JPH1157961A - Device for controlling cutting-off of cast slab in continuous casting - Google Patents

Abstract

PROBLEM TO BE SOLVED: To optimize the cutting-off of a cast slab obtd. by a continuous casting in a steelmaking process. SOLUTION: In a cast slab cutting-off control device, an evaluation function part 1, in which it is evaluated how degree is suitable to plural required lengths in the cut-off pieces obtd. by cutting off the cast slab in which plural abnormal quality parts develop, a genetic algorithmic search part 3, in which a gene-pool constituting the cast slab is made into an individual under condition of using the cut-off piece obtd. by cutting off the cast slab as gene and this individual is formed in many pieces at random, and pairing of the individuals, crossing of the gene between the paired individuals and mutation of the gene, are decided and the alternation of generation is executed and many patterns of the cut-off pieces are formed and the best evaluation function value is searched in the evaluation function part and the search is completed in a prescribed condition, and a hill-climb algorithmic search part 4, in which the further best evaluation function value is searched in the evaluation function part 1 by sequentially adjusting the lengths of plural cut-off pieces based on the best evaluation function value obtd. with the genetic algorithmic search part and the pattern of the cut-off piece, are provided.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a slab cutting control apparatus in continuous casting for automatically controlling the cutting of slabs obtained by continuous casting in a steel making process.

[0002]

2. Description of the Related Art As a slab cutting control device in the continuous casting described above, there is one disclosed in Japanese Patent Application Laid-Open No. 3-27853. This is because when a quality abnormality is found on a slab that has been poured from a tundish into a mold and cooled, a slab cutting control device that receives a signal from the process control device calculates a good piece length L from which an abnormal portion is removed, and The lower limit addition value obtained by summing the lower limit values of the cut piece lengths (designated lengths) is calculated as the lower limit addition value a
It is judged as above, and when L ≧ a, the cut pieces of the planned number (specified number) are sampled, and when L <a, the number of cut pieces obtained by subtracting one from the specified number is sampled. The control device calculates a difference value ΔL between a total value of each designated length and a good piece length L, and distributes the difference value ΔL to each cut piece. Thereby, generation of cutting debris can be prevented, and the collection rate of slab can be improved.

[0003]

However, in the above-mentioned slab cutting control device, it is good if the slab length has a sufficient margin. However, the size of the abnormal portion generated on the slab in the slab of a fixed length is small. The specified number and cutting debris must be optimized in consideration of the location, number of occurrences, and the length of each good piece on the slab. Further, from the viewpoint of quality, it is necessary to match the cut length with the designated length as much as possible even if the cut piece is within the tolerance. For this reason, it is necessary to consider various cutting patterns of the slab, and the control becomes very complicated.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a slab cutting control device in continuous casting that can easily control a complicated cutting arrangement in view of the above problems.

[0005]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention provides a slab cutting control apparatus for cutting a continuously cast slab so as to match a plurality of slabs of a specified length. An evaluation function unit that evaluates to what extent a cut piece obtained by cutting the slab having a plurality of abnormal quality portions is suitable for the plurality of specified-length cut pieces, and obtained by cutting the slab. A collection of genes constituting the slab as a solid with the virtual cut piece as a gene, and a large number of the solids are formed at random, pairing of the solids, gene prominence between the paired solids, Is determined, generation alternation is performed, a pattern of a number of virtual cut pieces is formed, the evaluation function section is searched for the best evaluation function value, and the search is terminated under predetermined conditions. A genetic algorithm search unit, Based on the value of the best evaluation function obtained by the algorithm and the pattern of the virtual cut pieces at that time, sequentially adjust the lengths of the plurality of virtual cut pieces to the evaluation function unit. A hill climb algorithm search unit for searching for a good evaluation function value. By this means, the coarse adjustment control is performed by the genetic algorithm search unit, and the fine adjustment control is performed by the hill climb algorithm search unit, so that the optimal cutting of the slab can be performed.

[0006]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing an example of a cutting optimization block of a slab cutting control device in continuous casting according to the present invention. As shown in the figure, an evaluation function unit 1 is provided for inputting calculation conditions and evaluating optimization of slab cutting. The evaluation function f is set in the evaluation function unit 1 as follows.

