KR970001319B1 - Method and apparatus for mixing ore - Google Patents

Abstract

Process and apparatus for blending a raw material of iron and steel by using a blending specialist system and a process computer which give a good operation and energy efficiency are described. The apparatus is formed of a blending specialist system(39)in which a blending pattern is determined by a knowledge base of an experimental knowledge, a file blending indication information(43) which is received from a factory production control business computer(37) and an equipment operation condition information(44) classified by system from a programmable controller(36) and a process computer(38) in which an equipment automatic operation control, a homogeneity identification control, and a blending result are administered by the blending pattern and the blending specialist system(39).

Description

Steel compounding device and method using expert system

1 is a steel raw material blending operation process chart.

2 is a general configuration of a conventional compound control system.

3 is an overall configuration diagram of a steel mixing apparatus using an expert system according to the present invention.

4 is a composition expert system configuration applied to the present invention.

5 is a functional flow chart of the steel mixing method using an expert system according to the present invention.

Figure 6 is a table of expert knowledge used in the formulation expert system.

7 is a constraint constraint table used in the formulation expert system.

8 is an explanatory diagram of compounding patterns as a result of inference of the compounding expert system.

9 is a measurement chart of SiO ₂ component data when the present invention is applied to a manufacturing industry.

* Explanation of symbols for main parts of the drawings

1: Imported Light Yard 4: Special Light Yard

5: Mixing file 7: Quantitative cutting device

8: belt conveyor 9: stack

11: fixed quantity control panel 17: belt conveyor operation control panel

20: stack operation control panel 38: process computer

39: compounding expert system

The present invention relates to a steel blending apparatus and method using an expert system that automatically determines the optimal blending pattern for mixing and mixing the steel raw material in the raw yard.

The raw material compounding process is divided into a compounding pattern determination process and a compounding plant operation control process for automatically controlling the raw material plant according to the compounding pattern. FIG. 1 is a process diagram of a steel raw material blending operation, and FIG. 2 is a conventional overall system configuration for controlling it.

Mixing pattern determination process consists of compound pile (5) unit, and supply and demand of raw materials among about 100 brands stored in imported mine (1), domestic light (2), secondary raw material (3) and special mineral yard (4) Depending on circumstances, 15-25 species are used per compound pile (5). Normally, about 70,000 to 90,000 tons are applied, and one compound pile (5) is divided into 3-4 blocks according to the properties of ingredients, particle size, angle of repose, specific gravity, etc. . This is a job scheduling to determine the optimal loading position and loading amount for each brand by the same component or loading amount for each divided block.

However, when an experienced driver manually decides the mixing pattern and mixes various kinds of ores in the compounding pile (5), it is difficult to determine the optimum pattern for determining the loading position and loading amount in the case of the mineral species with severe component variation. It was a factor of quality instability due to component variation. In addition, it is difficult to quantitatively formulate and formulate the blending pattern because many factors to consider such as facility conditions, operation conditions, and ore constellation conditions are required to optimize the blending pattern determination. In addition, it was not possible to change the mix pattern quickly when the operating conditions changed.

In addition, in the mixing facility operation control process, after the mixing pattern is determined as described above, the programmable controller 25 is manually inputted by manually inputting the brand-appropriate quantity setting value from the first floor with the dip switch of the operation control panel 26 installed in the central cab. The digital signal is output to the output control panel 11 and set to the set value. Then, when the start push button for the facility operation start is pressed in the operation control panel 26 connected to the programmable controller 25, the belt conveyor 8 of the fixed quantity cutting device 7 is interlocked so that the fixed quantity fixed control panel 11 and the belt conveyor operation control panel ( The brands stored in the hopper 6 are sequentially started by 17), and the cutting is started in the quantitative cutting device 7 and conveyed to the stack 9 by the belt conveyor 8. At this time, the mechanical weighing device, Merrick scale 14 detects the on / time when the brand is cut off, and is converted into a digital signal (4-20 mA) by the A / D converter 15 to be compared with the set value in the PID controller 16. The amount is controlled to be cut out.

