JPH0598325A - Device for controlling distribution of charging materials in blast furnace - Google Patents

Abstract

PURPOSE:To control the distribution of raw materials in a blast furnace automatically and to always stabilize a furnace condition by optimizing a gas flow distribution while considering the operational condition, such as the intensity and the shifting direction of gas flow, the waiting time for charging the raw material, the reduction of blasting, the action history of the raw material distribution, the residual iron quantity, the iron tapping condition in the blast furnace. CONSTITUTION:Knowledges relating to the blast furnace operation are resistered and corrected as a knowledge base and data from various kinds of sensors arranged in the blast furnace are fetched into a process computer in the prescribed period and the included noise is removed, and respective data based on the knowledge base with the operational experiences and the model experiment are inferred as the intensity and the shifting tendency of the gas flow in each sensor. This result is synthesized as the intensity and the shifting tendency of the gas flow in a center part, a peripheral part and a local part in the furnace based on the information and the knowledge relating to the time of charging the raw material and reducing the blasting, etc., and by various kinds of the decided matrix relating to the synthesized result and the charging material from the operational condition, the change in the raw material blending, etc., is corrected in advance and the action means and the action quantity in the distribution of the charging raw material are automatically controlled to stabilize the furnace condition.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control device for controlling a gas flow through a distribution control of a charge in a furnace of a blast furnace to stabilize the gas flow distribution and operation.

[0002]

2. Description of the Related Art FIG. 25 is a diagram showing the distribution control of the charge and the condition inside the furnace, and Table 1 is a table showing the position and shape of the melting zone and the operating condition.

[0003]

[Table 1]

Generally, a large amount of reducing gas (hereinafter referred to as gas) generated by the combustion of coke flows in the center of the blast furnace, and a large amount of the reducing gas flows in the furnace wall.
In addition, a large amount of flow locally is called local flow. Usually, coke has a larger grain size than ore as a raw material to be charged into a blast furnace, and since coke does not undergo reduction pulverization, coke has better air permeability. Further, in the zone where the ore is melted (hereinafter referred to as the melting zone), the ventilation resistance of the ore layer is 200 to 300 times that of the coke layer, and the gas passes through the coke layer (the coke layer in the melting zone is hereinafter referred to as the slit). Flowing.

When a large amount of coke is charged in the center of the blast furnace, it becomes a central flow, the melting zone shape becomes Mt. Fuji shape, the number of slits increases, and the gas flow becomes stable. Also, if a large amount of coke is charged into the furnace wall, it becomes a peripheral flow, which is said to be useful for removing deposits on the furnace wall. Furthermore, if a large amount of local flow occurs, it becomes a local flow, which causes a sudden abnormal reactor condition such as blow through or slip. Therefore, for stable operation, the distribution of the charge is controlled to stabilize the gas flow, and the local flow is detected by detecting the occurrence of the local flow and taking a locally different action. It is necessary to prevent growth.

In the case of a bell type blast furnace, charging of raw materials is performed by opening and closing a large bell. One large bell opening / closing operation is called a batch, and the coke and ore are charged in several batches (for example, 2 batches of coke and 3 batches of ore). This is called 5 batch charging, and 5 batches are collectively called 1 charge. In normal operation, the charging interval is about 10 minutes.

FIG. 26 is a diagram showing the installation status of the MA.
(A) is the top view, (b) is a side view. In order to control the landing position of the raw material that falls freely from the large bell, about 20 MAs are installed in the blast furnace at equal intervals in the circumferential direction, and the MA stroke moves from the furnace wall to the center. You can do it. Then, the charge distribution control is performed by changing the MA stroke in batch units with the charge cycle. Also, if SL (the distance from the large bell to the surface of the charge) is raised or lowered, the dropping position of the charging raw material changes, and if the ore base (the ratio of the ore to be charged and the coke) is changed, the surface property of the charging is changed. If the usage rate of the fine-grain raw material changes, the inclination angle of the charging surface changes, and if the time-series discharge speed changes, the flow of the raw material changes on the charging surface. It can be used for distribution control.

The relationship between MA and charging distribution has been disclosed in, for example, "Materials and Process (1) 1988, p74". However, the content is a formalization of the relationship between MA action and charge distribution based on the results of intermittent measurements of layer thickness distribution of coke and ore in the furnace with a sonde, and the results obtained in outdoor model tests. Therefore, the amount of action is determined from the strength of the gas flow in the actual operation and the tendency of transition, and M
A consistent system technology aiming at controlling A and optimizing the gas flow distribution has not yet been established.

A technique for predicting blow-through or slip based on sensor information is disclosed in, for example, Japanese Patent Laid-Open No. 62-2707.
No. 12 publication. However, this is a system for the purpose of predicting blow-through and slip, not a consistent system for optimizing the gas flow distribution in the blast furnace. Further, JP-A-2-182815 proposes a method of controlling the MA, SL, ore base, use ratio of fine grain raw material and time-series discharge rate by inferring using a knowledge engineering method. There are problems that the selection criteria for these actions and the criteria for determining the action amount are not concrete, and the influence of the amount of residual iron in the reactor is not taken into consideration.

