method of level estimation

eliminates the integration error which is typical of the conventional methods of
estimating liquid level, since no mathematical integration is performed. Instead the
fuzzy engine interprets the height in real time using the presently available medium
term and short term differential flows. However, this methodology requires initial tuning
of the fuzzy rules and membership functions to suit a particular hearth dimension. The
schematic diagram of the system is shown in Fig.l.
This method by which the liquid levels in the Blast Furnace hearth can be determined
by the short term and medium term differential flows using a fuzzy rule based
approach, directly eliminates,
a. The need to design a complicated filter for measurement noise elimination.
b. The need to know the exact hearth dimensions, but rather an "idea" of the size
of the hearth is insufficient.
It is a method for estimating the levels of liquid metal and slag inside the hearth based
on Fuzzy Rule based techniques which is independent of the dimensions of the hearth.
The fuzzy inference engine takes the short term and the medium term differential flows
as the inputs and uses them to interpret the levels of liquid metal and slag in the hearth
based on predefined set of fuzzy rules, membership functions and inference methods.
The application of fuzzy logic in liquid level estimation of the hearth has eliminated the
need for complicated filtering schemes, knowledge of the exact hearth dimensions, the
need to incorporate new hardware in the hearth and hence the integration errors
associated with the estimated liquid levels in the hearth.

WE CLAIM
1. Fuzzy logic based method of monitoring level of liquid metal and slag inside the
hearth of a blast furnace comprising:
analyzing of differential flow data to decide the short and medium time periods
and to establish an idea of the current level of liquids in the hearth;
employing sensor outputs of the plant, stored in the database to calculate the
differential flows of liquid metal and slag by program residing in the server;
storing the short and the medium term, cumulative differential flows data;
characterised in that,
the fuzzy logic tool box fetches the data from the database and estimates the
levels of liquids in the hearth through a set of prefixed rules, membership
functions and inference methods wherein all the input and output variables of
the fuzzy inference method is first fuzzified before they are implemented.
2. A method as claimed in claim 1, wherein the inputs are short term cumulative
differential flow (ST), medium term cumulative differential flow (MT) for liquid
metal and slag in cubic meters per unit time.
3. A method as claimed in claim 1 and 2, wherein the outputs are liquid metal level
in meters and slag level in meters.

4. A method as claimed in claims 1 to 3, wherein the differential flows are
categorized into five membership functions as negative large (NL), negative
small (NS), zero (Z), positive small (PS) and positive large (PL).
5. A method as claimed in claims 1 to 4, wherein the output height or level in the
hearth is categorized into five membership functions as very low (VL), low (L),
medium (M), high (H) and very high (VH).
6. A method as claimed in claims 1 to 5, wherein the set of rules of the Fuzzy
inference method decide the relationship between input variables and output
variables.
7. A method as claimed in claims 1 to 6, wherein the Fuzzy Associate Map (FAM) is
formed employing the corresponding output memberships and the weights for
each rule.
8. A method as claimed in claims 1 to 7, wherein for any given input data set (ST,
MT), one or more rules are fired and each rule after implications returns a
reshaped output membership function.

9. A method as claimed in claims 1 to 8, wherein the inputs to the aggregation
block are the output fuzzy sets corresponding to each rule and the output is a
combined single output fuzzy set or function.
10. A method as claimed in claims 1 to 9, wherein the input to the defuzzification
block is the aggregated fuzzy output set when the output value is the height of
the liquid metal and slag inside the furnace hearth in meters.
11. Fuzzy logic based system of monitoring level of liquid metal and slag inside the
hearth of a blast furnace as implemented according to the claims 1 to 10 for
implementing the method.

The invention relates to a fuzzy logic based system and method of monitoring level of
liquid metal and slag inside the hearth of a blast furnace. The said method comprising
the steps of analysing of differential flow data to decide the short and medium zone
time period and to establish an idea of the current level of liquids in the hearth, then
employing sensor outputs of the database to calculate the differential flows of liquid
metal and slag by program residing in the server and storing the short and the medium
term cumulative differential flows data when the fuzzy logic tool box fetches the data
from the database and estimates the levels of liquids in the hearth through a set of
prefixed rules, membership functions and inference methods wherein all the input and
output variables of the fuzzy inference method is first fuzzified before they are
implemented.