JP3071986B2 - Cast knitting device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a cast knitting apparatus, and more particularly to an apparatus for preparing a cast knitting plan in a continuous casting machine.

[0002]

2. Description of the Related Art FIG. 6 simply shows a continuous casting process by a general continuous casting machine. Molten steel (charge) stored in a container called ladle from the converter through molten steel processing is poured into a kind of saucer called a tundish, and then passes through a mold of a predetermined width, then gradually descends and hardens. Finally, it is cut into a predetermined steel slab (slab). Normally, when performing continuous casting of two charges having greatly different mold sizes, it is necessary to insert a jig called a dummy bar between the two charges, and the casting is interrupted during the preparation. If the steel type of two consecutive charges is different, it is necessary to replace the tundish from the viewpoint of product quality, and the casting is interrupted during the operation. Therefore, in order to increase the productivity of continuous casting,
It is necessary to determine the order of charge casting so that charges of the same steel type are continuous as much as possible, and the difference in size between the front and rear charge molds is as small as possible. At this time, from the insertion of the dummy bar to the next insertion,
Think of one unit and call it a cast. FIG. 7 shows an example of one cast. The horizontal direction represents the width of the mold, and one trapezoid represents one charge. Castings are cast in order from the upper charge in the figure. Therefore, in charging and casting, the upper part of the figure is called the bottom side, and the lower part of the figure is called the top side. As shown in the figure, casting is performed from a charge data having a wide mold width to a narrow charge, and the size of the mold varies from a wide one to a narrow one even within one charge. In the figure, since the steel type differs from A to B at the second charge and the third charge, the tundish exchange is performed. Further, if the tundish is used continuously for a long time, the nozzle becomes clogged, so that the maximum time that can be continuously cast in one tundish is set according to the steel type and the size of the mold. In the tundish exchange between the fifth charge and the sixth charge in the figure, the steel type B is changed to 6 in one tundish.
This is because the continuous casting exceeds the maximum casting time. In the casting of billets with such a continuous casting machine, the continuous casting machine continuously casts the production plan (charge knitting plan) for each steelmaking furnace given by the higher-level production planning system without inserting a dummy bar. Conventionally, when determining cast units (cast units) and determining the casting order of charges in each cast (cast knitting), some simple processing (such as combining charges of the same specification) is performed. Process) was automated. Based on this result, a skilled person who was familiar with the operational constraints and know-how regarding the determination of the casting order created a knitting plan of the cast based on that experience so that the plan would be as desirable as possible. However, in this case, since most of the planning work is performed manually, a large amount of time has to be spent for one planning in a situation of small-lot production of other products. Further, there was a problem that a plan could be made only by a limited number of skilled workers. Therefore, the present inventors have developed a cast knitting apparatus capable of automatically creating a cast knitting plan using an expert system that has the know-how of an expert as a rule (Japanese Patent Application No. 5-119460).

[0003]

In the cast knitting apparatus using the conventional expert system as described above, even if a qualitatively good solution is obtained, an optimum knitting result based on a quantitative index is obtained. It was difficult. Therefore, the composition policy could not be delicately changed depending on the operation state, and it was necessary to improve the result of the expert system to a certain level manually. Furthermore, even in this case, there is a problem that the quantitative optimality of the knitting result is insufficient, and it takes a long time for improvement. SUMMARY OF THE INVENTION An object of the present invention is to provide a cast knitting apparatus capable of improving a cast knitting apparatus and automatically creating a fine cast knitting plan according to an operation state in order to solve such problems in the conventional technology. It is assumed that.

