JPH0722811B2 - Constrained breakout prediction method for continuous casting - Google Patents

Description

Description: FIELD OF THE INVENTION The present invention relates to predicting constraining breakouts that can occur in continuous casting.

[Prior Art] There are various types of breakouts that occur in continuous casting.
It is difficult to predict and prevent it because it suddenly occurs even if there is no apparent change in operating conditions. In other words, the restraint breakout is considered to occur when a part of the shell formed by the solidification of molten steel adheres to the mold and is torn off. There is no change above.

A method of predicting a restraint breakout is disclosed in, for example, Japanese Patent Publication No. 63-47545. in this way,
The temperature of each part of the mold is measured by a plurality of thermocouples embedded in the mold, and when the detected temperature of one thermocouple rises above the average detected temperature of the thermocouple and then decreases, the other adjacent When a similar temperature change pattern is detected by the thermocouple, it is considered that there is a possibility of breakout.

[Problems to be Solved by the Invention] However, according to the conventional constraint breakout prediction method, the prediction accuracy is relatively low, and even when the breakout does not actually occur, there is a high possibility that the breakout is erroneously recognized as the breakout. .

If the occurrence of a deterministic breakout is predicted in continuous casting, the drawing of the cast steel must be stopped immediately in order to prevent it. However, the temporary stop of the withdrawal of the cast steel involves a partial change in the material quality of the cast steel, so the portion formed by the withdrawal stop cannot be made into a product, and the yield is reduced. Therefore,
If the possibility of false recognition of breakout is high, the yield will be reduced more than necessary.

Actually, the hit-out rate of breakout in the conventional prediction method is about 2 to 4%.

Therefore, it is an object of the present invention to improve the predictive accuracy of a deterministic breakout predicting method in continuous casting, prevent breakout in advance, and at the same time prevent a decrease in yield.

[Means for Solving the Problems] In order to solve the above problems, in the present invention, the temperature at each position detected by a plurality of temperature detecting means embedded at different positions on the mold wall of the continuous casting equipment is set. In the method of monitoring and predicting the restraint breakout: From the temperature information obtained at the same timing by each of the temperature detecting means, the past and present temperature deviations at each position are detected, and the peak of the abnormal temperature rise from the temperature deviations. Detect points,
By forming a position distribution pattern of information indicating a curve connecting each detected peak point and recognizing the similarity between this position distribution pattern and a reference pattern in which the position distribution patterns actually measured in the past are stored in advance. , Predict whether there will be a breakout.

[Operation] In the present invention, attention is paid to the fact that the position distribution pattern of the occurrence point of the abnormal change in the casting steel temperature (temporary temperature rise) that appears prior to the restraint breakout becomes V-shaped or U-shaped. ing. That is, by recognizing the pattern of this position distribution pattern, it is possible to identify whether or not the pattern is a breakout pattern and predict the breakout with a high accuracy rate.

In a preferred embodiment of the present invention described later, the accuracy of breakout prediction is further improved by using the following method.

(A) The position distribution pattern to be processed is configured by synthesizing the instantaneous detection patterns obtained at different times. Although there is only a single V-shape in the instantaneous detection pattern, by superimposing the time series data, a more complicated pattern is recognized, and the accuracy of prediction is improved.

(B) Regarding the position of each element of the position distribution pattern of the detection target and the reference pattern, the position of the element in which the abnormal temperature rise is detected first is matched with a predetermined reference point. In pattern recognition, the positional deviation between the reference pattern and the recognition target pattern causes a large error in the recognition result. In continuous casting, the position where the constrained breakout occurs is indefinite, so if the position of each element of the reference pattern and the element position of each detector are fixedly associated in a one-to-one correspondence, the actual pattern position change will occur. According to the above, a position shift is likely to occur and a pattern recognition error is likely to occur, but the accuracy of breakout prediction can be improved by aligning the reference pattern with the detection target pattern.

(C) After inputting the position distribution pattern at the time of breakout actually measured in the past and the position distribution pattern at the time when no breakout occurred to the neural network, and learning them as teacher data Then, the position distribution pattern of the detection target is input and the presence or absence of breakout is predicted. By learning both the position distribution pattern when the breakout occurs and the position distribution pattern when the breakout does not occur, the occurrence rate of erroneous recognition is reduced and the predictive hit rate is improved.

Other objects and features of the present invention will become apparent from the following description of embodiments with reference to the drawings.

[Example] Fig. 5a shows the appearance of a mold used in continuous casting. A passage of cooling water is formed inside the wall of the mold, and the molten steel in the mold is cooled by the cooling water. The molten steel injected into this mold is gradually solidified in the mold as cooling progresses, and is slowly withdrawn at a constant rate from below the mold. The casting width depends on the specifications of the product, but in this example, the maximum casting width is 222 cm.

Next, the progress status of the process when the restrictive breakout occurs will be described with reference to FIG. In FIG. 6, the process is (1)-(2)-(3)-
The process proceeds in the order of (4)-(5).

The molten steel in the mold E solidifies sequentially from the portion along the inner wall of the mold which is cooled to form a shell B. Since the shell B is pulled out from the lower side of the mold, it progresses downward, grows downward, and its thickness increases. Since there is an inclusion called powder between the shell B and the mold E, the shell B is separated from the wall surface of the mold E. However, in the vicinity of the meniscus F, sometimes a part A of the shell is in the mold E. It may stick.
Since the shell fixed to the mold E is constrained and does not move, when a force for pulling out the shell B from below is applied, the shell breaks from the middle, and a part of the shell A remains fixed to the mold wall, while the shell of the other part remains. B progresses downward (1).

