JPH0733531B2 - Blast furnace thermal controller support system - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a support system for a blast furnace heat control device for controlling the temperature of hot metal tapped from a blast furnace, and more particularly to arithmetic processing of weighting factors of various sensors.

[Prior Art] Conventionally, the temperature of the hot metal in the blast furnace is estimated and controlled.
As a control method, generally, a method in which a blast furnace operator qualitatively judges information from various sensors installed in the blast furnace to evaluate the situation of the blast furnace and optimally adjusts operation factors is adopted. ing. However, there are individual differences in the evaluation results due to the ability and experience of operators, and therefore it is difficult to standardize the operation actions and it is difficult to estimate the hot metal temperature because the evaluation is not quantitative. there were. There is a method proposed by the inventors in Japanese Patent Laid-Open No. 62-270708 in order to solve such a problem.

This method takes data from various sensors installed in the blast furnace at a predetermined timing, and based on this data, creates various data indicating the status of the blast furnace such as unloading speed, pressure loss, shaft pressure, and gas utilization rate. Then, this data is compared with the reference data, true / false data is created and temporarily stored. Then, the furnace heat level and the furnace heat transition are inferred from various knowledge bases based on the experience and results of the blast furnace and the truth data, and the action amount for the blast furnace is determined. The computer is made to do the above.

[Problems to be Solved by the Invention] In a conventional blast furnace heat control device, a total confidence is calculated based on a plurality of rule groups and the confidence of each rule, and the total confidence is equal to or higher than a predetermined reference value. When it became, it was treated as valid.

However, in actual operation, the relative position of the hot metal and the sensor may change, or the reliability of the sensor itself may change over time. There was a problem that the accuracy deteriorated.

The present invention has been made in order to solve such a problem, and represents the usefulness of the output data of the sensor by a weighting coefficient, and the weighting coefficient is obtained based on the operation results and reflected in the blast furnace control. It is an object of the present invention to provide a support system for a blast furnace heat control device that enables improvement of control accuracy.

[Means for Solving the Problem] A blast furnace heat control apparatus according to the present invention includes a data input means for receiving data from various sensors installed in the blast furnace at a predetermined timing, and a wing based on the data from the sensor. Create various data showing the condition of the blast furnace, such as mouth filling temperature, unloading speed, pressure loss, furnace top temperature, gas utilization rate, solution loss amount, etc., and compare the various data with the reference data, and compare the difference. A processing data creating means for creating data, and further, a storage means for temporarily storing the various data and difference data (hereinafter referred to as processing data), and storing processing data used in the past, and blast furnace operation Various knowledge bases based on experience, achievements, mathematical models, etc. are stored, and the knowledge base stores membership functions and various types of sensors that define the certainty factor of each rule. The knowledge base storage means containing the weighting coefficient of the server, the processing data of the storage means, the confidence level of the knowledge base of the knowledge base storage means, and the weight coefficient of various sensors are used to infer the furnace heat level and the furnace heat transition. And an inference calculation means for determining the action amount for the blast furnace.

In such a blast furnace heat control device, there is a weighting factor calculation means for obtaining the weighting factors of various sensors based on the operation results.

[Operation] In the present invention, after the processing data creating means creates various data indicating the condition of the blast furnace based on the blast furnace data from the data input means, the processing data is created based on the data. An inference operation as artificial intelligence is performed based on the processing data and the knowledge base to determine the action amount for the blast furnace.

In this inference operation, the certainty factor is calculated based on the membership function, and the certainty factor is multiplied by the weighting factors of various sensors to correct the certainty factor, and the action amount is determined based on the corrected certainty factor. Then, the weighting coefficient is sequentially marched based on the operation results.

[Embodiment] An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the concept of a blast furnace heat control device according to an embodiment of the present invention. FIG. 2 is a block diagram of a blast furnace heating apparatus according to an embodiment of the present invention. The blast furnace heat control device of the present invention comprises a data collection processing means 1, a processing data creation means 2, a processing data storage means 3, an inference operation means 4, a knowledge base storage means 5 and a weighting coefficient operation means 6.

An outline of the operation of this embodiment having the above configuration will be described.