[0004]

According to a first aspect of the present invention, there is provided a data input means for inputting a plurality of charges, which are constituent units of a cast continuously cast using a mold, Calculate the priority of dividing the charge divisions of the respective charges constituting the respective casts formed by the cast knitting means and the cast knitting means for creating a cast by connecting the charges inputted by the input means, A reference division point generating means for sequentially generating reference division points representing division units with priorities; a cast after a reference division point of a priority-based cast sequentially generated by the reference division point generation means; A cast replacement result evaluation means for evaluating the result of replacing a cast created by the means with each cast after each division point; A cast knitting apparatus comprising: a cast exchanging means for exchanging casts after the reference split point and casts after each of the split points based on a result evaluated by the cast exchanging result evaluating means. Have been. According to a second aspect of the present invention, there is provided a data input means for inputting a plurality of charges, which are constituent units of a cast continuously cast using at least a mold, and a cast in which the charges input by the data input means are connected. Casting means to create
Reference division point generating means for calculating the priority when dividing the divided parts of each charge constituting each cast created by the cast knitting means, and sequentially generating reference division points representing the divided parts with priority. When the cast after the reference split point of the casts in order of priority sequentially generated by the reference split point generating means and the cast after each split point of a certain cast created by the cast knitting means are exchanged. Determining means for determining whether or not the result satisfies a predetermined constraint condition; cast replacement result evaluating means for evaluating only the result determined to satisfy the predetermined constraint condition by the determining means; Based on the result evaluated by the cast replacement result evaluation means, the cast after the reference split point and the cast after each of the split points are exchanged. It is configured as a cast organizing unit formed by and a cast replacement execution means. In addition, the above-mentioned prescribed constraints include the allowable range of the difference between the sizes of the molds of two consecutive charges, the constraints for avoiding an increase in the load on a specific facility in the upper and lower processes, and the casting conditions. Of steel that cannot be cast unless it is at the end.

[0005]

According to the first aspect of the present invention, a plurality of charges, which are constituent units of a cast which is continuously cast using a mold, are input by the data input means. A cast combining the charges input by the data input means is created by the cast knitting means. The priority at the time of dividing the divided parts of each charge constituting each cast created by the cast knitting means is calculated by the reference division point generating means, and the reference division points representing the divided parts with priority are sequentially generated. Is done. The result obtained when the casts after the reference split point of the casts in the priority order sequentially generated by the reference split point generating means and the casts after each split point of a certain cast created by the cast knitting means are exchanged is as follows. It is evaluated by the cast replacement result evaluation means.
Based on the result evaluated by the cast replacement result evaluation means, the cast replacement execution means executes replacement of the cast after the reference division point and the cast after each of the division points. By repeating the operation of partially replacing the casts in this way, the initial casts created by the cast knitting means can be automatically and sequentially improved in accordance with quantitative evaluation criteria. Further, by changing the evaluation criteria according to the operation state, it is possible to automatically create a detailed cast knitting plan according to the operation state, which is difficult to be realized only by the conventional expert system. Also,
According to the second aspect, at least a plurality of charges, which are constituent units of a cast that is continuously cast using a mold, are input as data by the data input unit. A cast combining the charges input by the data input means is created by the cast knitting means. The priority at the time of dividing the divided parts of each charge constituting each cast created by the cast knitting means is calculated by the reference division point generating means, and the reference division points representing the divided parts with priority are sequentially generated. Is done. A predetermined result is obtained when the casts after the reference split point of the casts in order of priority sequentially generated by the reference split point generating means and the casts after each split point of a certain cast created by the cast knitting means are replaced. It is judged by the judging means whether or not the constraint condition is satisfied. Of the above results, only those determined by the determining means to satisfy the predetermined constraint condition are evaluated by the cast replacement evaluating means. Based on the result of the evaluation by the cast replacement evaluating means, the casting after the reference dividing point and the casting after each of the dividing points are replaced by the cast replacement executing means. In this way, only those satisfying the predetermined constraints are evaluated, and the cast replacement is performed based on the evaluation result. A composition plan can be created.

