Method for controlling an operation of a blast furnace

Abstract

A method for operating a blast furnace by inferring the operating conditions in the blast furnace, on the basis of data supplied from sensors set in the blast furnace and also on the basis of the knowledge base formed by experience on the operation of the furnace by a small-scale computer and various processing units provided within the computer. Prior to the inference, the data from the sensors are processed by a large-scale computer and various processing units provided within the large-scale computer. The inference is performed by using certainty factor values, thereby making diagnosis of slip, hanging and channeling.

Description

BACKGROUND OF THE INVENTION

The present invention relates to a method for controlling the operation of a blast furnace, and particularly to a method for diagnosing the furnace operating conditions, based on information output from sensor means provided in the blast furnace.

DESCRIPTION OF THE RELATED ART

In recent times, various methods have been proposed for the diagnosing and controlling of the operation of a blast furnace. Japanese Patent Disclosure (KOKAI) No. 84-64705, for example, describes a method wherein:

(a) factors, which are clearly known on the basis of accumulated experience, are derived from data output from various sensors;

(b) these factors are, in connection with the phenomena occurring in the furnace, arranged in order and numerically evaluated; and

(c) from the factors thus arranged and evaluated in both a short and long term, the operating conditions of the furnace are detected.

This method, however, is disadvantageous in that it requires an analysis model to be maintained by means of modifications thereto in compliance with the changes the blast furnace undergoes throughout its life. Moreover, the modification itself is quite a time-consuming and complicated task, as the analysis model is quite complex.

SUMMARY OF THE INVENTION

It is an object of the present invention to provide a method for diagnosing a blast furnace operation, wherein an analysis model is easily modified in compliance with the change the blast furnace undergoes during its life.

According to the present invention, a method is provided for controlling an operation of a blast furnace, which comprises the steps of:

storing first data output from sensor means provided in the blast furnace into a central processing unit (CPU);

processing said first data, thereby preparing second data showing the operating conditions in the blast furnace; and

preparing true-and-false data, as the third data, by comparing said second data with standard data, and inferring a furnace operation condition, from the third data and a knowledge base formed by accumulated experience on the operation of the blast furnace.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic representation showing a method for diagnosing furnace operating conditions according to the present invention;

FIG. 2 is a schematic block representation of an apparatus for performing the method of the present invention;

FIG. 3 is a flowchart showing the method of the present invention;

FIG. 4 is a schematic representation illustrating the processes of inferring operating conditions of a blast furnace according to the present invention;

FIG. 5 is a graphic representation showing the results obtained by an example of the present invention; and

FIG. 6 is a graphic representation showing the relationship between the diagnosis results of an example obtained by the present invention and the actual operating conditions.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

An embodiment according to the present invention will now be described, with reference to FIGS. 1 to 4.

FIG. 1 schematically represents a method for diagnosing blast furnace operating conditions according to the present invention. Reference numeral10 denotes a large-scale computer. Computer 10 includes sequential processing means 12, which sequentially processes the data output from various sensor means 11, sequential filing means 13, sensor-data processing means 14, and interface buffer means 15. Reference numeral 20 denotes a small-scale computer which includes knowledge base means 21 for diagnosing the operating conditions of a blast furnace, common data buffermeans 22, and inference engine means 23. Reference numeral 30 denotes a cathode ray tube (CRT) which displays the results calculated by the inference engine means.

FIG. 2 schematically illustrates an apparatus for performing the method according to the present invention. Reference numerals 11a, 11b, and 11c each indicate sensors corresponding to sensor means 11 shown in FIG. 1. Large-scale computer 10 includes the following devices:

41: interface; 42: CPU; 43: read-only memory (ROM) storing program; 44 and 45: random access memories (RAMs); and 46: interface.

CPU 42 and ROM 43, which store the programs to be executed by CPU 42, constitute sequential processing means 12 and sensor-data processing means14, both shown in FIG. 1. RAM 44 constitutes sequential filing means 13 shown in FIG. 1. RAM 45 temporarily stores the data output from the sensormeans. RAM 45 and interface 46 constitute interface buffer means 15, shown in FIG. 1.