On the other hand, the driver located in the stack (9) the internal cab receives a driving start request through a mutual wired contact with the driver of the quantitative cutting device (7) of the central cab to the stack operation control panel 20 installed inside the stack (9) After the stack 9 is moved to the initial stowage position through the connected operating control panel 27, the machine is started and started at the time when the brand is cut out from the metering device 7 and arrives at the stack 9 Pressing the pushbutton starts loading operation and receives the limit switch type position detector 21 signal installed at both ends of the compounding head (5) from the relay board 22, and the stack 9 receives the compounding file (5). When the first floor loading amount on the mixing pattern is reached while the loading of the head and the tail portion of the head) is reached, the operation operation panel 26, 27 presses the equipment operation termination pushbuttons to perform loading completion processing. Thereafter, the above operation is repeated until the last floor of the compounding pattern to complete the loading of one compounding file 5.

Therefore, the stack 9 cannot be operated manually by the on-site driver located inside the stack and cannot detect the driving position continuously, and the stack 9 is not systematically online with the quantitative cutting device 7. 9) It is difficult to control the homogenization load that automatically adjusts the spirit cutting according to the driving position of 9) so that the loading work is terminated at the head and the tip of the mixing file 5. There is a problem that the quality of the compounding light is degraded and the manpower is consumed because the components in the longitudinal direction of the compounding pile (5) are not uniform due to the dropping of the pile due to the completion of the loading in the middle of the).

An object of the present invention is to determine the optimal mixing pattern with the compounding expert system 39 to minimize the variation in the composition of the mixed light, which is the most important in the steel raw material mixing operation, to produce high quality mixed light and also to respond quickly to changes in operating conditions. To maintain the concurrency of planning and production, and to automatically control the mixing facility with the process computer (38), to provide a steel mixing device and method using an expert system to improve operation efficiency and contribute to performance through labor reduction. It is.

Hereinafter, the present invention will be described based on the accompanying drawings.

3 is an overall configuration diagram of the steel mixing apparatus using the expert system according to the present invention, which is a file mixing instruction information 43 (see FIG. 4) and a programmable controller received from a steel mill production control business computer 37 (see FIG. 4). 36) and the set value by the compounding expert system 39 and the compounding pattern determined by comprehensive inference of the compounding pattern based on the knowledge base composed of the equipment operation status information 44 (see FIG. 4) and the expert's empirical knowledge of each device. Process computer (38) that manages automatic operation control, homogenization loading control, and mixing performance according to the equipment operation status information of each equipment, and displays the operating status monitoring and operation performance data of the mixing pattern and various devices on the CRT screen. The VDU 40 is connected to a computer network to form an upper system, and the lower system receives the operation status signal (4-20mA) from the sensors for each device. The programmable controller 36, which transmits to the process computer 38 and receives the operation command signal of the process computer, performs start and stop processing for each unit, and detects the cut-off weight of the brand cut out from the hopper 6. 38) to the operation command signal of the quantitative control panel 11 and the process computer 38, which perform the quantitative cutting control through the PID control panel 16 by comparing the set amount automatically calculated and transmitted and the current working amount. The belt conveyor control panel 17 which performs the operation and stop control of the belt conveyor 8, the rotary encoder position detector 29 which continuously tracks the travel position of the stack 9, and the RPM signal detected by the said detector digitally. Programmable controller 33 for handling the starting and stopping of the stack driving motor 23 according to the position signal input from the synchro transmitter 30 for converting to (4-20 mA), the stack running Induction wireless cable for wirelessly transmitting the operation status signal such as the value signal and the start stop and the operation command signal of the process computer 38 between the stack operation control panel 20 of the stack cab and the programmable controller 36 of the central cab. 34 and the wireless antenna 31 and the signal converters 32 and 35 for converting the wireless digital signals are interconnected.

The configuration and operation of the formulation expert system applied to the present invention will be described in detail with reference to FIG. 4, in which a driver receives a business computer 37 for a steel production project from a VDU 40 which is an operating CRT of the process computer 38. When the file compounding instruction information is checked and the reasoning request is clicked with a mouse, the data preprocessing process of the process computer 38 is started to request the reasoning to the expert system 39. In this case, the file compounding instruction information 43 necessary for the reasoning is performed. The compounding expert system (39) through the computer network for the inventory information of the yards (1, 2, 3, 4) by the type of light based on the brand-specific property information and the mixing facility operation status information (44) received from the programmable controller (36). After that, the expert system converts the information received from the process computer 38 into object type data, and the order of severe fluctuations in components for each species (special → additives → domestic light → imported light) and ores granted to prioritize reasoning as a large order SiO ₂ component by going to review the knowledge base consisting of constraint constraint functions consisting of experiential knowledge of experts based on the importance of knowledge by The optimal solution is searched by the weight of the constraint function, and the optimal load position and load setting value for each brand are transmitted to the process computer 38 to guide the CDU VDU 40 to the programmable controller 36. The setpoint value is used as setting data for compounding plant operation control to automatically operate the compounding plant (7, 8, 9).