The data collection processing means 1 is for performing input processing in time series from various sensors such as a temperature sensor, a pressure sensor, a gas sensor and the like via a data scanner (not shown) to the file means 1a.

The data collection processing means 1 includes a computing means, and the sensor data stored in the file means 1a is exponentially smoothed and then stored again in the file means 1. Then, in the processed data creating means 2, the average value thereof is calculated every 20 minutes, for example.
Also, the difference data between the average value and the reference value is obtained and sent to the processed data storage means 3.

The inference operation means 4 determines the furnace heat level and the furnace heat transition based on the knowledge of the knowledge base storage means 5 to obtain a theoretical action amount, and corrects the action amount as necessary, and a blast furnace control device (not shown). It sends the action to be taken to.

As shown in FIG. 3, the knowledge base storage means 5 includes a knowledge base 51 and a membership function knowledge base 52. The knowledge base 51 is a reactor heat level determination KS group (KS; knowledge source) reactor heat transition level determination KS group. , Action judgment KS, action correction judgment KS, comprehensive judgment KS group, operation status judgment
It is formed from units of each knowledge base of KS.

The furnace heat level judgment KS group is a knowledge base used by the inference calculation means 4 for determining what level the furnace heat level of the blast furnace is, and judges the furnace heat level with the hot metal temperature as the main judgment factor. . "Hot metal temperature-furnace heat level K
S "," sensor-furnace heat level KS ", etc. for determining the furnace heat level with the measured amount of other sensors as the main judgment factor.

In each of these KS, each measured quantity and furnace heat level are independent variables, and a combination of them is established (confidence factor = CF below).
Membership function (hereinafter referred to as value) as a dependent variable (hereinafter
MF function) and a group of rules that determine the procedure for using the MF function.

The furnace heat transition determination KS group is a knowledge base used by the inference calculation means 4 to determine what level the furnace heat transition of the blast furnace is, and the furnace heat transition is mainly based on the transition of the hot metal temperature. It includes the "hot metal temperature-furnace heat transition KS" that determines the transition of "," and "sensor-furnace heat transition KS" that determines the furnace heat transition based on the transition of the measured amount of other sensors as the main determination factor.

These KS also have MFs with each measured quantity and furnace heat transition level as an independent variable, and the probability that a combination thereof occurs as a dependent variable.
It consists of a function and a set of rules that determine the procedure for using that MF function.

The action determination KS is made up of a group of rules for determining the action amount based on the combination of the furnace heat level and the furnace heat transition level.

The action correction amount KS group is made up of a group of rules for correcting the current action amount based on the information on the actions taken in the past and the disturbances that have occurred in the past.

The comprehensive judgment KS is composed of a group of rules for determining the final action amount based on the result of the action judgment and the result of the action correction amount judgment.

Then, the operation state determination KS determines whether or not the operation of the furnace is normally performed based on, for example, the furnace heat level, and if normal, sends the above action to the control system, and if abnormal, that It is made up of a group of rules that display a message to guide the operator.

The inference calculation means 4 executes each knowledge base described above, determines the furnace heat level and the furnace heat transition, and then determines the action amount based on these judgment results. This action amount is subjected to a predetermined correction, and the result is stored in the processing data storage means 3 and is also transmitted to a blast furnace control device (not shown) to control the blast temperature, the blast humidity, etc., so that the hot metal temperature is desired. Controlled by value.

Next, the structure of each knowledge base and the outline of its inference will be described with reference to FIG.

(A) Reactor heat level judgment KS (Knowlege Source) group: This furnace heat level judgment KS group is a "hot metal temperature-furnace heat level KS" as described above in the knowledge base for judging the state of the furnace heat at the inference start time. , "Sensor-furnace heat level K
It consists of KS groups such as “S”, and as shown below, the CF value distribution is obtained by the method described below for each furnace heat level divided into 7 stages from high to low levels for each KS group, The maximum confidence level is the furnace heat level at the current time.

Furnace heat level Evaluation 7 Large heat 6 Medium heat 5 Normal 4 Small cold 3 Medium cold 2 Large cold 1 Extra large cold Here, an example of "hot metal temperature-furnace heat level KS" is explained. The KS is composed of an IF section that sets conditions and a THEN section that sets the instruction content when the conditions are satisfied.
An example is as follows.