[0010]

SUMMARY OF THE INVENTION The present invention has been made in view of such a situation, and the strength and transition tendency of the gas flow in the furnace of the blast furnace, charging waiting, reducing wind, charging distribution Blast furnace charge distribution that enables automatic control by determining the action means and action amount for optimizing the gas flow distribution by taking into consideration the action history of the system, the operating conditions such as the amount of remaining pig iron and the tapping situation, etc. Aiming to provide a control device

[0011]

A blast furnace charge distribution apparatus according to the present invention includes means for registering and correcting knowledge about blast furnace operation as a knowledge base, and data from various sensors installed in the blast furnace at predetermined intervals. A means to take in the process computer, a means to remove the noise contained in the taken data, and the noise-removed data based on the operating experience and the knowledge base obtained from model experiments. A means to infer as a transition tendency, and a means to synthesize the inference result as intensity and transition tendency of the gas flow regarding the central flow, peripheral flow, and local flow based on information and knowledge base about charging and wind reduction, etc. Based on the combined result, determine the action means of the distribution of the charge based on the operation status and operation policy at that time, and
A action decision matrix, SL action decision matrix, ore-based action decision matrix,
A means for determining the action amount by applying it to either the fine grain raw material usage change action determination matrix or the time-series discharge rate action determination matrix, and the action amount, the distribution of charge distribution action history, change of blast pressure, furnace A means for adding a temporary factor such as the amount of residual iron in the interior, a means for preemptively compensating for future events such as a composition change, and a means for automatically transmitting the corrected amount of the charge distribution action to the PLC. And a means for guiding the operator of the amount of action of the corrected distribution of the charge and the amount of action.

[0012]

According to the present invention, knowledge about blast furnace operation is registered / corrected as a knowledge base, and data from various sensors installed in the blast furnace is taken into the process computer at a predetermined cycle. Then, the noise included in the acquired data is removed, and the noise-removed data is inferred as the intensity and transition tendency of the gas flow for each sensor based on the operating experience and the knowledge base obtained from the model experiment. The inference result is synthesized as the strength and transition tendency of the gas flow with respect to the central flow, peripheral flow and local flow based on the information and knowledge about charging and wind reduction. The action means of the distribution of the charge is determined from the synthetic result and the operation situation and operation policy at that time, and the MA action decision matrix, the SL action decision matrix, the ore-based action decision matrix, the fine grain raw material usage action decision matrix, and the time are determined. Apply to any of the series emission rate action determination matrices to determine the action amount. Furthermore,
The action amount is corrected by taking into account future factors such as changes in the charge distribution action history and changes in blast pressure, transient factors such as the amount of residual iron in the furnace, and changes in the composition in advance. Then, the action means and the action amount of the corrected distribution of charges are set to P
Transmit to LC for automatic control. Further, the operator is instructed on the action means and the action amount of the corrected distribution of the charge.

[0013]

1 is a conceptual diagram of a blast furnace charge distribution control apparatus according to an embodiment of the present invention. In the figure, 10 is a blast furnace, in which a fixed sonde (measurement of temperature and component) and a shaft thermometer are installed in the circumferential direction on the upper part of the furnace wall, and a horizontal sonde (radial gas temperature, component) is installed below it. ing. Also,
At the top of the furnace, the throat sonde (radial gas temperature, composition) throat TV (measured as the surface temperature distribution on the charge surface), charge sonde (radial charge surface shape, layer of coke and ore) Various sensors such as thickness distribution) and MA are installed.

Reference numeral 20 is a process computer conventionally used for controlling the blast furnace. This computer 20
Is a means 21 for fetching data from various sensors installed in the blast furnace into a process computer at a predetermined cycle, a means 22 for removing noise contained in the fetched data, a charging standby, a wind reduction 23, a charging. File means 26 for saving and storing information such as action history 24 of the object distribution and tapping situation 25 at any time, means 27 for guiding the operator of the inference result and process, and the charging result by transmitting the guidance result to the PLC 40. A means 28 for automatically controlling the distribution and a means 2 for transmitting the data that the process computer is waiting for to the knowledge processing computer 30 and receiving the result of the knowledge processing.
9 and 9 are built in, and they are realized by the system program.

Reference numeral 30 denotes a knowledge processing computer, which registers and corrects knowledge about the operation of the blast furnace as a knowledge base 31, and based on the data received from the process computer 20 and the knowledge base 31, a central flow, a peripheral flow, and Means 32 for deducing the strength and transition tendency of the gas flow related to the local flow, and means 33 for synthesizing the result as the strength and the transition tendency of the gas flow based on the information on charging and wind reduction and the knowledge base, The action means for distribution of the charge is determined from the synthesis result and the operation status and operation policy at that time, and the MA action decision matrix, SL action decision matrix, ore-based action decision matrix, fine grain raw material usage change action decision matrix, and time Series emission rate Action decision Apply to any of the matrix Action amount The means 34 for determining and the action amount are corrected by taking into account transient factors such as change of charge distribution action history and change of blast pressure, amount of residual iron in the furnace, and future events such as composition change in advance. It incorporates means 35 and means 37 for receiving data necessary for inference from the process computer 20 and transmitting inference results to the proseca computer, and these are realized by the system program thereof.

As described above, the apparatus is divided into the process computer 20 and the knowledge processing computer 30.
This is because there is a part to be processed by the conventional system technology and a part to be processed by the artificial intelligence application technology, and it is convenient for system development to divide these parts, and it is not essential to the present invention. Therefore, if one computer is logically divided and both means are realized, it can be realized by one computer.