In FIG. 2, small-scale computer 20 includes keyboard 47, interface 48, CPU 49, ROM 50, RAMs 51 and 52, and interface 53. CPU 49 and ROM 50, which store the programs to be executed by CPU 49, constitute inference engine means 23 shown in FIG. 1. RAM 51 constitutes knowledge base means 21, alsoshown in FIG. 1. The data stored in RAM 51 can be altered by operating keyboard 47. New data can be added to these data by inputting the new databy keyboard 47 via interface 48. RAM 52 constitutes common data buffer means 22, as shown in FIG. 1. The data stored in RAM 45 of large-scale computer 10 are transferred to RAM 52 via interface 46. The results obtained by CPU 49 are supplied to CRT 30 through interface 53, and are displayed. The operation of the embodiment according to the present invention will now be described, in conjunction with the flowchart shown in FIG. 3.

(1) Firstly, the first data, output from sensor means 11, are read in a predetermined sequence by sequential processing means 12, and then stored in sequential filing means 13 (Step 1). Actually, this work is completed by supplying the data from sensors 11a, 11b, and 11c to RAM 44 through interface 41 under the control of CPU 42.

(2) The first data stored in sequential filing means 12 is processed by CPU42, thereby forming second data showing furnace operating conditions. This processing step produces data showing a rate of change, comparison of levels, dispersion of values and an integral value of the first data within a designated time interval. This work is actually carried out by CPU 42 (Step 2).

(3) The second data obtained in Step 2 are compared by CPU 42 with standarddata, thereby providing true-and-false data. The true-and-false data are stored in interface buffer means 15. More specifically, these data are stored in RAM 45 in FIG. 2 (Step 3).

(4) The true-and-false data stored in interface buffer means 15 are transferred to common data buffer means 22 (Step 4). More precisely, the stored data in RAM 45 are transferred to RAM 52.

(5) Inference engine means 23 infers the operating conditions of the furnace, based on the data stored in knowledge base means 21 and the true-and-false data stored in common data buffer means 24 (Step 5). This work is achieved as CPU 49 executes the program designated by the data stored in RAMs 51 and 52.

(6) Subsequently, the total judgement made by CPU 49 is supplied to CRT 30 through interface 53 and then displayed (Step 6).

(7) Then, it is determined whether a stop signal has been given or not. If YES, the processing is stopped. If NOT, it returns to Step 1. In the latter case, the aforementioned Steps 1-7 are repeated at predetermined intervals of, for example, 2 minutes.

Sensor means 11 detecting channeling and ship are largely classified into those for checking burden decent behavior, those for detecting pressure, those for detecting temperatures, and those for detecting gas compositions. In addition, residual slag data, obtained from information regarding pressure and temperature, also affect the channeling and slip. Table 1 shows the sensor means which detect the various conditions in the furnace.

TABLE 1______________________________________Items Sensor Means______________________________________A. Burden descent Stock line level soundings behaviorB. Pressure Top gas pressure guages Shaft pressure guages Blast pressure guagesC. Temperature Top gas thermometers Under-bell probes at throat level Horizontal probes set in Shaft thermometers Blast thermometersD. Gas constituent Top gas analyzersE & F. Residual slag Under-bell probes at throat level Shaft pressure guages Blower pressure guages______________________________________

The sensor means are not limited to those listed in Table 1. In addition tothose, a vertical probe, a differential type burden thickness and speed meter and a tuyere-nose-flame brightness measuring device can be used.

The step of processing the first data includes the following data-processings:

(a) Calculating the first regression coefficient for the past 30 minutes:

This is applicable to the temperature measured by horizontal probes, and top gas utilization ratio, and the temperature measured by under-bell probes.

(b) Calculating the balance between a value measured at present and that measured t.sub.k minutes before:

This is applicable to the shaft temperature.