Rule NO.1 [IF part] Pan order = 1 NOT (a large amount of residue) NOT (elapsed time after completion of wind reduction ≤ 180 minutes) Judgment of Si, S is "low" [THEN part] Hot metal by normal 3D function Obtain the CF value of the temperature-furnace heat level.

This rule No.1 is that when the blast furnace is operating in a steady state, the CF of the hot metal temperature-furnace heat level is calculated by the normal 3D function.
It indicates that the value is to be obtained. That is, the CF value of the hot metal temperature-furnace heat level is usually ~ f _MTN , somewhat high ~ f _MTH , high f
There are three types of _MTEH , and the inference operation means 4 selects which function to use by the above logic.

In other words, the hot metal temperature-furnace heat level KS is the order of the pots (the number indicating the number of the ladle used to measure the hot metal temperature from the start of tapping) and the amount of residue Which of the above-mentioned three types of functions is to be adopted is determined according to the elapsed time from.

All three MF functions of hot metal temperature-furnace heat level are functions of pan order LN, hot metal temperature MT and furnace heat level FHL.

That is, f _MTN = f _MTN (LN, MT, FHL) f _MTH = f _MTH (LN, MT, FHL) f _MTEH = f _MTEH (LN, MT, FHL) The inference calculation means 4 is a CF selected by the above rule. Regarding the value function, LN and MT are applied to the actual measurement values, and FHL is applied to the numerical values of the furnace heat levels from 1 to 7 described above to obtain the CF values corresponding to the respective furnace heat levels.

Figure 4 shows the CF value function for one pot order, and since the pot order is fixed, the CF value is a function of the hot metal temperature and the furnace heat level. For example, by rule f
_{When MTN} is selected, CF = 0.2 when the furnace temperature is 4 when the hot metal temperature is 1400 ° C, and when the hot metal temperature is 1480 ° C.
In case of the furnace heat level 7, the maximum CF value is CF = 0.4.

Next, of the sensor-furnace heat level KS, the tuyere embedded temperature-furnace heat level KS and the solution loss amount-furnace heat level KS will be described.

These KS are the rules that determine whether to use each KS, and the CF value function f _HT , where the tuyere embedding temperature and furnace heat level, the solution loss amount and furnace heat level are two independent variables. It consists of f _SL .

f _HT = f _HT (HT, FHL) f _SL = f _SL (SL, FHL) The tuyere embedding temperature HT and the solution loss C amount SL used to calculate this function are the measured values (the exponential smoothing thereof). Value or moving average) is used. Use of these functions,
An example of a rule for requesting non-use is shown below.

Rule No. 1 [IF part] NOT (there is a lot of residue) [THEN part] (1) Obtain the CF value of the tuyere embedding temperature-furnace heat level based on the tuyere embedding temperature.

(2) Obtain the CF value of the solution loss C amount-furnace heat level based on the solution loss C amount.

In this case, the inference operation means 4 uses both functions.

Rule No. 2 [IF part] NOT (there is a lot of residue) [THEN part] (1) The CF value of the tuyere embedding temperature-furnace heat level is set to [0] depending on the tuyere embedding temperature.

(2) Based on the solution loss C amount, set the solution loss C amount-the furnace heat level CF value to "0".

In this case, the CF value of the tuyere embedding temperature-furnace heat level is constant regardless of the furnace heat level.

Similarly, for gas utilization rate-furnace heat level KS, unloading speed-furnace heat level KS, furnace top gas temperature-furnace heat level KS, and blast pressure-furnace heat level KS, the certainty factor for each furnace heat level is the same. Ask for.

FIG. 5 is an explanatory diagram showing the operation of the inference calculation means 4, in which the CF value for each furnace heat level is obtained based on the hot metal temperature-furnace heat level KS, and the sensor-furnace heat level KS Based on this, the CF value for each furnace heat level is calculated. Then, the CF value of each level based on the sensor-furnace heat level KS is multiplied by the weighting coefficient α _{1 to} α ₆ of each sensor (for example, if the weighting coefficient of the tuyere embedding temperature is α ₁ , then α ₁ × Tuyere embedding temperature −
CF value of each level of furnace heat level), CF of each level after multiplication
Add values.