(1) Preprocessing Method Various sensor data are taken in at a predetermined cycle by the constant cycle processing function of the process computer 20, and the file 2
After storing in 11, remove the noise contained in the data,
Pre-processing is performed to extract meaningful information as control information. In the pre-processing, data processing is performed by exponential smoothing, linear regression, etc. in order to consider the type of sensor and the state of noise contained in the sensor data, and to make the data effective for gas flow determination.

A. Exponential smoothing

## EQU00001 ## Sn = (1-1 / t) * Sn-1 + Rn / t Sn: Value after exponential smoothing at time n Sn-1: Value after exponential smoothing at time n-1 Rn: Exponential smoothing at time n Previous value t: Specified for each sensor with time constant 1 ≤ t n, t Unit is minute

B. Primary regression

[Equation 2]

Τ ≦ ti ≦ 0: reference point is inference execution time, τ: time constant, n: number of valid data Xo: first-order regression data, and Sn or Xo is used to judge the strength and transition tendency of the gas flow. ..

(2) Judgment of strength of gas flow The result of pre-processing is sent to the knowledge computer 30, and the strength of gas flow is judged based on the reference value learned and controlled for each sensor. Below is an example of how to determine the reference value and how to determine the strength of the gas flow. (2-1) How to determine the reference value The reference value is determined using the exponential smoothing method based on the daily average value,
We have made it possible to easily respond to changes over time in equipment and operations.

[0022]

[Equation 3]

The exponential smoothing constant α (0≤α≤1) is
It is a function of the charging waiting time, the wind reduction time, the number of MA actions, etc. Furthermore, the charging waiting time, the wind reduction time, the number of MA actions, etc. are based on operating experience and online experiments. It is expressed by the function shown in (h) to remove the special operation factor.

[0024]

Α = α1 × α2 × α3 × α4 × α5 × α6 × α7 × α8 α1 = f (charging waiting time) α2 = f (wind reduction time) α3 = f (number of MA actions) α4 = f ( SL action number) α5 = f (ore base action number) α6 = f (fine grain raw material change number) α7 = f (time-series discharge rate change number) α8 = f (all action number) where charging time ( The interval from charging to the next charging) is
The target waiting time is longer than the average charging waiting time, and the charging waiting time is calculated by the following formula.

[0025]

[Equation 5]

The wind reduction time (the time when the air flow rate is reduced from the target air flow rate) is empirically calculated by the following equation based on the time when the air flow is reduced and the air flow reduction amount.

[0027]

[Equation 6]

(2-2) Judgment of strength of gas flow Sensor information for measuring gas temperature and components in the radial direction such as horizontal sonde, furnace mouth sonde, and furnace TV can be used for judgment of central flow and peripheral flow. Also, a fixed sonde can be used to judge the surrounding flow. In addition, furnace TVs and fixed sondes can be used to determine local flow. FIG. 3 shows the relationship between the gas flow intensity and the gas temperature and components. Generally, when the gas flow is strong, the gas temperature is high and the gas utilization ratio (CO ₂ / CO) is high.
Will be lower.

(2-2-1) Determination of central flow and peripheral flow a. Horizontal probe (Fig. 4) The installation situation is shown in Fig. 4 (a). The horizontal sonde inserts the sonde into the interior of the furnace several times a day to measure the gas temperature and components in the furnace radial direction (gas flow rate will be added in the future). Generally, if the heat level in the furnace is high, the gas temperature is measured high, and conversely, if the heat level is low, the gas temperature is measured low, but the heat level has little influence on the gas temperature and the distribution of components. For this reason, it was decided to determine the gas flow by comparing the measured data with the patterns obtained from operational experience and the previously measured data. FIG. 4 (b) is a comparative example of the measured data and the pattern obtained from the operation experience, in which the central flow is strong,
It indicates that the peripheral flow is weak.

B. Furnace mouth sonde (Fig. 5) Installation status is shown in Fig. 5 (a), and measurement data is shown in Fig. 5 (b).
Are shown respectively. Furnace sondes are installed in the east-west and north-south directions above the surface of the charge at the top of the furnace, and the gas temperature and gas components are continuously measured by multiple thermometers and analyzers. Here, in order to accurately determine the gas flow, when the gas temperature in the peripheral part is higher than the central part by a certain value (the component CO ₂ / CO is low), the lower (higher) sensor is used based on the operating experience. After judging as abnormal and removing the data,
The average of north, south, east and west is taken. Then, the gas temperature is compared with the furnace top gas temperature, and the gas components are judged in the same manner as in the horizontal sonde to determine the gas flow condition. An example of processing relating to gas temperature is shown below. FIG. 5C shows the gas flow condition when the gas temperature intensity (X) is plotted on the horizontal axis, and in this example, the central flow is strong and the peripheral flow is weak. here,

## EQU00007 ## X = (Xi-t)-(Xib-Tb) Xi: average value of i furnace opening gas temperature, t: furnace top gas temperature Xib: reference value of furnace opening gas temperature, Tb: furnace top gas temperature Is the standard value, and the throat gas temperature is the average of north, south, east, and west. Since the furnace top gas temperature is measured at the place where the gas generated in the blast furnace gathers, it is more accurate than the value obtained by weighted averaging the furnace port gas temperature with the surface area of the furnace port.