(c) Calculating the balance between the two values obtained by the first regression coefficient formula by substitution of two values measured at present and t.sub.k minutes before:

This is applicable to the top gas utilization ratio.

(d) Calculating the balance between two values obtained at present and t.sub.k minutes before:

This is applicable to burden decent speed, pressure loss, shaft temperature, burden temperature at horizontal probe levels, shaft pressureand blast pressure.

(e) Calculating a standard deviation:

This is applicable to burden descent speed and pressure loss.

(f) Integrating a deviation from a theoretical value:

This is applicable to burden descent speed.

The second data, which have been pretreated, are shown in Table 2.

TABLE 2______________________________________Block Pretreated Data (Second Data)______________________________________A. Burden descent 201 Burden descent speed behavior 202 Dispersion of burden descent 200 speeds 203 Delayed of burden descent 204 Average of burden descent speedsB. Pressure 211 In-furnace pressure loss 201 212 Average of in-furnace pressure losses 213 Dispersion on in-furnace pressure losses 214 Average of shaft pressures in stable operationC. Temperature 221 Rise of shaft temperature 220 222 Average of shaft pressures in stable operation 223 Rise of burden temperature measured by horizontal probes set in a shaft 224 Average of burden temperatures in stable operation measured by the horizontal probesD. Gas composition 231 Fall of gas utilization ratio 230 232 Amount of fall of gas utilization ratioE. Residual slag (1) 241 Fall of temperature at the in- 240 furnace center part measured by under-bell probes set at a furnace throat level 242 Rise of temperature at the in- furnace peripheral part measured by under-bell probes set at a furnace throat levelF. Residual slag (2) 251 Rise of shaft pressure 250 252 Rise of blast pressure 253 Amount of residual slag______________________________________

The second data comprises:

Data 201 to 204 regarding burden descent behavior 200;

Data 211 to 214 regarding pressure 210;

Data 221 to 214 regarding temperature 220;

Data 231 and 232 regarding gas constituent 230;

Data 241 and 242 regarding residual slag (1) 240; and

Data 251 to 253 regarding residual slag (2) 250.

The items forming the second data are not limited to those described above.The residual slag (1) is affected by a burden distribution device, and the residual slag (2) is not affected by the same.

The second data are compared with the standard data to prepare true-and-false data. The standard data are composed of the average of those measured on a previous day, the average of those measured for an hour in stable operation, and a theoretical value.

In the case of burden descent behavior, for example, if the burden descent speed (a present moment value) minus burden descent speed (a previous day average value) is less than -K.sub.1 (where K.sub.1 is a constant) with respect to the four directions of sounding, the true-and-false data are "Burden descent speed is slow and a certain factor value is CF.sub.201." Thus, the true-and-false data contain this certainty factor.

In the case of in-furnace pressure loss, if the "pressure loss (a present moment value) minus pressure loss (a previous day average value is larger than K.sub.2 (where K.sub.2 is a constant)", the true-and-false data are "Pressure loss is large and a certainty factor value is CF.sub.211." The true-and-false data contain this certainty factor.

The items of true-and-false data, thus obtained, are listed in Table 3. Thetrue-and-false data include, in addition to data corresponding to the second data, data 361, based on human judgement, "Residual slag is great",and data 371 based on a previous alarm, "The previous alarm is high", wherea previous alarm is a total judgement made in the previous inference process.

TABLE 3______________________________________ The True-and-False Data (The Third Data)______________________________________A. Burden descent 301 Burden descent speed is slow behavior 302 Dispersion of burden descent 300 speed is large 303 Delay of burden descent is great 304 Average of burden descent speeds is greatB. Pressure 311 Pressure loss is large 310 312 Average of pressure losses is large 313 Dispersion of pressure losses is large 314 Shaft pressure is highC. Temperature 321 Shaft temperature is rapidly 320 rising 322 Amount of shaft temperature rise is high 323 Burden temperature measured by horizontal probes set in a shaft is rapidly rising 324 Amount of burden temperature rise measured by the horizontal probes is largeD. Gas constituent 331 Gas utilization ratio is rapidly 330 falling 332 Amount of gas utilization ratio is largeE. Residual slag (1) - 341 Temperature of in-furnace sensor judgement central area measured by 340 under-bell probes set at a furnace throat level is falling 342 Temperature of in-furnace peripheral area measured by under-bell probes set at a furnace throat level is risingF. Residual slag (2) - 351 Rise of shaft pressure is n or different sensor more judgement 352 Blast pressure is rapidly rising 350 353 Residual slag is greatG. Residual slag (3) - 361 Residual slag is great human judgement 360H. Previous alarm 371 The previous alarm is high 370______________________________________