The CF value of each level of the sensor-furnace heat level KS thus obtained and the CF value of each level of the hot metal level KS are added. In this way, for each furnace heat level (7-1)
CF value is calculated.

(B) Furnace heat transition determination KS group; This furnace heat transition determination KS group includes hot metal temperature-furnace heat transition
KS is included, and according to the degree of change from the past to the present, the furnace heat transition is divided into 5 stages from sudden rise to constant to sudden fall, and the CF value is calculated for each rank. The stage position of the value of is the furnace heat transition state at the current time.

Level Contents 5 Rapid rise 4 Upward rise 3 Sideways 2 Lower down 1 Rapid fall Here, the hot metal temperature-furnace heat transition KS is explained.

This KS is an MF function fΔ _MT = fΔ _MT (△ MT, VFHL) with two independent variables: the difference in the hot metal temperature between preceding and following taps (one tapping), MT, and the furnace heat transition level VFHL. And this FM
The function and the pre-processing and post-processing rules of this MF function are stored.

An example of how to use this KS will be described below.

Rule No. 0 [IF part] (Initial setting) [THEN part] Set the following values as the CF value of the hot metal transition.

Level 1 2 3 4 5 CF value 0 0 0 0 0 Rule NO.1 [IF part] (1) NOT (Si, S judgment is "somewhat high") (2) NOT (Si, S judgment is "high" (3) Stability flag is ON (state of furnace is stable) [THEN part] (1) △ MT = (hot metal temperature of current tap-hot metal temperature of previous tap) for each furnace heat transition level Obtain the CF value and use it as the "current hot metal temperature-furnace heat transition CF value".

(2) Next, the CF value is calculated for each furnace heat transition level as ΔMT = (hot metal temperature of the current tap-hot metal temperature of the previous tap),
"Hot metal temperature of current tap-hot metal temperature of front tap".

(3) Multiply the "current hot metal temperature-furnace heat transition CF value" and "previous hot metal temperature-furnace heat transition CF value" by weighting factors and add to the "hot metal temperature-smooth furnace heat transition CF value".

(These calculations are performed separately for each level of the furnace heat transition.) In other words, in the stable state of the furnace conditions, up to the latest data are used for estimating the furnace heat transition.

Next, the sensor-furnace heat transition KS will be described. The sensor-furnace heat transition KS is a CF value function that uses the transitions of the measured values of various sensors and the furnace heat transition as two independent variables, and the rules that determine how to use it are stored. It is provided in. That is, fΔ _si = fΔ _si (ΔSi, VFHL) (i = 1 to n: for each corresponding sensor) An example will be described in which the sensor has the tuyere embedded temperature.

Rule No.1 [IF part] NOT (there is a lot of residue) [THEN part] △ Si = (tuyere embedding temperature-60 minutes before tuyere embedding temperature) CF value of tuyere embedding temperature-furnace heat transition Ask for.

Rule No.2 [IF part] A large amount of residue CF value of tuyere transition level 4> 0 [THEN part] Overwrite setting of "0" to CF value of tuyere transition level 4.

Level 1 2 3 4 5 5 CF value * * * 0 * (* indicates that the value remains unchanged) Figure 6 shows the tuyere embedding temperature (difference from the reference value) and furnace heat transition level. Tubing embedding temperature-furnace heat CF function as an independent variable is shown.

Note that the hot metal temperature-furnace heat transition KS can be divided into “short-term transition” and “long-term transition” and set as rules.
In addition to the tuft-embedded KS, other sensor transitions such as unloading, blast pressure, gas utilization rate, solution
This information is also taken into consideration in addition to each KS such as loss amount.

The inference calculation means 4 obtains a CF value for each transition for each furnace heat transition level based on each rule of the hot metal temperature-furnace heat transition KS, and also for each sensor based on each sensor-furnace heat transition KS rule. The CF value for the transition of is calculated. The CF value of each sensor is multiplied by the weighting coefficient α _{1 to} α ₆ in the same manner as in the case of the furnace heat level. Then, the CF values of these KS are added for each furnace heat transition level to obtain the CF value of each furnace heat transition level.