C. Furnace TV (Fig. 6) The measured raw data is shown in Fig. 6 (a). Furnace TV continuously measures the surface temperature distribution of the charge as a thermal image at the top of the furnace. The charging surface temperature rises gradually after the charging is completed, and reaches the maximum before the next charging starts. In addition, the surface temperature of the charging material cannot be accurately captured due to dust generation immediately after charging. For this reason, data ((b) in FIG. 6) corresponding to the temperature measurement point of the furnace opening sonde is created based on the thermal image after 4 to 5 minutes after charging, and then the same method as the processing of the furnace opening sonde is performed. To determine the gas flow situation.

B. Fixed sonde temperature (Fig. 7) The installation status is shown in Fig. 7 (a), and the measured data is shown in Fig. 7
(B) of. A plurality of fixed sondes are installed in the furnace wall above the shaft of the blast furnace in the circumferential direction (sometimes several steps further in the vertical direction) to measure the gas temperature and composition inside the charge. Therefore, it can be used to judge the surrounding flow. In general, when the sensor becomes defective or there is a deposit, the gas temperature and components become extremely low. Therefore, the adjacent sensor data are compared with each other, and when the difference between them is equal to or larger than a certain value, the lower sensor is determined to be abnormal, and the average value (Xt) of the remaining sensor data.
To calculate. In addition, the peripheral flow intensity (X) is an average value (Xt)
And the gas flow condition is determined by calculation based on the reference value (Xo).

## EQU8 ## X = Xt-Xo

E. Others Several shaft temperatures are installed in the furnace wall in the circumferential direction and in the vertical direction, and the temperature inside the furnace wall brick is measured, so it can be used to determine the surrounding flow in the same way as a fixed sonde. However, since the gas temperature is not directly measured, the accuracy of the gas flow information is lower than that of the fixed sonde. Moreover, the shape of the melting zone is sharp when the central flow (see FIG. 8).
(A)), the number of slits in the coke through which the reducing gas passes increases, the gas flow becomes stable, the ventilation resistance index and blast pressure are low, and heat loss and fluctuations in sensor data also decrease.
Gas utilization ratio _(CO 2 / CO) increases. On the contrary, in the peripheral flow, the shape of the melting zone becomes flat ((b) in FIG. 8),
Since the number of slits also decreases, the gas flow becomes unstable, the ventilation resistance index and blast pressure are high, heat loss and fluctuations in sensor data also increase, and the gas utilization rate (CO ₂ / CO) decreases.
Therefore, the gas flow status can be determined by comparing the current value and the reference value with respect to the gas utilization rate, ventilation resistance index, blast pressure, heat loss, fluctuations in sensor data, and the like.

(2-2-2) Determination of local flow a. Furnace TV (Fig. 9) Furnace TV continuously measures the temperature distribution on the surface of the charge as a thermal image at the top of the furnace. FIG. 9A shows raw data measured by the furnace TV when the charging surface of the blast furnace is viewed from directly above. The charging surface temperature rises gradually after the charging is completed, and reaches the maximum before the next charging starts. In addition, the surface temperature of the charging material cannot be accurately captured due to dust generation immediately after charging. Therefore, based on the thermal image obtained 4 to 5 minutes after charging, the image is equally divided in the circumferential direction according to the number of MAs, and the position of the MA is corresponded based on the weighted average value of the temperature of the peripheral portion. Data (FIG. 9 (b)) is created to determine the gas flow in the circumferential direction. For example, FIG. 9B shows that the gas flow on the north side is strong. here,

ΔXi = Xi−X Xi: average value of charging surface temperature corresponding to the position of MA X: average value of charging surface temperature X is calculated as the average of Xi.

B. Fixed sonde temperature (Fig. 10) The installation status is shown in Fig. 10 (a), and the measured data is shown in Fig. 1.
0 (b) respectively. A plurality of fixed sondes are installed in the furnace wall above the shaft of the blast furnace in the circumferential direction (sometimes several steps further in the vertical direction) to measure the gas temperature and components inside the charge. Therefore, if the relationship between the measurement position and the position of MA is taken into consideration in consideration of the distribution in the circumferential direction, and the data is converted into data corresponding to the position of MA using a linear regression equation or the like, the gas flow corresponding to the position of MA can be determined. Available. For example, FIG. 10B shows that the gas flow on the north side is strong. Here, ΔXi is obtained by the following equation.

[0036]

[Equation 10] X is calculated as the average of Xi. It should be noted that the fixed sonde can also analyze the gas component, and if this value is used, it can be used to determine the local flow in the same manner as the temperature data.

C. Others Several shaft temperatures are installed in the circumferential direction of the furnace wall and in the vertical direction, and the temperature inside the brick of the furnace wall is continuously measured. Therefore, it can be used to determine local flow in the same way as the fixed sonde temperature. However, since the gas temperature is not measured directly, the accuracy of the information is lower than that of the fixed sonde.

(2-3) Judgment of Transition Tendency (2-3-1) Judgment of Transition Tendency of Central Flow and Peripheral Flow The transition tendency of the dregs flow is a time series of the sensor information used for the determination of the gas flow intensity. The transition can be used as it is. For example, in this apparatus, it is determined that the central flow is in the transition tendency when the information shown in FIG. 11A has an upward tendency and the information shown in FIG. 11B has a downward tendency. Further, when the information shown in (c) of FIG. 11 has a rising tendency and when the information shown in (d) of FIG. 11 has a falling tendency, it is determined that there is a tendency of transition of the peripheral flow.