The knowledge base is composed of knowledge units corresponding to the true-and-false data. Each knowledge unit indicates an operator's knowledgeand experience on the operating production process, in the form of "if . . . , then . . . ". In this embodiment, the reliability of inference is raised by introducing to the inference process a certainty factor (CF) which indicates the uncertainty degree of each rule for the operating production process.

In the case of, for example, a rule for burden descent, the expression is: "If burden descent speed is slow, then, the operation has tendency toward channeling".

In the case of a rule for pressure loss, the expression is: "If pressure loss is large, then, the operation has a tendency toward channeling".

An example of inference of judgement will now be given with reference to FIG. 4.

The processed results indicated by "OR" are the sum of the true-and-false data, into which CF values are taken, not a logical sum. The processed results indicated by "AND" are a logical product.

(1) The fourth data are obtained by inferences, based on the true-and-falsedata (the third data) and the knowledge base, and also based on CF values.

In the case of a burden descent rule, inference is performed, in combination with consideration of the CF values, on the basis of both the knowledge base, and the true-and-false data which are composed of:

"Burden descent speed is slow" (301);

"Dispersion of burden descent speed is large" (302);

"Delay of burden descent is large" (303);

"Average of burden descent speeds is slow" (304).

Thus, the inference results are a sum 400 of the items of data used in the inference process, i.e., the fourth data, which is accompanied by a CF value and which indicates: "Burden descent is out of successful operation".

In case of a pressure loss rule, inference is performed in the same manner corresponding CF values, on the basis of both the knowledge base and the true-and-false data of 311 to 314. thus, the inference results are a sum 410 of the items of data used in the inference, i.e., the fourth data, which is accompanied by a CF value summed, and which indicates "Pressure dispersion is large".

In the case of a temperature rule, the inference results, obtained from thetrue-and-false data of 321 to 324, are also a sum 420, or the fourth data. The fourth data are accompanied by a summed-up CF value and indicate: "Temperature dispersion is large".

In the case of a gas utilization rule, the inference results obtained from the true-and-false data of 331 and 332 are a sum 430, or the fourth data. The fourth data are accompanied by a summed-up CF value and indicate: "Gasutilization ratio is falling".

In the case of the inference regarding the amount of residual slag, based on sensor information (440) "Residual slag is great", which is processed by sensor judgement, a logical product is obtained from the following:

(1) "Temperature of in-furnace central area, measured by under-bell probes set at a furnace throat level, is falling" (341).

(2) "Temperature of in-furnace peripheral area, measured by under-bell probes set at a furnace throat level, is rising" (342).

If either of the two is false, the logical product is false. Except for this respect, the inference is performed in the same manner as described in the foregoing.

In the case of a residual slag rule other than the above-mentioned, the following true-and-false data are used:

(1) "Rise of shaft pressure is n or more" (351).

(2) "Blast-pressure is rapidly rising" (352).

(3) "Residual slag is great" (353).

Moreover, in some rules for residual slag, human judgement based on the operator's judgement is incorporated. From this processing, data 450, which constitutes the fourth data, is obtained. These data represent whether residual slag is great (450) (human judgement).