F = a × (number of requested cuts N / specified number No) n−bΣ | (specified length li−cut length Li) | / slab length Li = 1n′−CΣcut debris length Li ′ / cast Piece length L ... (1) i '= 0 Here, the first claim is an evaluation section for the number of cut pieces required, and designates a planned cut piece (or slab) from which one slab obtained per casting is cut off systematically. For the number No, the ratio of the number N of requested cuts regarded as the designated length of the planned cut piece among the cut lengths of the cut pieces requested to be cut is shown. Only when the cut length Li of the requested cut piece is within the tolerance αi of the specified length li, it is regarded as the planned cut piece of the specified length li. In the first aspect, the evaluation value of the evaluation function f increases as the number N of requested cuts approaches the designated number No of planned cut pieces.

The second section is a request cutting length evaluation section, and shows a cumulative value of a ratio of one cast piece to a cast piece length L between a designated length li and a cut length Li of a cut piece to be cut. Here, n is the requested number of cuts (excluding the number of cut debris). The second term is an evaluation function f as its cumulative value decreases.
Evaluation value increases. The third term is an evaluation section of the cutting debris length, and shows the cumulative value of the ratio of the cutting debris length Li ′ to the slab length L in which the requested cut piece is regarded as the cutting debris. Here, n 'is the number of cut pieces regarded as cutting debris. In the third term, the evaluation value of the evaluation function f increases as the cumulative value decreases.

A, -b, and -c are weight coefficients of the first, second, and third terms of the evaluation function f. The weight coefficient is set by inputting calculation conditions according to the operation needs. The larger the value of the evaluation function processed by the evaluation function unit 1, the better the optimization of the cutting of the slab is. In this evaluation function section 1, the value of the evaluation function is obtained by forming a pattern of a cut piece requested for a plurality of designated lengths of one cast piece. Become huge. For this reason, the genetic algorithm search unit 3 for performing the inhibition evaluation via the switch 2
Then, the hill climb algorithm search unit 4 for performing the fine evaluation searches the pattern of the cut piece.

First, the genetic algorithm search unit 3 will be described. The genetic algorithm by the genetic algorithm search unit 3 is an artificial model that simulates the mechanism of the evolution of an organism. The outline of the genetic algorithm is as follows when an organism evolves. Natural selection: only those that adapt to the environment survive.

[0011] Predominant inheritance: crosses between individuals inherit excellent traits between generations. Mutation: A mutation that is completely different in nature. The genetic algorithm will be described below using a specific example. Assuming that the slab length L obtained by casting is 50.3 m, the designated number No of the planned cut pieces (slabs) to be obtained by planned cutting of the slab is 9, the designated length li and tolerance αi of each planned cut piece are Are set as follows. The designated length li of one planned cut piece is 10 m at the maximum.

[0012]

[Table 1]

Next, one gene is made to correspond to one requested slice. One solid corresponds to one slab. In the above example, providing the nine cut pieces claimed in one slab corresponds to providing nine genes in one solid. As a first generation, one solid is composed 9
The length of one gene is variously changed, only 100 solids are formed, and the value of each evaluation function is calculated as in the following example.

Contents of gene Evaluation function value Solid No. 1 (L _1,1 , L _1,2, L _1,3 ,..., L _1,9 ) f1 Solid No. 2 (L _2,1 , L _{2, 2,} L _2,3 ,…, L _2,9 ) f2 Solid No.3 (L _3,1 , L _3,2, L _3,3 ,…, L _3,9 ) f3 Solid No.4 (L _{4 , 1, L 4,2, L 4,3} , ..., L 4,9) f4 ... solid _{No.100 (L 100,1, L 100,2,} L 100,3, ..., L 100,9) f100 Here, L _{n, 1 +} L _{n, 2 +} ... ₊ L _{n, 9} = L (= 50.3 m).

FIG. 2 is a diagram for explaining the arrangement of a gene as a cut piece on a cast piece. As shown in this figure (a), when a cut piece to be claimed as a gene Ln, _{1 ,} Ln, _2, ... _, Ln, ₉ (n = 1,100) is placed on a normal slab, As shown in FIG. 2B, for example, when there is a quality abnormality of length Δ between the genes L _{n, 2} and L _{n, 3} on the slab, the gene L _{n, 2} becomes Δ ₂ ,
Gene L _{n, 3} is shorter by delta _3. In addition, there may be a plurality of abnormal quality portions at arbitrary positions on the slab, and their lengths are also generally different.