(C) Action judgment KS; This action judgment KS is a knowledge base for determining the action to be taken by obtaining the furnace heat state at the current time on a matrix having the furnace heat transition and the furnace heat level as axes.

The inference operation means 4 is based on the above action determination KS,
The product of the CF value of the furnace heat level and the CF value of the furnace heat transition is calculated and written in the matrix.

FIG. 7 shows an example thereof, and in this example, the peak (maximum value) of the CF value is that the furnace heat level = 4 and the furnace heat transition = 3. The action type and the amount of action at each position on the matrix are stored as knowledge in the frame in advance.

FIG. 8 is a diagram showing an example of the action type. FIG. 9 is a diagram showing an example of the action amount.

When adopting an action type action at a predetermined position, it is necessary that the CF value reaches a predetermined value. Further, in principle, all of the action amount is automatically controlled, but a part of the action amount can be manually controlled (for example, the action amount G in FIG. 9).

(D) Action correction amount determination KS group: In this action correction amount determination KS group, the action or disturbance taken in the past is determined, and the correction action amount is determined in consideration of the influence amount at the present time. Includes various KS for making decisions. The contents are composed of rules and the like which detect and respond to changes in manipulated variables such as blast humidity, blast temperature, liquid fuel, and coke ratio, and disturbances such as coke water content and falling of adhering substances.

For example, when the blast humidity is changed, the change time and the changed amount are automatically detected by the "manipulation amount change detection" rule, and the subsequent influence amount is considered as a function of time by the "blast humidity" rule. In addition, when the deposits on the furnace wall fall off, the “falling wall” rule automatically determines the amount of influence on the falling place, the heat on the furnace and the tuyere descent time, determines the operation time and amount of the preliminary action, and makes a correction calculation. Incorporated into.

The inference operation means 4 executes the above rule to obtain the necessary correction action amount and operation time.

(E) Comprehensive judgment KS; This comprehensive judgment KS is a knowledge base for comprehensively judging the amount of action to be taken based on the judgment results of (C) and (D). Then, by the inference operation means 4, this KS
When the determination result is obtained by inferring, the determination result is input to the operating state determination KS to determine the operating state, and is displayed on a CRT (not shown) to instruct the operator to take an action amount, and at the same time, guidance is given. A predetermined automatic control is performed by feeding back to the digital instrumentation device.

By the way, the learning control method described below is executed because it is difficult to uniquely adjust and set the CF value and the weighting factor α described above.

FIG. 10 is a flowchart showing the processing of the learning control method. If the hot metal temperature has reached a predetermined target value, and if it has reached the target value, it is determined whether the size of the standard deviation is below a predetermined value, and if it is below a predetermined value. For example, the current state is maintained (case A), and if the standard deviation exceeds a predetermined value, the sensor weighting factor α is changed (case B). If the hot metal temperature does not reach the predetermined target value, it is determined that the certainty factor is inappropriate, and the MF function that defines the certainty factor is not appropriate. Rebuild the MF function.

Here, the method of constructing the MF function will be described first.

FIG. 11 is an explanatory diagram of a method for constructing a sensor-level three-dimensional MF function.

(1) The hot metal temperature pattern changes depending on various conditions as shown in FIG. Therefore, first, only the measured values in the stable region are extracted.

(2) Correlation analysis is performed between the hot metal temperature group selected in (1) above and the sensor data group.

(3) A linear regression analysis by the least squares method is performed using the data group for which the highest correlation function is obtained in (2) above, and the obtained straight line is used as the standard line (line C).

(4) The frequency of a limited area near the standard line is projected on the xy plane, and the linear regression analysis by the least square method is performed between the maximum frequency and the sensor data to determine the straight line z1z2. This straight line z1z2 is moved to the straight line C to form the line A.

(5) The line B and the line D are determined by the degree of variation in the data group.

The knowledge base (51) is prepared by adjusting the straight line group constructed with the above steps using the production rule in the same way as the MF function created by the conventional operator so that there is no unnatural intersection of each straight line. To store. The reconstructed MF function is used in the subsequent arithmetic processing.

Next, a processing method for changing the weighting factor (case B) will be described.

In the case B, the average value of the hot metal temperature during the learning period is within the target value, but the variation is large, and at least the following two factors are considered as possible factors in this case.