[0006]

Embodiments of the present invention (first and second inventions) will be described below with reference to the accompanying drawings to facilitate understanding of the present invention. The following embodiment is an example embodying the present invention and is not intended to limit the technical scope of the present invention. FIG. 1 is a block diagram showing a schematic configuration of a cast knitting apparatus 0 according to one embodiment of the present invention.
Is a flowchart showing the operation procedure of the cast knitting apparatus 0, FIG.
FIG. 4 is a diagram showing a partial replacement state between casts, FIG. 4 is a table showing an improved state of cast knitting results, and FIG. 5 is a diagram showing a direct connection state between casts. As shown in FIG.
The cast knitting apparatus 0 according to the present embodiment includes a data input unit 1 (corresponding to a data input unit) for inputting data relating to the charge to be planned determined by the higher-level production planning system to the apparatus, and a data input unit 1. Charge data storage unit 2 for storing input charge data
And an expert system that uses rules based on the knitting know-how of the cast knitting operator for the charges stored in the charge data storage unit 2 to automatically create a cast in which the charges are linked or manually create the cast. A knitting unit 3 (corresponding to cast knitting means) for performing so-called knitting, a knitting result registering unit 4 for storing the progress of knitting by the cast knitting unit 3 and a final result, and a knitting result registering unit 4 In the case where one cast is divided by a certain division unit with reference to the charge data in the charge data storage unit 2 for the cast registered in the group, an evaluation function for determining the priority of division in the division unit Referring to the evaluation function registration unit 5 to be registered and the charge data storage unit 2, the evaluation registered in the evaluation function registration unit 5 is referred to. Executing the above-mentioned priority using a function to determine the above-mentioned priority, and sorting all the divided parts of all the target casts, thereby sequentially generating reference division points representing the divided parts with the above-mentioned priorities. Part 6
(Corresponding to a reference division point generating means), a candidate division point storage section 7 for storing division points of all casts that can be divided for casts registered in the knitting result registration section 4, and a division point unit. A reference division point storage unit 8 for storing the reference division points of the casts in order of priority sequentially generated by the sort execution unit 6, and a reference division point of the cast stored in the reference division point storage unit 8. The result obtained when the subsequent casts and the casts after each of the specified split points stored in the candidate split point storage unit 7 are exchanged is determined by the operation or quality in continuous casting. When there are various constraint conditions, a constraint check unit 9 (corresponding to a judging means) for judging whether or not the constraint conditions are satisfied, a constraint condition registration unit 10 for registering the constraint conditions, , The above constraints An evaluation function calculator 11 (corresponding to a cast replacement result evaluation means) for calculating an evaluation value when a part of the cast is replaced according to the candidate division point determined to be satisfied and the reference division point, and the evaluation function is registered. An inter-cast portion for performing the exchange between the cast after the reference split point and the cast after each of the split points based on the evaluation value calculated by the evaluation function registration unit 12 and the evaluation function calculation unit 11. The knitting result registration unit 4, the reference division point storage unit 8, and the candidate division unit 13 are based on the result of the improvement of the knitting plan by the replacement execution unit 13 (corresponding to the cast replacement execution unit) and the partial replacement by the inter-cast partial replacement execution unit 13. A data updating unit 14 for updating each data stored in the point storage unit 7, a data output unit 15 for outputting a finally improved cast knitting result, It is configured.

[0007] Among the above components, as the expert system used for the cast knitting section 3, for example, those described in the conventional example can be applied. In the first invention, at least the data input unit 1, the cast knitting unit 3,
A split point unit sort execution unit 6, an evaluation function calculation unit 11, and an inter-cast partial replacement execution unit 13 are provided. In addition to the first aspect of the present invention, a constraint checking unit 9 is provided. Hereinafter, regarding the operation procedure of the cast knitting apparatus 0,
A brief description is given with reference to FIG. The charge to be planned determined by the higher-level production plan creation system is taken into the apparatus by the data input unit 1 of FIG. 1 and registered in the charge data storage unit 2 (S1). If there are various constraints on operation or quality in making a casting plan, the parameters are registered in the constraint registration unit 10 in advance. The above-mentioned constraints include, for example, the allowable range of the difference between the mold sizes of two consecutive charges, constraints to avoid increasing the load on a specific facility in the upper and lower processes, and the condition that the cast must be at the end of casting. There are restrictions such as the fact that there are steel types that cannot be cast. First, for the charge of the charge data storage unit 2, the cast knitting unit 3
, A primary plan draft is automatically or manually created and stored in the knitting result registration unit 4 (S2). All casts CAST A,
The following data is created for CAST B,. [CAST A, 0, C0], [CAST A, TD
1, C1], [CASTA, TD2, C2], [CAS
TA, TD3, C3], ...