After the compounding pattern is determined as described above, automatic operation and homogenization loading control of the compounding facility is performed by the process computer 38 from the first floor.

The driver located in the central cab checks whether there is an abnormality of the facility on the compounding operation comprehensive screen of the V.D.U 40 connected to the process computer 38 and clicks READY with the mouse if there is no error.

When the operation READY is clicked, the process computer 38 uses the programmable controller 36 to determine the appropriate amount of each brand through the programmable controller 36, using 4-5 brands as a unit of work from the first floor of the compounding pattern determined by the expert system 39. A digital signal is outputted to (11) to set the amount of cutout to the quantitative cutting device (7), and the stack operation control panel (20) outputs the stack (9) initial loading position signal to initialize the stack (9). Move up After the operation preparation operation is completed as described above, the driver checks the equipment operation preparation completion state on the screen of the VDU 40 and clicks the operation START with a mouse, and the metering device 7 and the belt conveyor 8 interlock with each other. The brands stored in the hopper 6 are sequentially started by the exit control panel 11 and the belt conveyor operation control panel 17, and the cutting is started by the quantitative cutting device 7 and the stack 9 is carried out by the belt conveyor 8. Is carried by.

At this time, the loader cell 28, which is an electronic weighing device, detects the tone / time when the brand is cut out and is converted into a digital signal (4-20 mA) by the A / D converter 15 to be compared with the set value in the PID controller 16. Quantitative cutting control is performed.

On the other hand, the stack (9) waits until the brand is cut out from the quantitative cutting device (7) at the initial site and arrives at the stack (9), and the brand is placed on the brand arrival detection sensor installed in the stack (9). Is detected, the stack driving control panel 20 receives the driving position signal from the rotary encoder position detector 29 which continuously tracks the stack traveling position in units of 1 cm. When the first floor quantity on the mixing pattern is reached while reciprocating between the US and the US, the process computer 38 sends the floor end signal through the programmable controller 36 to operate the fixed quantity cutting control panel 11 and the belt conveyor. The stacking operation of the first floor is completed by outputting to the control panel 17 and the stack operation control panel 20 to stop the equipment operation sequentially. Thereafter, the above operation is repeated to the last floor of the compounding pattern to complete the compounding process of the compounding file 5.

In addition, once the compounding pile 5 is properly applied, the compounding light is emitted in the longitudinal direction by the recreational device 10, which is a dispensing device, and is sent to the sintering plant. Therefore, the uniformity of the longitudinal components of the compounding pile 5 is very important. Do. To this end, in order to control the loading and the end portion of the compounding pile (5) to always start and end the loading work through the rotary encoder position detector 29 to continuously track the running position by the stack (9) (5) When the loading operation is performed while repeatedly reciprocating the head and the tail portion, tracking the amount of cut out of the quantitative cutting device 7 from the start of loading on the process computer 38 to the present time from the loader cell 28. Based on this value, the planned residual amount to be loaded in the future is calculated, and if it is difficult to finish loading at the head and tail of the compound pile (5) with the current cut amount set value, By changing the cutting amount, the stack 9 always controls the homogenization of loading so that the loading can be terminated at the tip or the tip of the tip. Day 5 and promote uniform component in the longitudinal direction.

The most essential aspect of the present invention will be described in detail the reasoning process and the storage form of the knowledge base of the compounding expert system 39 as follows.

The flowchart of FIG. 5 describes the process of inferring using the knowledge base stored in the constraint form of FIG. 7, and according to the present invention, the pro-scheduling method using a scheduling expert system development technique for determining an optimal compounding pattern (PER-SCHEDULING). And Constraints Satisfaction Scheduling are applied sequentially.

The prescheduling technique is a preliminary preparation for creating a schedule line, which sets an arrangement strategy for variables and variable values. The variable declaration and initial value setting step (41), the inference request data reception and the object change step (42), and each brand Inference priority determination step 43 is divided.