(2-3-2) Judgment of Transition Trend of Local Flow For the transition tendency of the gas flow, the time-series transition of the sensor information used for the determination of the gas flow intensity can be used as it is. For example, when the temperature at a specific point tends to increase due to the surface temperature of the charging material or the temperature of the fixed sonde measured by the furnace TV, it indicates that the local flow is growing. FIG. 12 shows an example of determining the tendency of the gas flow transition depending on the fixed sonde temperature. From this figure, it can be seen that the gas flow is growing on the north side.

(3) Synthesis of intensity and transition tendency of gas flow (3-1) Synthesis of intensity and transition tendency of central flow and peripheral flow Sensor data are sensor installation status, operation status (charging, charging,
It greatly changes due to wind reduction. In addition, there are some things that show almost the same tendency such as the furnace mouth sonde temperature and the furnace mouth TV, but there are differences in accuracy, and when finalizing the gas flow situation, consider these factors well and systemize. There is a need. Here, a device with high accuracy is set as follows.

(3-1-1) Synthesis of Sensor Information Measured at Plural Points in Radial Direction Sensor information measured at plural points in the radial direction such as horizontal sonde, furnace port sonde, furnace port TV is at the center (or furnace). The accuracy of the sensor information is improved by not only judging the gas flow only by the sensor on the wall) but also by synthesizing it considering the measurement position. For example, the central flow and the peripheral flow are calculated by the following equation.

[0042]

[Equation 11]

For example, when i = 5, a1 = 0.7, a2 =
0.3, a3 = a4 = a5 = 0.0, b1 = b2 = b3
= 0.0, b4 = 0.3, b5 = 0.7, the intensity and transition tendency of the gas flow are synthesized.

(3-1-1) Synthesis of Strength and Transition Trend of Central Flow and Peripheral Flow Separate synthesis of intensity and transition tendency of the central flow and peripheral flow is performed. Table 2 shows the status of the sensor data used for synthesis. In Table 2, the larger the O mark (smaller Q), the stronger the strength and the tendency to change. The following is an example of central flow intensity synthesis.

[0045]

[Table 2]

[0046]

[Equation 12]

Here, ai is the degree of contribution of each sensor, and fi is the degree of certainty in consideration of waiting for charging and wind reduction. Figure 2 individually
It is determined as the product of (a) and (b). n is the number of O marks,
m is the number of sensors. The strength of the peripheral flow, the central flow, and the trend of the peripheral flow are determined in the same way.

(3-2) Synthesis of Local Flow Intensity and Transition Trend The sensor data largely changes depending on the installation condition and operation condition (charging standby, wind reduction, etc.). Also, furnace TV and fixed sonde temperature show almost the same tendency, but there is a difference in measurement accuracy. Therefore, when making a final decision on the gas flow, it is necessary to systematize these factors carefully. Here, a device with high accuracy is set as follows. Table 3 shows the status of the sensor data used to determine the local flow.

[0049]

[Table 3]

(3-2-1) Synthesis Method For the local flow, first, the sensor data is linearly regressed as shown in FIG. 13 to be converted into data corresponding to the position of MA, and then the local flow is synthesized by the following equation. For the position of MA (the temperature corresponding to (a) in FIG. 13), the linear regression curve ((c) in FIG. 13) of the measured temperature of the fixed sonde arranged as shown in (b) in FIG. 13 is obtained. , Obtained by reading its corresponding value.

[Equation 13] Here, aj is determined as the product of FIGS. 2 (a) and 2 (b).

(4) Standardization of Synthesis Result In this apparatus, the control is performed as close to the operator's operating experience as possible, so the synthesis value in the item (3-1) is standardized to a level of 1 to 5. did. FIG. 14A shows a standardization example. This equipment has a judgment reference value k based on the operation experience.
1, k2, k3, and k4 are decided, and the judgment is made in five stages from 1 to 5. Numerical values in FIG. 14A indicate the following states. 1: Very weak, 2: Weak, 3: Good, 4: Strong, 5: Very strong In addition, it is judged by expressing with a function as shown in FIG. 14 (b) created based on operating experience. It is obvious that more detailed judgment can be made by doing so, but here, a judgment method close to that of the operator is adopted.

(5) Determination of charge distribution action means (5-1) Radial direction distribution action means As a general method for controlling the central flow and the peripheral flow by controlling the charge distribution in the radial direction. (1) MA action pattern change, (2) SL position change, (3) Ore base change, (4) Fine grain raw material usage amount change, (5) Time-series discharge rate change, etc. Conventionally, the action means was determined only by depending on the operation policy at that time, but in order to obtain a more stable gas flow distribution in this device, the action shown in FIG. A gas flow evaluation table, and an operation evaluation table shown in FIG. 15 (b) so that the operation policy of the operator can be added,
Furthermore, the action propriety evaluation function shown in (c) of FIG. 15 was devised based on operating experience and offline experiments. The charged action means in the radial direction uses these table evaluation functions to calculate the evaluation value of each action means by the following formula. Note that Cij in FIG.
Can be determined based on operational experience and model experiments, and learning can be done by using a neural network. Aij in FIG. 15B is arbitrarily set based on the operation policy of the operator. Figure 1
In (c) of FIG. 5, it is shown that the operation becomes more difficult as bi is closer to 1.

[0053]

[Equation 14]

Here, i: action means, j: central flow (j = 1), peripheral flow (j = 2) ai: operation evaluation value, bi: action availability evaluation function value Cij: evaluation value regarding gas flow intensity , Xj: strength of gas flow Cij ′: evaluation value regarding transition tendency of gas flow, Xj ′:
This is the strength of the transition tendency of the gas flow. Then, the action means (i) that maximizes Ti is selected.