Here, TD1, TD2 and TD3 are all tundishes included in the cast CAST A, respectively, and 0 is a dummy tundish virtually set at the bottom of the cast CAST A. Cast CA
The pair of ST A and tundish TDn (n = 1, 2, 3) indicates that cast CAST A is split immediately after tundish TDn, and that the two at both ends are not actually split. I have. The third element Cn (n
= 1, 2, 3) represents an evaluation value when the cast CAST A is divided immediately after the tundish TDn, and is calculated by the evaluation function registered in the evaluation function registration unit 5 in FIG. The specific evaluation function can be represented by the following equation, for example. Cost1 = αA1 + βA2 + γA3 (α, β, γ ≧
0) Here, A1: difference in mold size at the dividing point A2: number of charges in the cast A3: casting time of the entire cast α, β, γ: weight parameter The larger the evaluation function Cost1, the higher the priority as the object of division. (That is, the initial reference target data). Furthermore, information [CAST A, 0], [CAST A, TD1], information of all division points from which the evaluation value of the division point data is removed.
[CAST A, TD2], [CAST A, TD
3],... As initial candidate target data,
(S3). Next, the division target unit sorting execution unit 6 sorts the reference target data in descending order of the evaluation value (S4). If the number of reference target data> 0, the head of the sorted reference target data is extracted and stored as a reference division point in the reference division point storage unit 8 of FIG. 1 (S5, S6). Assuming that a part of the cast is to be exchanged as shown in FIG. 3A, a candidate target division point that satisfies the constraint condition is extracted from the reference division point and the division points of all the candidate target data. (S7, S8).

An evaluation value is calculated for all of the extracted candidate target division points in accordance with the evaluation function registered in the evaluation function registration unit 12 of FIG. 1 (S9). The evaluation function Cost2 used at this time is for evaluating the degree of improvement before and after the replacement, and is represented by the following equation, for example. Cost2 = J (A, B) -J (A ', B') Here, J (A, B) is the difference between cast CAST including the reference division point before the replacement operation and cast CAST B including the candidate division point. Similarly, J (A ', B') is a total evaluation function of cast CAST A 'including the reference division point after replacement and cast CAST B' including the candidate division point. This is the partial evaluation function for one cast. The partial evaluation function J used at this time
(A, B) is represented by, for example, the following equation (partial evaluation function J
(The same applies to (A ', B')). J (A, B) = α '(I (A) + I (B)) + β' | N
(A) −N (B) | + γ ′ | T (A) −T (B) |
(Α ′, β ′, γ ′ ≧ 0) where I (X) (X = A, B) is the cast CAST
Difference in front and back of mold size at division / connection point of X, N (X)
Is the number of charges in cast CAST X, and T (X)
Is the casting time of the entire cast CAST X, α ', β',
γ ′ is a weight parameter. Therefore, the smaller the difference between the front and rear mold sizes at the split / connection point of the two casts, the smaller the difference in the number of charges between the two casts, and the smaller the difference in the casting time between the two casts, the smaller the partial evaluation function J (A, B) and J (A'B ') become smaller. Furthermore, the larger the degree of improvement by the replacement, the larger the evaluation function Co
st2 increases.