This arrangement strategy has a decisive effect on the schedule performance and the number of backtracking occurrences. As shown in FIG. 5, first, the inference request data receiving and object changing step 42 receives the inference request data of the process computer 38 through network transmission and converts the inference request data into object type data, which is a language that can be inferred. .

File Combination Instruction Data: Brand Name, Accumulated Total Amount by Brand, Accumulation Period, and Accumulation Pattern (Choose from 4 types)

Property data by brand: SiO ₂ component, specific gravity, particle size, angle of repose, adhesion, mineral species and yard stock

* Mineral species (special ores, subsidiary materials, domestic or imported)

Combined pile performance data: brand name and quantity applied by floor, stack operation status

In the inference priority ranking step 43 by brand, an initial value is set in a parameter table of various objects and constraints included in the knowledge base, and the inference priority of the brand to be scheduled is set to have a large SiO ₂ component in which the components of each brand are unstable. After determining the order of mineral instability, subsidiary materials, domestic light, and imported light, the order and quantity of parts to be applied to the floor are calculated and stored in the order list buffer.

The number of split orders and loads is calculated by looping the calculation tasks by the number of floors of the blended pattern yarn using the following formula.

In addition, a GANTT-CHART-BOOY screen for displaying the inference result of FIG. 8 form is created for the screen inquiry for the driver to easily see the inference result processing object.

Constraint Satisfaction Scheduling Scheme is the most essential step to specifically set the value of each variable according to the arrangement strategy set in the prescheduling step, which corresponds to the comprehensive reasoning step 44 of FIG.

When a variable is selected according to the array priority, a higher priority value is assigned among the values satisfying the constraint constraint function associated with the selected variable according to the weight of the constraint function. This reasoning process is repeated in variable order until all variables have a value.

If in the process of inferring a value is assigned according to priority, if a specific variable is not satisfied with the related constraints, it is best to assign the value of other variables before the current variable. We thought that we couldn't allocate it, so we applied the backtracking technique that tries to reassign it first, and then skipped if it doesn't work out.

By inference by the inferential priority stored in the order list created in the inferred right decision step by brand (43), it determines the placement position of each floor with respect to the order, which can be called the final target variable. A parser array buffer for function check and adjustment is generated as many as the number of floors of the blending pattern, and the floor to which the highest value (weight) is assigned to this floor array is determined as an appropriate position.

First, the value of 1-10 should be set so that the middle value is large and the value at both ends is reduced based on 1/2 of the length of the passivation array so that the SiO ₂ component having a very unstable ore phase as the default value is located in the middle of the floor. Set the weight value as follows:

Number of floors 1 2 3 4 5 6 7 8 9 10 11 12

Array value 1 2 3 4 5 6 6 5 4 3 2 1

After that, all 20 constraint constraint functions are checked for each order with high priority of the order list, and if the relevant conditions are satisfied, the value specified in the number of priliters (11 in Fig. 7) of the corresponding constraint is determined. The floor with the largest weight is added to the buffer as a weight to determine the appropriate position of the order, and then the order is removed from the order list, placed in the schedule list, and converted into a schedule bar. The floor on the GANTT-CHART-BODY screen is displayed. Mark on.

Repeat this process for all variables registered in the order list to complete the appropriate location determination by brand.

FIG. 6 shows an empirical knowledge specification of a driver applied to the compounding expert system of the present invention, and FIG. 7 shows an example of a method of storing it in the form of a knowledge base.

The driver's empirical knowledge related to the compounding pattern determination process can be roughly divided into operating conditions, equipment conditions, brand characteristics, and other types of loading conditions as shown in FIG.

The operating conditions are 25 or less brands per batch (5), and the suitability type is block-floor, 3 block-12 floor, 3 block-9 floor or 2 block-10 floor And loading conditions are equally stacked.

The equipment conditions should be less than or equal to the number of floor-mounted suitable brands, and check the light type yards (1-4) in which the brand to be loaded in the hopper 6 is stored. There are conditions to assume.

There are conditions in which starbursts that do not have stickiness by brand are not combined with each other, and brands with a large proportion are minimized per floor, and brands with a large angle of repose are imported with.

Other types of loading conditions are that special lights do not overlap on the same floor, but are stacked in the middle of the compound pile (5), and even if a small amount is divided into two or more floors.