(A) When there is a hot metal temperature deviation at each tap.

(B) When fluctuations occur due to defective furnace thermal action.

FIG. 13 is a flowchart showing a data collection method for changing the weighting coefficient α in consideration of the above factors,
The points will be calculated for each item described below.

First, i = 1 is set as a process for tap No. 1 (S11), and it is judged whether or not the target hot metal temperature is reached (S12). Add 1 point to the "Maintenance" item (S13).

If it is determined that the target hot metal temperature has not been reached, it is next determined whether or not the action instructions other than CR have been ignored within 3 taps (S14). +1 point for the item of "" (S15).

That is, even when the fully automatic control is performed, an action that does not depend on the action instruction of this system or the instruction may be ignored. Furnace heat fluctuations caused by these are not directly related to the function of this system and must be removed from the learning target. From this point of view, the point of “Ignore action” is added every time action is ignored, and when it becomes equal to or more than the reference value within the learning target period, detailed description is omitted in this embodiment. However, the learning is stopped.

If the action is not ignored, then it is determined whether the hot metal temperature of the tap and the target temperature are the same, low, or high (S16).

When it is determined that the hot metal temperature of the tap is the same as or lower than the target temperature, the hot metal temperature of the previous tap is next low (hot metal temperature level ≤ 4) and the target value (hot metal temperature level =
5) or high (hot metal temperature level ≧ 6) is determined (S17).

When it is determined that the hot metal temperature of the previous tap is low, the hot metal temperature of the next tap is also low (hot metal temperature level ≤ 4), target value (hot metal temperature level = 5), or high (hot metal temperature). It is determined which of the levels ≧ 6) (S18).

Even when it is determined that the hot metal temperature of the front tap is the same as or higher than the target value, the hot metal temperature of the hot metal temperature of the next tap is also low (hot metal temperature level ≤ 4), the target value (hot metal temperature level = 5), Alternatively, it is determined whether it is high (hot metal temperature level ≧ 6) (S19), (S20).

When it is determined in step (S18) that the hot metal temperature of the next tap is low, the following 2-A-1 process is performed (S21).

First, the average hot metal temperature level for the maximum temperature of 3 taps is calculated (S41). Then, it is determined whether or not the temperature level indicated by the total certainty factor matches the measured average hot metal temperature level (S42). If the level indicated by the total certainty factor (calculation result of each furnace heat level and the furnace heat level on the action matrix in Fig. 8) matches the measured average hot metal temperature level, the next action type is NO
It is determined whether it was an action (S43). If the action is NO, the action type is inappropriate and the action type of the level and the transition is determined to be inappropriate (+1 for the item "action type defect").
To point. ) (S44). This determination process is used to provide guidance after totaling which part on the action matrix of FIG. 8 has a bad action type.

If the action is not NO, the action amount is determined to be inappropriate and the level and transition action amounts are determined to be inappropriate (+1 point is added to the item of "action amount defect") (S45). . This judgment process is also the 8th
It is used to provide guidance after totaling which part of the action matrix in the figure has an inappropriate action amount.

If the level indicated by the total confidence does not match the measured average hot metal temperature level, the sensor k is judged to be larger or smaller than the hot metal temperature level (S4).
6).

If the sensor level is the same as or higher than the hot metal temperature level, add 1 to the item "Lower the weight coefficient of sensor k".
Point (S47). When the sensor level is lower than the hot metal temperature level, the item "increase the weighting coefficient of the sensor k" is incremented by 1 point (S48). Therefore, the weighting factor is not changed at this point, and the weighting factor is changed based on these items after the aggregation is completed.

Here, returning to FIG. 13 again, if it is determined in step (S18) that it is the target value, the process of 2-A-2 is performed (S2).
twenty two). Here, the reason for 2-A-2 is to maintain the status quo, and +1 point is added to the item "Maintain the status quo".

Furthermore, if it is determined that the value is high in step (S18), 2-A-3 processing is performed (S23).

FIG. 14 is a flowchart showing 2-A-3 processing. t It is determined whether or not there is a heat reduction action within a predetermined time (S51), and if it is determined that there is a heat reduction action, then it is determined whether the action originates from the level or the transition ( S52). This judgment is made by storing the action immediately before the action and comparing it on the action matrix.