Next, a candidate that maximizes the evaluation function Cost2 is selected from the set of candidate target division points (S10). The largest evaluation function Cost2 selected here is Cost2> 0
, It indicates that the knitting result is improved. Therefore, as shown in FIG. 3, the part after the two cast division points indicated by the reference division point and the candidate division point are exchanged (S1).
1, S12). FIG. 4 shows an example of improvement of the knitting result. In this example, the reference is a case where the cast CAST1 is divided immediately after the tundish TD2. On the other hand, only the candidates 3, 4, 5, and 6 satisfy the constraint condition when the casts after the respective division points are replaced with the casts after the reference division point, and the evaluation function for the candidates 3 to 6 is satisfied. Of the values of the candidate 3 is 97. This means that, as shown in FIG. 4, the overall evaluation value is improved by 97 compared to before the replacement. Also,
Since the value of the evaluation function of the candidate 6 is negative, the replacement result does not improve the knitting result. As a result of the partial cast replacement, all the division points of the newly generated cast are added to the candidate target data, and all the division points of the lost cast are deleted from the candidate target data (S13). As a result of the partial cast replacement, all the division points of the newly generated cast are added to the reference target data, and all the division points of the disappeared cast are deleted from the reference target data (S14). Then, the process returns to step S4. Here, if the number of reference target data = 0 in step S5, a termination condition is determined (S15). As the end condition, for example, the number of repetitions (the number of times the end condition is determined) may be registered in advance. If the end condition is satisfied, the knitting result at that time is output as an improvement plan (S16), and the process ends. If the termination condition is not satisfied, the process returns to step S3. If there is no candidate satisfying the constraint condition in step S8, the process returns to step S5 without calculating the evaluation function.
The largest evaluation function Cost selected in step S11
If 2 is Cost2 ≦ 0, it indicates that the knitting result is not improved, and the process returns to step S5 without performing the partial replacement of the cast.

By repeating the above operation,
The primary planning is performed in the direction indicated by the evaluation function Cost2, that is, in the direction in which the difference in the size of the mold between the tundishes in the cast in the cast becomes smaller, and the variation in the number of charges per cast and the casting time of all the casts becomes smaller. The plan is gradually improved. Furthermore, the balance of each evaluation item is determined by the weighting parameters α ′, β ′, γ ′ of the evaluation function Cost2.
Can be adjusted by changing Of course, depending on the setting of the weighting parameters α ', β', and γ ', even if the improvement of a certain evaluation item is negative, if the other evaluation item with a large weight is positive, the cast can be partially replaced in that direction. May be done. As described above, in the cast knitting apparatus 0, a qualitatively good solution qualitatively determined automatically by an expert system in which the knitting know-how possessed by a skilled worker is obtained, that is, the primary knitting plan (or the (A primary knitting plan manually created by), another evaluation function Cost based on one cast selected based on a certain evaluation function Cost1
By repeating the operation of replacing part of the cast with the cast selected by 2,
The primary knitting plan can be automatically and sequentially improved according to the quantitative criteria evaluated by the evaluation functions Cost1 and Cost2. In addition, by changing the parameter settings of the evaluation functions Cost1 and Cost2 in the operation state, it is possible to create a fine-tuned cast knitting plan according to the operation state that is difficult to realize only with the conventional expert system. However, when there are few constraints, the evaluation function may be directly calculated without performing the constraint check. In the above embodiment, the split point of the cast is set in the middle of the cast. However, in actual use, the split point may be extended to both ends of the cast (in this case,
It is not actually split). As a result, the following can be performed in addition to the partial replacement between the casts. (1) Movement of only a part of one cast to another cast. (2) Two casts are connected in series (two casts become one cast) (see FIG. 5). In this way, the degree of freedom for improving the plan is increased, and the possibility of obtaining a better knitting result is increased. As a result, it is possible to obtain a cast knitting apparatus capable of automatically creating a detailed cast knitting plan according to the operation state.