TABLE 4______________________________________400: Burden descent behavior is out of successful operation410: In-furnace pressure dispersion is large420: In-furnace temperature dispersion is large430: Gas utilization ratio is falling440: Residual slag is great (based on sensor judgement)450: Residual slag is great (based on human judgement)______________________________________

Among those rules for burden descent behavior, pressure loss, temperature, gas composition and residual slag, the burden descent behavior and the pressure rules affect most slips and channelings. There are three residualrules: the rule for residual slag (1), the rule for residual slag (2), and the rule for residual rule (3). This is because a single rule for residualslag (1) is insufficient when an operation of a burden distribution device is modified. Human judgement of residual slag (3) can be replaced by slag flow meter judgement.

(2) The fifth data are obtained on the basis of the fourth data.

An inference is performed, in consideration of the summed-up CF value accompanied by each of the results inferred in the fourth data-making process in respect to the burden descent behavior, pressure loss temperature and gas composition, on the basis of the following:

"Burden descent behavior is out of successful operation" (400);

"In-furnace pressure dispersion is large" (410);

"In-furnace temperature dispersion is large" (420);

"Gas utilization ratio is falling" (430).

Thus, the inference results, or a sum 500, i.e., the fifth data, is accompanied by a CF value and indicates: "Channeling (sensor judgement)" (500).

In the same manner, an inference is performed, in consideration of the sum of the CF values accompanied by the results inferred in the process of making the fourth data in respect to the rest of the items on the basis ofthe following:

"Residual slag is great (sensor judgement)" (440);

"Residual slag is great (human judgement)" (450).

Thus, the inference results or data 510, i.e., the fifth data, are accompanied by a CF value summed up in the inference process and indicates: "Channeling (residual slag)" (510).

TABLE 5______________________________________500: Sensor judgement data510: Residual slag judgement data530: Previous alarm data______________________________________

(3) Total judgement is made on the basis of the fifth data.

Finally, an inference is performed, in consideration of the summed-up CF values accompanying the results inferred in the process of making the fifth data, on the basis of the following:

"Channeling (sensor judgement)" (500);

"Channeling (residual slag judgement)" (510);

"Channeling (previous alarm)" (520).

The inference results, or a sum of these items of data indicates "Channeling (total judgement)".

The embodiment explained above is for the diagnosis of channeling. Diagnosis of slip and hanging is made fundamentally in the same manner as the diagnosis of channeling, except that the CF values for slip and hanging are different from those for channeling.

In the embodiment, the certainty of foreseeing slip (CF.sub.f.sup.s) and the certainty of foreseeing channeling (CF.sub.f.sup.c) are obtained by carrying out inferences in three steps that is, the step of obtaining the fourth data from the true-and-false (third) data; the step of obtaining the fifth data from the fourth data; and the step of making the total judgement from the fifth data. This step-by-step inference gives the following advantages:

(1) Inference becomes speedy.

(2) Causes for slip and channeling are easy to clarify.

(3) CF values can be well maintained, even if modified.

The following substitutions can be employed in the invention:

(a) a method wherein a total judgement is inferred directly from the true-and-false (third) data;

(b) a method wherein a total judgement is inferred from the fifth data prepared directly on the basis of the true-and-false (third) data; and

(c) a method wherein a total judgement is made directly from the fourth data prepared on the basis of the true-and-false (third) data.

However, the method of inference described in the embodiment is more desirable, by reason of the aforementioned (1), (2), and (3), than the substitutions.

When a total judgement is made, the furnace is controlled in accordance with the results (CF.sub.f.sup.s, CF.sub.f.sup.c) of the judgement. The control is achieved by reducing the blast supplied through tuyeres, by changing the distribution of burden charged through a furnace top, by controlling the fuel ratio, by adjusting the composition of slag or, by increase or decrease of the temperature of the tuyere-nose flame. The reduction of the blast is effective, particularly for channeling.

The reduction of blast is standardized according to probabilities of slip. The reduction is set, for example, to 300 Nm.sup.3 /min. (4 to 5% of totalblast volume) when CF.sub.f.sup.s is 60 to 80%, and to 500 Nm.sup.3 /min. (7 to 8% of total blast amount) when CF.sub.f.sup.s is greater than 80%.