If it is determined that the sensor is out of the level, the transition value of each sensor is obtained (S53). Then, it is determined whether or not the transition is the same as the transition of the hot metal temperature, and the one having the same tendency as the hot metal temperature is added to the item of "increasing the weighting coefficient of the sensor k".
If there is one point and the tendency is opposite to the hot metal temperature, "sensor k
Decrease the weighting coefficient of “” is added by 1 point (S54).

When it is determined that an action has occurred from the transition, it is determined that the action amount is insufficient, and the process is performed. That is, +1 point is added to the item "insufficient action amount" (S55). This factor item will also be used for guidance after completion of the aggregation.

If an action is issued by both level and transition, it is determined that the amount of action is insufficient, as in step (S55) above, and the item "Insufficient amount of action" of the corresponding level and transition is determined as a factor item. Add +1 point (S56).

If it is judged that there was no heat reduction action within the predetermined time (S51), it is judged whether the overall level of the sensor is good or not by comparing it with the level of the hot metal temperature (S57), and both do not match. That is, if it is determined that the weight is not good, the weighting factor of the sensor is checked (S58). Here, the checking of the weighting factor of the sensor means, specifically, the step (S4
7) and (S48) processing.

If it is determined to be good, the action type is changed. That is, +1 point is added to the item of "action type defect" of the corresponding level x ₁ transition y (S59).

Here, returning to FIG. 12 again, although the same processing is performed after the determinations of step (S19) and step (S20) ((S24) to (S29)), the steps (S21) to (S23) The relationship with each process is as follows.

2-B-1 processing (S24); 2-A-1 processing 2-B-2 processing (S25); 2-A-2 processing 2-B-3 processing (S26); 2-A-3 processing 2- C-1 treatment (S27); 2-A-1 treatment 2-C-3 treatment (S28); 2-A-2 treatment 2-C-3 treatment (S29); 2-C-3 treatment It is a flow chart showing details of processing (S30) when it is determined that the tap hot metal temperature is higher than the target hot metal temperature in step (S16). This process is basically the same as steps (S17) to (S29), but a judgment is made on the hot metal temperature of the front tap and the hot metal temperature of the next tap (S60) to (S63) ((S60) to (S63)) performs 2-D-1 processing (S64) to 2-F-1 processing (S70) as shown. Here, the relationship between these processes and each process 2-A-1 to 2-A-3 in FIG. 12 is as follows.

2-D-1 processing (S64); 2-D-1 processing 2-D-2 processing (S65); 2-A-2 processing 2-D-3 processing (S66); 2-D-3 processing 2- E-1 processing (S67); 2-A-3 processing 2-E-2 processing (S68); 2-A-2 processing 2-E-3 processing (S69); 2-E-3 processing 2-F- 1 process (S70); 2-F-1 process 2-F-2 process (S71); 2-A-2 process 2-F-3 process (S72); 2-A-1 process where 2-D In the -1 process (S64), t is set to i-1 to i, and the same process as the 2-A-3 process is performed.
Is set to i to i + 1, and "t" in the step (51) of the 2-A-3 processing is
The same process as the 2-A-3 process is performed by replacing the judgment process "whether there was a heat reduction action within the time" with the judgment process "whether a heat treatment action has occurred".

In the 2-D-3 process (S66), the same process as the 2-A-3 process is performed with t being set between i-1 and i tap. 2-E-3 processing (S69)
Performs the same processing as the 2-A-1 processing after limiting the target taps to i-1 and i, and then performs the same processing as the 2-A-3 processing. 2-F-1
In the processing, t is set between i and 1 to i taps and the same processing as the 2-A-3 processing is performed.

Here, returning to FIG. 12 again, when the arithmetic processing for the i tap is completed, i =
It is set to i + 1 (S31), and the processes after step (S12) are repeated again.

As described above, the arithmetic processing for each tap is performed for each sensor, each site on the action matrix, etc., and the points of each item are totaled as follows.

(1) Maintain the status quo △ △: Number of points (2) Ignore actions △ After the points for each item are collected as described above, it is examined whether or not the weighting coefficient is changed.