(5-2) Circumferential radial direction action means For controlling the local distribution by controlling the radial distribution of the charge, the suction condition of fuel and air in the lower part of the furnace is adjusted for each tuyere. Although a method of changing the MA pattern and a method of changing the tap hole for tapping work can be considered, a method of locally changing the MA pattern is common, and this method is used in this apparatus.

(6) Determination of charge distribution action amount (6-1) MA action pattern change (6-1-1) Central flow / peripheral flow action (a) Inference of MA pattern change amount FIG. 16 An example of inference of the MA pattern change amount will be shown. The MA pattern change amount is inferred using a strength judgment matrix and a transition tendency judgment matrix devised based on operational experience. For example, the intensity of the gas flow when the central flow intensity is +2 and the peripheral intensity is +3 is W1 from the intensity determination matrix. In addition, the transition tendency of the gas flow when the transition tendency of the central flow is +2 and the transition tendency of the peripheral flow is +4
It becomes 1. If this result is applied to the MA action decision matrix, it becomes +2. This means that only +2 MA patterns No. oriented to the central flow are selected from the current MA pattern. Also in this case, the MA pattern No. is obtained by rounding off the result determined by the function shown in FIG.
It is obvious that finer control can be achieved by determining the change amount of. In this apparatus, the result obtained by the model experiment on the distribution of the charged material is registered as the MA pattern No. from the peripheral flow toward the central flow (FIG. 18).

(B) MA Action Amount Diagnosis When the MA pattern No. is changed, the gas flow in that portion gradually changes as the charge drops. Then, it usually takes about 12 hours before the influence of the change of the MA pattern No is spread. Therefore, it is necessary to consider the operation history of the MA in order to determine the gas flow corresponding to the position of the MA in the actual operation. In addition, during tapping work, the air flow is disturbed due to a temporary large change in the blowing pressure, which is caused by a reduction in wind for several hours. It is also necessary to eliminate the factors associated with these transient factors. In this device, these elements are added as follows.

A. MA Operation History FIG. 19 shows a correction example of the MA action in consideration of the MA operation history. The horizontal axis represents time, and the vertical axis represents the change amount of the MA pattern No. For example, in the item (6-1-1) (a), the instruction of +2 as the MA pattern change amount continues for 10 hours (see FIG.
(A)), it is assumed that the MA pattern No is changed by +2 at time 0 ((c) in FIG. 19). The result then gradually appears in the gas flow as the charge drops (Fig. 1
9 (b)). In actual operation, taking this point into account, the effect of changing the MA pattern No. by +2 appears for only +1 (about 6
After 6 hours, MA pattern N will be restarted after 6 hours.
Change o by +1. Note that the MA pattern N is also used here.
If the change amount of o is ambiguous, the value obtained in (b) of FIG. 19 can be used to perform finer control, but this device is used to approximate the operation of the operator. This concept is the same for the action means described in (6-2) to (6-5).

B. Temporary factors such as blast pressure Since the wind is blown for several minutes during tapping work, the blast pressure and other factors change temporarily. Therefore, the actual action is taken only when the instruction to change the MA pattern No. in the same direction continues twice or more to eliminate the transient factor associated with the wind reduction. When the amount of residual iron is large, the gas flow is disturbed. Therefore, the residual pig iron amount is expressed by a membership function as shown in FIG. 20, and the action amount is multiplied by the value to correct.
This concept is the same for the action means described in (6-2) to (6-5).

C. Elements that cannot be judged only by the gas flow, such as composition changes, FIG. 21 (a) shows an example of preemption of elements that disturb the gas flow, such as composition changes. The horizontal axis represents time, and the vertical axis represents the amount of change in the MA pattern No. For example, if the sinter ore with a slightly smaller grain size is used instead of the direct sinter, the central flow tends to gradually collapse. For this reason, this device calculates the MA pattern No change amount (Y) based on the increase amount (X) of the usage ratio of the yard sinter, and starts several hours before the time when the yard sinter is actually used. Until the sinter melts into molten pig iron (6-1-1) (b) b. With respect to the value determined in the item (1), Y is selected as the pattern No. directed to the central flow. Here, too, if the action amount is corrected by expressing it with a membership function as shown in FIG. The concept of anticipating future events such as changing the composition is the same as in the action means described in (6-2) to (6-5).

D. Execution of action (6-1-1) (b) The result of the item of c is the display device (CR
T) and an alarm device are notified to the operator, and when the control mode is automatic, the signal is transmitted to the MA control device (PL).
It is transmitted to C), and the stroke of MA is automatically controlled, and it is linked to the distribution control of the charged material in the radial direction. Further, in the manual mode, the operator changes the stroke setting of the MA control device by looking at the display, which is connected to the radial charge distribution control. This concept is the same for the action means described in (6-2) to (6-5).

(6-1-2) Local Flow Action (a) Inference of MA Pattern Change Amount The MA pattern change amount is determined using the MA action decision matrix (FIG. 17) determined based on operational experience and model experiments. Reason. For example, if the intensity of the gas flow at a point corresponding to a certain MA is +2 and the intensity of the transition tendency is also +2, the value of the MA action decision matrix is +.
It becomes 2. This means that only +2 MA patterns No. for the central flow are selected from the current MA pattern. Here again, it is obvious that finer control can be performed by rounding off the result determined by the function shown in FIG. 14B and determining the change amount of the MA pattern No. In addition, in this device, the results obtained by the model experiment regarding the distribution of the charge are registered as the MA pattern from the peripheral flow toward the central flow in the same manner as in the action of the central flow / peripheral flow (FIG. 18). ..