FIG. 3 shows each solid, for example, solid No. 9 is a flowchart illustrating calculation of an n evaluation function. When genetic information is provided from the genetic algorithm search unit 3 to the evaluation function unit 1, the evaluation function unit 1 first initializes A1 = 0, A2 in steps S1 and S2 as shown in FIG.
= 0, A3 = 0, N = 0. In step S3, it is determined whether or not the gene length L _{n, i} is within the tolerance range (± αi) of the designated length L _i . In step S4, if the number falls within the tolerance range, the number N of bills cut is counted, and in step S5, A1 = N / No is calculated. In step 6, A = A2 _{+ 1} | _{Li-Ln , 1} | is calculated. In step 7, when the length L _{n, i} of the out-of-tolerance gene is smaller than L _i− βi, it is determined to be the length of the cutting residue. Here, βi is the judgment length of the cutting debris length. In step 8, A3 = A3 + Ln _{, i} is calculated based on the determination of the cutting residue length. Step S7
In the case where L _i −β _i ≦ L _{n, i} <L _i −α, L _{n, i} > L _i + α _, the gene of L _{n, i} is used as a spare material, and the process proceeds to step 6 to calculate A2. . In steps S9 and S10, this calculation is repeated, and as a result, in step S1
In the solid No. 1 using the formula (1), evaluation function f of n
n is obtained as fn = a.times.A1 / 9-b.times.A2 / 50.3-c.times.A3 / 50.3 (2).

Next, from 100 solids, for example, a roulette method (other than a random method and a substrate method).
Pairing of 2 solids × 50 sets is performed according to the above. Next, the values of the paired solid genes are exchanged according to the formal method. The solid No. 3 and the solid No. 7 that have been paired are shown below.
Here is an example of a one-point fork. Solid No.3 (L _3,1 , L _3,2, L _3,3 , L _3,4 , L _3,5 , L _3,6 , L _3,7 , L _3,8 , L _{3 , 9} ) Solid No. 7 (L _7,1 , L _7,2, L _7,3 , L _7,4 , L _7,5 , L _7,6 , L _7,7 , L _7,8 , L _{7 , 9} ) After forging Solid No.3 (L _3,1 , L _3,2, L _3,3 , L _3,4 , L _7,5 , L _7,6 , L _7,7 , L _7,8 , L _7,9 ) Solid No. 7 (L _7,1 , L _7,2, L _7,3 , L _7,4 , L _3,5 , L _3,6 , L _3,7 , L _3,8 , L _3,9 ) An example of a two-point fork is shown below.

Solid No. 3 (L _3,1 , L _3,2, L _3,3 , L _3,4 , L _3,5 , L _3,6 , L _3,7 , L _3,8 , L _3,9 ) Solid No. 7 (L _7,1 , L _7,2, L _7,3 , L _7,4 , L _7,5 , L _7,6 , L _7,7 , L _7,8 , L _7,9 ) After forging Solid No. 3 (L _3,1 , L _3,2, L _3,3 , L _7,4 , L _7,5 , L _7,6 , L _3,7 , L _3,8 , L _3,9 ) Solid No.7 (L _7,1 , L _7,2, L _7,3 , L _3,4 , L _3,5 , L _3,6 , L _7,7 , L _7,8 , L _7,9 ) Next, the gene of the mutation is determined, and the value of the gene is changed according to the degree of change. For example, a gene is randomly selected at a probability of 3% from a total of 900 genes. Selected gene value (virtual section length)
Is changed to a value of [original value] × 0.7 + [random value between 0 and 10 m] × 0.3 (3) with the degree of change being ± 30%.

The 100 solids changed in this way are regarded as 1.5th generation solids, and their evaluation functions are obtained in the same manner as described above. As the second generation, a solid is formed by combining one solid having the best evaluation function from the first generation solids and 99 solids excluding the worst one solid of the 1.5th generation. To change generations. If the evaluation function of the second generation satisfies the end condition specified below, the process is terminated. Otherwise, the above procedure is repeated, and
Generation, 3rd generation.

It should be noted that two evaluations of the first and 1.5th generations
From 100 solids, 100 solids with good evaluation may be used as second generation solids. Alternatively, 100 solids of the 1.5th generation may be simply used as the second generation. As an ending condition, the process is executed up to the generation whose number of generation alternations is specified. Or the second
If the evaluation value of the best individual of the generation is better than the specified value, the process is terminated. Alternatively, if the invariable number of the best solid is specified, that is, if the situation that the best solid of the previous generation is the same as the best solid of the current generation continues for the specified number of times or more, the process ends.