The change of the weighting factor is determined in the order having the highest correlation with the hot metal temperature. This relationship is determined in advance as follows.

“Increase weighting factor”; a ₁ (number of points) “Decrease weighting factor”; a ₂ (number of points). For each sensor Ask for. And Whether or not the weighting coefficient is changed is determined by the judgment of. That is, if it is larger than the reference value ε, the weighting coefficient is changed. Among the sensors to be changed, Δα is changed for the sensor having the highest correlation with the hot metal temperature. Here, if a ₁ −a ₂ is positive, the Δα weighting coefficient is increased, and if a ₁ −a ₂ is negative, the Δα weighting coefficient is decreased.

When the Δα weighting coefficient is increased, the sensor (n) having a lower correlation with the hot metal temperature than the sensor with the changed weighting coefficient is corrected by Δα and lowered. (If n is 2, do Δα / 2 each).

When the Δα weighting coefficient is lowered, it is added to those having a high correlation with the hot metal temperature.

FIG. 16 is an explanatory diagram showing the values of the weighting factors before and after the change.
FIG. 17 is a characteristic diagram of hot metal temperature and Si at that time. As is clear from these figures, the changes in both the hot metal temperature and Si are small after changing the weighting factor.
The effect of changing the weighting factor can be confirmed.

[Effect of the Invention] Since the present invention is configured to sequentially update the weighting factors of various sensors and control the blast furnace according to the output data thereof in accordance with the actual operation, it is possible to improve the management accuracy of the hot metal temperature and Si in the hot metal. Is now possible.

[Brief description of drawings]

FIG. 1 is a block diagram showing the concept of a blast furnace heat control system according to the present invention, FIG. 2 is a block diagram of a blast furnace heat control system according to an embodiment of the present invention, and FIG. 3 is a blast furnace according to the present invention. Block diagram of the knowledge base storage function of each heat control device, 4th
The figure is a graph showing the MF function for one pan order, 5th
The figure is an explanatory diagram showing the operation of the inference function, Fig. 6 is a graph showing the tuyere embedded temperature-furnace heat CF function, and Fig. 7 is the furnace heat level.
The figure which shows the matrix which wrote in the product of the CF value and the CF value of the furnace heat transition, Figure 8 is a figure showing an example of action type, Figure 9 is a figure showing an example of action amount, and Figure 10 is FIG. 11 is a flow chart showing a learning control method, FIG. 11 is an explanatory view of a method for constructing a sensor-level three-dimensional MF function, FIG. 12 is a characteristic diagram showing transition of hot metal temperature in a tap, and FIGS. A flow chart showing a data collection method for changing
FIG. 17 is an explanatory diagram showing the values of weighting factors before and after the change, and FIGS. 18 and 19 are characteristic diagrams of hot metal temperature and Si. 1 ... Data input means, 2 ... Processing data creation means, 3
…… Processing data storage means, 4 …… Inference calculation means, 5 ……
Knowledge base storage means, 6 ... Weighting factor calculation means, 51 ...
Knowledge base, 52 ... MF function creation knowledge base.

Claims (1)

[Claims] 1. A data input means for fetching data from various sensors installed in a blast furnace at a predetermined timing, and tuyere embedding temperature, unloading speed, pressure loss, furnace top temperature based on the data from the sensor. Processing data creating means for creating various data indicating the state of the blast furnace such as the gas utilization rate and the solution loss amount, comparing the various data with the reference data, and creating the difference data, the various data and Difference data (hereinafter referred to as processing data)
In addition to temporarily storing processing data used in the past, various knowledge bases based on experience, achievements, mathematical model, etc. about blast furnace operation are stored, and the confidence of each rule is stored in the knowledge base. Knowledge base storing means including a membership function that defines the degree and weighting factors of various sensors, processing data of the storing means, a knowledge base of the knowledge base storing means, its certainty factor, and weighting factors of various sensors. A furnace temperature control device for inferring the furnace heat level and the furnace heat transition based on the inference calculation means for determining the action amount for the blast furnace, and outputting the action amount, wherein the weighting factors of the various sensors are A support system for a blast-furnace heat control device, characterized by having a weighting factor calculation means obtained from an operation record.