(B) MA Action Quantity Diagnosis When the MA pattern No. is changed, the gas flow in that portion gradually changes as the charge drops. Then, it usually takes about 12 hours before the influence of the change of the MA pattern No is spread. For this reason, it is necessary to consider the operation history of the MA in order to determine the gas flow corresponding to the position of the MA during operation. In addition, during tapping work, air is blown for a long time to temporarily reduce the blowing pressure, which disturbs the gas flow. In this device, the elements associated with these transient factors are removed in exactly the same way as during the action of the central and peripheral flows.

(6-2) SL position change Central position / peripheral flow can be controlled by changing the position of SL. In this apparatus, the SL position change is inferred by using the strength determination matrix and transition tendency determination matrix shown in the item (6-1-1) (a) and the SL action determination matrix devised based on operating experience. For example, the central flow intensity is +2,
The intensity of the gas flow when the peripheral flow intensity is +3 is W1 according to the intensity determination matrix. Also, the trend of central flow transition is +2,
The gas flow transition tendency when the peripheral flow transition tendency is +4 is W1 from the transition tendency determination matrix. If this result is applied to the SL action decision matrix, it becomes +2. This means to select only +2 central stream oriented SL actions from the current SL. Again in FIG.
Round the results of the function shown in (b) and SL
It is obvious that finer control can be performed by determining the change amount of In this device, the result obtained by the model experiment on the distribution of the charged material is registered as the SL position change evaluation function, and the operation amount is determined from this function ((a) in FIG. 22). For example, if SL is at the reference point and the peripheral flow action is +2, increase SL by 0.8 m.

(6-3) Change of ore base By changing the ore base (or coke base), the central flow and peripheral flow can be controlled. In this equipment, the change of the ore base is inferred using the strength judgment matrix and transition tendency judgment matrix shown in the paragraph (6-1-1) (a) and the ore base action decision matrix devised based on the operation experience. .. For example, the central flow intensity is +2 and the peripheral flow intensity is +
The strength of the gas flow at 3 is W1 from the strength judgment matrix.
Becomes Further, the transition tendency of the gas flow when the transition tendency of the central flow is +2 and the transition tendency of the peripheral flow is +4 is W1 from the transition tendency determination matrix. Applying this result to the ore-based action decision matrix yields +2. This means that only +2 ore-based action pattern Nos oriented to the central flow are selected from the current composition base.
Here again, it is obvious that finer control can be performed by rounding off the results determined by the function shown in FIG. 14B and determining the ore-based pattern No change amount. In this device, the result obtained by the model experiment on the distribution of the charge is registered as an ore-based action evaluation function, and the operation amount is determined from this function ((b) in FIG. 22). For example, if the ore base is at +1 position from the reference and the peripheral flow strength is +
When strengthening by 1, change the ore base from +1 to +1.6.

(6-4) Change in amount of fine grain raw material used The central flow / peripheral flow can be controlled by changing the amount of fine grain raw material used. In this device, the amount of fine-grain raw material used is (6-1-1) (a)
Reasoning is performed by using the strength judgment matrix and transition tendency judgment matrix shown in the section (4) and the action decision matrix for fine grain raw material usage devised based on operational experience. For example, when the central flow intensity is +2 and the peripheral flow intensity is +3, the intensity of the gas flow is W1 from the intensity determination matrix. Further, the transition tendency of the gas flow when the transition tendency of the central flow is +2 and the transition tendency of the peripheral flow is +4 is W1 from the transition tendency determination matrix. If this result is applied to the fine grain raw material usage action determination matrix, it becomes +2. This means that only +2 action patterns No. of the fine grain raw material usage oriented to the central flow are selected from the present ore base. Here again, it is obvious that finer control can be performed by rounding off the results determined by the function shown in FIG. 14 (b) and determining the pattern No change amount of the fine grain raw material usage amount. In this apparatus, the result obtained by the model experiment on the distribution of the charged material is registered as the action evaluation function of the amount of the fine grain raw material used, and the operation amount is determined from this function ((a) in FIG. 23). For example, when the fine grain raw material usage amount is at the reference point and the peripheral flow action is taken by +2, the fine grain raw material usage amount is increased by 1.6%.

(6-5) Change of time series discharge pattern If the time series discharge rate of ore and coke is changed, the central flow
The peripheral flow can be controlled. In this device, the time-series emission rate is changed by using the strength determination matrix and transition tendency determination matrix shown in (6-1-1) (a) and the time-series emission action determination matrix devised based on operating experience. Reason. For example, when the central flow intensity is +2 and the peripheral flow intensity is +3, the intensity of the gas flow is W from the intensity determination matrix.
It becomes 1. Further, the transition tendency of the gas flow when the transition tendency of the central flow is +2 and the transition tendency of the peripheral flow is +4 is W1 from the transition tendency determination matrix. If this result is applied to the time series emission action decision matrix, it becomes +2. This means that only +2 pattern Nos directed to the central flow are selected from the current time-series discharge pattern. Here again, it is obvious that finer control can be performed by rounding off the result determined by the function shown in FIG. 14 (b) and determining the time series emission pattern No change amount. In this device, the result obtained by the model experiment on the distribution of the charged material is registered as a time-series discharge rate evaluation function, and the operation amount is determined from this function ((b) in FIG. 23). For example, when the discharge speed is +1.7 seconds from the reference value and the central flow action is taken by +1, the discharge speed is increased from +1.7 seconds to +4 seconds.