[0012]

Since the cast knitting apparatus according to the present invention is constructed as described above, the initial cast created by the cast knitting means can be performed by repeatedly performing the operation of exchanging a part between the casts. Improvements can be made automatically and sequentially according to quantitative evaluation criteria. Further, by changing the evaluation criteria according to the operation state, it is possible to automatically create a detailed cast knitting plan according to the operation state, which is difficult to be realized only by the conventional expert system. In addition, by evaluating only those satisfying the predetermined constraints and performing the cast replacement based on the evaluation result, a fine cast knitting according to the operation state can be performed even in the continuous casting process having the above predetermined constraints. A plan can be created. As a result, it is possible to obtain a cast knitting apparatus capable of automatically creating a detailed cast knitting plan according to the operation state.

[Brief description of the drawings]

FIG. 1 is a cast knitting apparatus 0 according to an embodiment of the present invention.
FIG. 2 is a block diagram showing a schematic configuration of FIG.

FIG. 2 is a flowchart showing an operation procedure of the cast knitting apparatus 0.

FIG. 3 is a diagram showing a state in which casts are partially exchanged.

FIG. 4 is a chart showing an improved state of a cast knitting result.

FIG. 5 is a diagram showing a direct connection state between casts.

FIG. 6 is a view showing a general continuous casting process.

FIG. 7 is a diagram showing a structural example of a cast.

[Explanation of symbols]

0: Cast knitting device 1: Data input unit (corresponding to data input means) 3: Cast knitting unit (corresponding to cast knitting means) 6: Dividing point unit sorting execution unit (corresponding to reference dividing point generating means) 9: Constraint check Unit (corresponding to a judgment unit) 11 ... Evaluation function calculation unit (corresponding to a cast replacement result evaluation unit) 13 ... Partial inter-cast replacement execution unit (corresponding to a cast replacement execution unit)

──────────────────────────────────────────────────の Continuing on the front page (72) Inventor Watani, etc. 1 Kanazawacho, Kakogawa City, Hyogo Prefecture Kobe Steel, Ltd. Inside the Kakogawa Works (72) Inventor Yoshitsugu Yamada 1 Kanazawacho, Kakogawa City, Hyogo Prefecture Kobe Steel, Ltd. Location Kakogawa Works (56) References JP-A-6-328212 (JP, A) (58) Field surveyed (Int. Cl. ⁷ , DB name) B22D 11/16, 46/00

Claims (2)

(57) [Claims] 1. A data knitting means for inputting a plurality of charges, which are constituent units of a cast continuously cast using a mold, and a cast knitting for creating a cast in which the charges input by the data input means are connected. Means and means for calculating the priority when dividing the divided parts of each charge constituting each cast created by the cast knitting means, and sequentially generating reference division points representing the divided parts with priority. The point generation means and the casts after the reference division point of the casts in order of priority sequentially generated by the reference division point generation means are exchanged with the casts after each division point of a certain cast created by the cast knitting means. A cast replacement result evaluation means for evaluating the result of the cast replacement, and a result based on the result evaluated by the cast replacement result evaluation means. A cast knitting apparatus comprising: a cast change execution unit for performing a changeover between a cast after the reference split point and a cast after each of the split points. 2. A data input means for inputting a plurality of charges which are constituent units of a cast continuously cast using at least a mold, and a cast for creating a cast by connecting the charges input by the data input means. A standard for calculating the knitting means and the priorities at the time of dividing the divided parts of each charge constituting each cast created by the cast knitting means, and sequentially generating reference division points representing the divided parts with priority The dividing point generating means, a cast after the reference dividing point of the casts in order of priority sequentially generated by the reference dividing point generating means, and a cast after each dividing point of a certain cast created by the casting knitting means. Determining means for determining whether or not the result of the replacement satisfies a predetermined constraint condition;
A cast replacement result evaluating means for evaluating only those determined to satisfy predetermined constraints by the determining means; a cast after the reference division point based on a result evaluated by the cast replacement result evaluating means; A cast knitting apparatus comprising: a cast replacement execution unit that performs replacement with a cast after each division point.