Next, when the processing of the genetic algorithm search section 3 is completed, the switch 2 is switched to the hill climb algorithm search section 4. The hill climb algorithm search unit 4 is a trial and error search, and proceeds to a certain state. If the value of the evaluation function is good in the previous step and the present step, the hill climb algorithm search unit 4 proceeds as it is, and if it is bad, returns to the original state and proceeds to another state. .

FIG. 4 is a diagram for explaining the relationship between the genetic algorithm search unit 3 and the hill climb algorithm search unit 4. As shown in the figure, the genetic algorithm search unit 3
Then, the solution space of the optimal solution of the evaluation function is limited, and the hill climb algorithm search unit 4 searches for a local solution in the limited local solution space. That is, coarse adjustment control is performed by the genetic algorithm search unit 3, and fine adjustment control is performed by the hill climb algorithm search unit 4. For example,
As described below, the hill climb algorithm search unit 4 makes the length of the cut piece closer to the designated length, thereby improving the quality.

FIG. 5 shows a hill climb algorithm search unit.
6 is a flowchart illustrating an operation example of Example No. 4; Steps
In S21, the hill climb algorithm search unit 4
Is the best evaluation function than the genetic algorithm search unit 3.
Value f_GABAnd the calculation conditions at that time. Step S22
In the search result of the genetic algorithm search unit 3
As shown in FIG. 2B, the gene L_{n, 1}
And L_{n, 2}The cutting position 1 to the right by Δ (for example, L_TwoTolerance α
_Two1/10). In step S23
And the value f of the evaluation function_GABStep S2 is applied to the condition where
The value of the evaluation function is recalculated by adding the movement condition of 2,
_HA11And In step S23, f _GHA11Is f
_GABIs determined to be better. If yes, go to step S22
Go back and repeat the above calculation to get this f_HA21Before
Times f_HA11Is determined to be better. In step S24
If the above determination is not good, move the cutting position 1 to the left by Δ
(Eg L₁Tolerance α₁ 1/10) of the
In steps S25 and S26, the same calculation as above is repeated. So
Then, in step S27, further measurement is performed at the cutting position of the next gene.
Perform the calculation.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of a cutting optimization block of a slab cutting control device in continuous casting according to the present invention.

FIG. 2 is a diagram illustrating the arrangement of a gene, which is a cut piece, on a cast piece.

FIG. 3 is a flowchart illustrating calculation of an evaluation function of each solid.

FIG. 4 is a diagram illustrating a relationship between a genetic algorithm search unit 3 and a hill climb algorithm search unit 4;

FIG. 5 is a flowchart illustrating an operation of a hill climb algorithm search unit 4;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Evaluation function part 2 ... Switch 3 ... Genetic algorithm search part 4 ... Hill climb algorithm search part

Claims (3)

[Claims] 1. A slab cutting control device for cutting a slab to be continuously cast so as to match a plurality of slabs of a specified length, wherein the slab is obtained by cutting the slab having a plurality of abnormal quality parts. An evaluation function unit that evaluates how much the cut piece is suitable for the cut pieces of the plurality of designated lengths, and a gene that constitutes the cast piece as a virtual cut piece obtained by cutting the cast piece. The collection is made a solid, a large number of these solids are formed at random, pairing of the solids, gene formalism between the paired solids, gene mutation are determined, and generational changes are performed. A genetic algorithm search unit that forms a pattern of a simple cut piece, causes the evaluation function unit to search for the value of the best evaluation function, and ends the search under predetermined conditions; and the best evaluation function obtained by the genetic algorithm. And its value Based on the virtual cutting piece pattern at the time,
A hill climb algorithm search unit that sequentially adjusts the lengths of the plurality of virtual cut pieces and causes the evaluation function unit to search for a better evaluation function value. . 2. The evaluation function unit is characterized in that the requested number of cut pieces relative to a specified number of cut pieces, the length of cut pieces relative to a specified length of the cut pieces, and the amount of cutting debris are items of evaluation. The slab cutting control device in continuous casting according to claim 1. 3. The search according to claim 1, wherein the hill climb algorithm search unit determines whether or not the value of the evaluation function is good each time the position of each cut is sequentially moved left or right. A slab cutting control device in continuous casting according to the above.