(7) Operation record FIG. 24 shows the transition of the operation index after the control by this apparatus is started. It can be seen that the realization of the control by this device makes the gas flow distribution in the radial direction and the circumferential direction appropriate, and the effect on the operation side is gradually expanded.

[0069]

INDUSTRIAL APPLICABILITY As described above, according to the present invention, the charge distribution control in the radial direction and the circumferential direction, the central flow / peripheral flow and the local flow are advanced for the purpose of standardization and transmission of operation technology and automation of operation. It is a systemization of various operational knowledge, and has the following effects. (1) High-level standardization and transmission of operation technology (2) Reduction of operator workload, automation of operation (3) Optimization of gas flow judgment by advanced processing of sensor information (high accuracy) (4) Real-time monitoring and control Realization of (5) (As a result of (1) to (4) above) Realization of a control that is narrower than the conventional operator control (6) (As a result of (1) to (4) above) Operation effect shown in FIG. Confirmation of

[Brief description of drawings]

FIG. 1 is a block diagram of a blast furnace charge distribution control device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a decision function of an exponential smoothing constant.
(A) is the charging waiting time, (b) is the wind reduction time, (c) is the number of actions, (d) is the number of SL actions, (e) is the number of ore-based actions, and (f) is the amount of fine grain raw material used. Number of change actions, (g) is the number of time-series discharge rates, (h)
Is the load distribution action count.

FIG. 3 is a diagram showing an example of determination of gas flow intensity.

FIG. 4 is a diagram showing an installation example and a data example of a horizontal probe used for determination of a central flow and a peripheral flow.

FIG. 5 is a diagram showing an installation example, a data example, and a processing data example of a furnace port sonde used for determination of a central flow and a peripheral flow.

FIG. 6 is a diagram showing an example of raw data and an example of processed data of a furnace mouth TV used for determination of central flow / peripheral flow.

FIG. 7 is a diagram showing an arrangement example and measurement data example of a fixed sonde used for determination of central flow / peripheral flow.

FIG. 8 is a diagram showing an example of gas flow rate determination based on other information used for determination of central flow / peripheral flow.

FIG. 9 is a diagram showing a furnace port TV used for determining a local flow and a processing example.

FIG. 10 is a diagram showing an arrangement example of fixed sondes used for determination of local flow and an example of measurement data.

FIG. 11 is a diagram showing a sensor used for determining a transition tendency of a gas flow and a determination example (Examples 1 to 4).

FIG. 12 is a diagram showing a sensor and a determination example (example 5) used for determining a transition tendency of a gas flow.

FIG. 13 is a diagram showing an example of temporary regression of fixed sonde temperature data.

FIG. 14 is a diagram showing an example of standardization of a synthesis result.

FIG. 15 is a diagram showing a judgment method of a charge distribution action means.

FIG. 16 is a diagram showing an inference example of MA actions for a central flow and a peripheral flow.

FIG. 17 is a diagram showing an example of inference of MA action for a local flow.

FIG. 18 is a diagram showing a MA pattern No registration status.

FIG. 19 is a diagram showing a correction example in which MA operation history is added.

FIG. 20 is a diagram showing an example of a correction function for the amount of residual pig iron.

FIG. 21 is a diagram showing a preemption example of a future event such as a combination change.

FIG. 22 is a diagram showing an SL position change evaluation function and an ore-based action evaluation function.

FIG. 23 is a diagram showing a fine grain raw material usage evaluation function and a time-series discharge rate evaluation function.

FIG. 24 is a diagram showing an operation result of the example.

FIG. 25 is a diagram showing a charge distribution control and a situation in a furnace.

FIG. 26 is a diagram showing an installation state of MAs.

[Explanation of Codes] 10 Blast Furnace 20 Process Computer 30 Knowledge Treatment Computer

Claims (3)

[Claims] 1. A means for registering / correcting knowledge about blast furnace operation as a knowledge base, a means for fetching data from various sensors installed in the blast furnace into a process computer at a predetermined cycle, and noise included in the fetched data. And a means for inferring noise-removed data based on operating experience and a knowledge base obtained from model experiments as the intensity and transition tendency of gas flow for central flow, peripheral flow and local flow for each sensor. , The means of synthesizing the inference result as the strength and transition tendency of the gas flow based on the information on charging and wind reduction, and the knowledge base on the gas flow, and charging the synthesis result from the operation status and operation policy at that time. Movable Armor (hereinafter M
Apply to any one of the action decision matrix (referred to as A), stock line (hereinafter referred to as SL) action decision matrix, ore-based action decision matrix, fine grain raw material usage change action decision matrix, and time-series emission rate action decision matrix. A blast furnace charge distribution control device having means for determining an action quantity and means for transmitting the action quantity to a programmable controller (hereinafter referred to as PLC) to automatically control the charge distribution. 2. The action amount is corrected by taking into account transient factors such as change of charge distribution action history, change of blast pressure, amount of residual iron in the furnace, or future event such as change of composition. The blast furnace charge distribution control device according to claim 1, further comprising means. 3. The blast furnace charge distribution control device according to claim 2, further comprising means for guiding an operator of an action amount of the charge distribution.