These empirical knowledge is stored in the form of constraint function, which extracts and formalizes the knowledge of experts. FIG. 7 shows an example of converting expert knowledge of adhesion conditions among floor properties in the form of a constraint constraint function in the following case of brand properties. It is a knowledge base expression form that does not combine with special light by causing mixed bushing chemical reaction on the same floor and causing component change.

here

CONSTRAINTS NAME represents each constraint name.

The CONSTRAINTS TYPE is divided into two types: hard and soft. If it is hard, it is a necessary condition. If it is soft, it shows a good condition.

CONSTRAINTS CLASS is a rangepoint class.

CONSTRAINTS SOURCE is a kind of constraint function.

CONSTRAINTS VALUE is the value of the constraint function.

VALUE CONDITION is more, less, equal, not-equal, no-mean.

COUNTER is the same number of constraints allowed.

APPLICATE TYPE is absolute: a numerical value in absolute meaning

relative: Number of relative meaning.

APPLICATE RANGE indicates that the range to which the constraint function applies is full, except, or only.

APPLICATE VALUE indicates the value of the applicator type.

PRIORITY is meaningful only when the constraint type is soft and can be weighted in the range of 1-100, with larger numbers indicating higher weights.

This is shown in the general form as follows.

As such, the brand appearance conditions are realized in the form of constraint constraint function, which is composed of 18 kinds of brand size, specific gravity, repose angle, and adhesiveness.

The effect of the present invention is to determine the optimum compounding pattern by the compounding expert system 39 in producing high quality compounding light by mixing and mixing various kinds of cheaper materials in the process of manufacturing the compounding light of steel raw materials and according to this pattern. The process computer 38 realizes the complete automation and homogenization control of the compounding facility, thereby improving production line and reducing the manpower and work time through high quality compounding light.

9 is data obtained by analyzing the results of applying the present invention to the manufacturing industry based on the SiO ₂ component.

The ultimate goal of the branding operation in the steel industry is to minimize the fluctuation of the SiO ₂ component. If the fluctuation of the SiO ₂ falls by 0.05 as shown in the result of the above operation, the cost reduction effect may actually be hundreds of millions.

Claims (2)

In the steel raw material blending method, the variable declaration and initial value setting step, receiving the inference request data from the process computer by network transmission and converting the object type data into the inferred language as follows; File Combination Instruction Data: Brand Name, Accumulated Total Amount by Brand, Accumulation Period, and Accumulation Pattern (Choose from 4 types) Property data by brand: SiO ₂ component, specific gravity, particle size, angle of repose, adhesion, mineral species and yard stock * Mineral species (special or minor, domestic and imported) Combination pile performance results data: brand name and quantity applied by floor, stack operation status. Set the initial value in the parameter table of the various objects and constraints included in the knowledge base, determine the inference priority of the brand to be scheduled in order of unstable component, and divide the order to be applied to the floor in the order of the highest priority. Computing the number and loading amount in the order list buffer, and inferring by the order of inference priority stored in the order list created in the above step to determine the placement position of each floor with respect to the order, which can be called the final target variable. Comprehensive reasoning step per order that creates the number of floors of the mixing pattern for the constraint constraint function check and adjustment for the target order and decides the floor that is assigned the most weight to the array for each floor. Expert system characterized in that provided Steel raw material blending method using. In the steel raw material mixing device, a compounding expert system for determining a steel material compounding pattern and a compounding pattern determined by the compounding expert system are calculated, and automatic control of the facility is performed according to the equipment operation status information for each device, and homogenization appropriate control. And a process computer for managing the compounding performance, a VDU displaying the compounding facility comprehensive operation screen of the process computer and the facility operation readiness status screen as a CRT screen, and an operation command signal of the process computer to be started and operated for each unit. Programmable controller that performs stop processing and receives the operation status signal from the sensor for each device and transmits it to the process computer, and the set amount which is automatically calculated and transmitted from the process computer which is the control signal of the programmable controller and the brand cut out from the hopper Current detecting the weight A quantitative cutout control panel which compares the amount of cuts in operation and performs a quantitative cutout control through a PID control panel, a belt conveyor control panel which controls the operation and stop of the belt conveyor according to the operation command signal of the process computer, and the process computer. A steel raw material mixing device using an expert system, characterized in that it comprises a stack operation control panel for receiving the operation command signal to the wireless antenna through a signal converter and an induction wireless cable to perform operation and stop control of the stack driving motor.