EXAMPLE 1

Diagnosis of slip was carried out for a 4,663 m.sup.3 -blast furnace on theoperating conditions shown in Table 6. The diagnosis is made every 2nd minute. The results of the diagnosis at 12.degree.10' are shown in Table 7.

TABLE 6______________________________________Blast volume 6500 Nm.sup.3 /min.Blast pressure 3.4 to 3.5 kg/cmGas temperature 900.degree. C.Tuyere nose flame temperature 2000.degree. to 2100.degree. C.Fuel ratio 514 kg/T .multidot. pigSi contained in molten metal 0.3 to 0.4%S contained in molten metal 0.02 to 0.025%Molten metal temperature 1480.degree. to 1500.degree. C.______________________________________ TABLE 7______________________________________Inference reasons for slip Probability______________________________________Burden descent behavior judgement 22%teaches tendency toward slipPressure loss judgement teaches 13%tendency toward slipTemperature judgement teaches 22%tendency toward slipPossibility that residual slag is 0%great is lowSensor judgement: Aggregatetendency toward slip 47%Influence coefficient of 30%previous alarm probabilityProbability of slip in this Aggregateoperation 63%______________________________________

EXAMPLE 2

Diagnosis of channeling was carried out every 2nd minute for 3 hours. the blast furnace was operated under the conditions shown in Table 6. The results are represented in FIG. 5, wherein (a) designates the burden descent behavior, (b) denotes the probability of slip (CF.sub.f.sup.s), and (c) represents the probability of channeling (CF.sub.f.sup.c)--all in relation to the lapse of time.

Remarkable increases of the burden descent speed occurred at points of X, Y, and Z where the probability of slip (CF.sub.f.sup.s) was high. This teaches that slips have occurred. Thus, a close relationship was proved toexist between CF.sub.f.sup.s and slip occurrence.

EXAMPLE 5

The relationship between probability of slip (CF.sub.f.sup.s) and the conditions are the same as in the case of Example 1. The results of the diagnosis at 12.degree.10' are shown in Table 8.

TABLE 8______________________________________Inference reasons for channeling Probability______________________________________Burden descent behavior judgement 10%teaches tendency toward channelingPressure loss judgement teaches 8%tendency toward channelingTemperature judgement teaches 10%tendency toward channelingPossibility that residual slag is 0%great is lowSensor judgement: tendency toward Aggregatechanneling 25%Probability of channeling in this Total aggregateoperation 25%______________________________________

EXAMPLE 3

Diagnosis of slip and channeling was carried out. The actual slip occurrence ratio is shown in FIG. 6.

The data-sampling period was 10 days. The actual slip ratio is given by formula of N/F, where N is the number of highest CF.sub.f.sup.s levels obtained during the 10-day period, and F is the number of slips occurring during the same period. Each of these levels is highest of the fifteen levels detected within thirty minutes. They were counted at a ten-percent pitch.

When CF.sub.f.sup.s was greater than 60%, it was ascertained that slips occurred with a probability of greater than 70%.

Claims

1. A method for preventing channeling, slip and hanging in a blast furnace, comprising the steps of:

measuring a burden-descent speed by a first sensor means;

measuring an in-furnace pressure by a second sensor means;

measuring a burden temperature by a third sensor means;

measuring an in-furnace gas temperature by a fourth sensor means;

measuring a composition of an in-furnace gas by a fifth sensor means;

measuring a pressure of a blast gas blown in through tuyeres set in the blast furnace by a sixth sensor means;

storing first data from the first to sixth sensor means;

processing said first data, and preparing second data corresponding to the conditions of the channeling, slip and hanging in the blast furnace;

preparing true-and-false data, as third data, by comparing said second data with standard data;

inferring conditions of channeling, slip and hanging, from the third data and from a knowledge base formed by accumulated experience on an operation of the blast furnace; and

repeating the above-mentioned steps at predetermined intervals.

2. The method according to claim 1, wherein said predetermined intervals are about two minutes.

3. A method according to claim 1, wherein said step of inferring comprises a step of processing the true-and-false data, in consideration of a CF value corresponding to an individual item of the true-and-false data, to prepare a total judgement datum.

4. A method according to claim 1, wherein said step of processing the first data includes at least one selected from the following processings:

calculating a rate of change within a designated time interval;

calculating an integral value within a designated time interval;

calculating dispersion of values within a designated time interval; and

calculating a balance between a value measured at a present time and a value measured before.

5. A method according to claim 1, wherein said second data includes at least one selected from the following:

data representing burden-descent behavior;

data representing in-furnace pressure loss;

data representing in-furnace temperature;

data representing composition of in-furnace gas; and

data representing residual slag.

6. A method according to claim 5, wherein said data representing burden descent behavior includes at least one selected from the following:

burden-descent speed;

dispersion of burden-descent speed;

delay of burden-descent; and

average of burden-descent speeds.

7. A method according to claim 5, wherein said data representing in-furnace pressure loss includes at least one selected from the following:

in-furnace pressure loss;

average of in-furnace pressure losses;

dispersion of in-furnace pressure losses; and

average of shaft pressures in stable operation.

8. A method according to claim 5, wherein said data representing in-furnace temperature includes at least one selected from the following:

rise of shaft temperature;

average of shaft temperatures in stable operation;

rise of burden temperature, measured by horizontal probes set in a shaft; and

average of burden temperatures in stable operation, measured by the horizontal probes.

9. A method according to claim 5, wherein said data representing gas composition includes at least one of a drop of a gas utilization ratio and the drop of gas utilization amount.

10. A method according to claim 5, wherein said data representing residual slag includes at least one selected from the following:

a fall of temperature of the in-furnace central area at a furnace-throat level, measured by under-bell probes;

a rise of in-furnace temperature of the periphery area at a furnace-throat level, measured by under-bell probes;

a rise of shaft pressure;

a rise of blast pressure; and

an amount of residual slag.

11. A method according to claim 1, wherein said knowledge base comprises at least one selected from the following:

a rule for burden-descent;

a rule for in-furnace pressure loss;

a rule for in-furnace temperature;

a rule for in-furnace gas constituent; and

a rule for residual slag.

12. A method according to claim 1, wherein said step of inferring comprises the steps of:

producing true-and-false data, in consideration of a certainty factor (CF) value corresponding to an individual item of the true-and-false data, to prepare at least one item of data constituting fourth data, selected from the following:

data showing levels of burden descent;

data showing levels of in-furnace pressure dispersion;

data showing levels of in-furnace temperature dispersion;

data showing levels of gas utilization ratio;

data showing levels of amount of residual slag by sensor judgement; and

data showing levels of amount of residual slag by human judgement;

processing said fourth data, in consideration of a CF value corresponding to an individual item of said fourth data, to prepare fifth data including data obtained by sensor judgement and data obtained by human judgement; and

processing said fifth data and a datum of a previous alarm, in consideration of a CF value corresponding to an individual item of the fifth data, to prepare a total judgement datum.

13. A method according to claim 1, wherein said step of inferring comprises the steps of:

processing the true-and-false data, in consideration of a CF value corresponding to an individual datum of the true-and-false data, to prepare at least one item of data constituting fourth data, selected from the following:

data showing levels of burden descent;

data showing level of in-furnace pressure dispersion;

data showing levels of in-furnace temperature dispersion;

data showing levels of gas utilization ratio;

data showing levels of amount of residual slag by sensor judgement; and

data showing levels of residual slag by human judgement;

processing said fourth data and a datum of a previous alarm, in consideration of a CF value corresponding to an individual item of said fourth data, to prepare a total judgement datum.

14. A method according to claim 1, which further comprises the step of controlling an operating condition of the blast furnace in accordance with results of the inferring step.

15. A method according to claim 14, wherein said step of controlling an operating condition of the blast furnace comprises reducing a blast blown into the blast furnace through tuyeres set in the blast furnace.