Aperture formed by breaking the shell (between A and B)
Then, since the molten steel directly contacts the mold wall and is rapidly cooled, a new shell C is formed at that portion (2).

The new shell C adheres to both the lower shell B and the upper restrained shell A, but the lower shell B moves downward, so that the new shell C having a small thickness is torn off by the force. , A portion Ca fixed to the lower shell B and a portion Cb fixed to the upper shell A (3).

Molten steel penetrates again between the separated shells Ca-Cb, contacts the mold wall, cools to form a new shell, and the shell is torn off again. These processes are repeated (4).

The position where the new shell is formed gradually moves below the mold. When the position reaches the lower end of the mold E, breakout occurs (5).

Fig. 7 shows the appearance of the slab in which the constrained breakout occurred during casting. Referring to FIG. 7, it can be seen that the portion where the constrained breakout has occurred forms a substantially V shape. Since the thickness of the shell is very thin at the breakout portion, the temperature at the mold wall becomes higher than usual. The portion where such an abnormal temperature rise occurs (thick portion of the shell) gradually moves downward.

Therefore, in the present invention, the restraint breakout is predicted by measuring the temperature at each position of the mold, paying attention to the portion where the abnormal temperature rise occurs, and recognizing the position distribution pattern thereof. .

In this embodiment, as shown in FIGS. 5b and 5c, a large number of thermocouples TC are embedded in each of the four faces of the mold. Specifically, a rod made of constantan is covered with an insulating material except for the tip, and the rod is inserted and fixed in holes formed at respective positions on a copper plate constituting the mold. By processing the electrical signal obtained from each thermocouple,
The temperature of each part can be measured.

In this embodiment, of the many thermocouples shown in FIGS. 5b and 5c, only those at the positions of the first to fourth stages are used, and the remaining thermocouples are not used. .

FIG. 8 shows an example of the change in temperature detected by the thermocouple at each position when the restraint breakout occurs. The eighth
In the figure, the reference point is one measurement point, the upward-related point is one measurement point located above the reference point, and the lateral-direction related point is one measurement point located laterally to the reference point. is there. In the case of a constrained breakout, when the shell breaks, the temperature rises more abnormally at that location and then drops. In FIG.
Abnormal temperature rises appear at all of the reference point, the upward direction related point, and the lateral direction related point. However, the timing at which the temperature rises differs from each other.

In this embodiment, the processing shown in FIG. 1 is performed to predict the restraint breakout. Although FIG. 1 is in the form of a flow chart, FIG. 1 shows the flow of information, and the order in which each process is executed is not accurate in FIG. 1. For example, a plurality of processes are executed simultaneously. Please note that it may occur.

Each process will be described with reference to FIG. In the process P1, the voltage output from each thermocouple attached to the mold is sampled at predetermined intervals, the obtained voltage is converted into digital data, and the data is further processed to obtain temperature information.
In addition, the time series temperature information obtained from each thermocouple is
When using it as learning data, it is temporarily stored and saved in an external storage device.

In the process P2, the time series information of many temperatures obtained in the process P1 is processed to detect an abnormal temperature rise at each point. That is, the time series change of the temperature data at each position is monitored, and it is discriminated whether or not it satisfies the condition of the following formula (1).

Tb (t) -Tba (t ') ≥Ts (1) Tb (t): Current detected temperature (instantaneous value) Tba (t'): Moving average temperature one cycle (2 seconds) before Ts: Temperature Threshold of deviation The moving average temperature is obtained as an average value of all the instantaneous temperature values detected at the past n points at each time point (in this example, n =
16:32 seconds average).

When the condition of the expression (1) is satisfied, it is further determined by the following expressions (2) and (3) whether or not a predetermined temperature drop appears within a predetermined time.

(Tb (t) −Tb (t ′)) / 2 <b …… (2) tb ₁ −tb ₀ ≦ c …… (3) Tb (t): Current detected temperature (instantaneous value) Tb (t ′) ): Previous detected temperature (instantaneous value) Tba (t '): Moving average temperature one cycle (2 seconds) before tb ₁ : Time when the condition of expression (2) is satisfied tb ₀ : Expression (1) Times when the conditions are satisfied b, c: Constants Equation (2) must satisfy the conditions continuously for a predetermined time Process P4
Then, for each of the measurement points that satisfy all the conditions of the above equations (1), (2), and (3), the time at which the time series change of the temperature peaks is identified, and at that time, The isotherm carp passing through the measurement point is obtained by the detection temperature at each measurement point and calculation.

That is, for example, as shown in FIG. 2a, an isotherm (C1) passing through a measurement point (X1) indicated by a black circle is obtained on the plane coordinates where the mold is developed in the lateral direction. Five isotherms C1 to C5 are drawn in FIG. 2a, and each of these isotherms is generated based on the temperature distributions at different time points. That is,
Since the process P4 of FIG. 1 is repeatedly executed based on the temperature distribution data at different time points, a plurality of isotherms are obtained. These isotherms are converted to 1 by the process P5 in FIG.
Overlaid on one processing plane. Thereby, for example, a group of isotherms as shown in FIG. 2a is obtained.

In process P6 of FIG. 1, an input data area (in a two-dimensional array) having cells that correspond to the respective cells of the input layer of the neural network described later in a one-to-one correspondence is generated. That is, a memory area for data arrangement corresponding to each position of the cell is generated so that each position coordinate of the two-dimensional pattern to be recognized matches each input cell of the neural network. In this embodiment, this input data area is composed of 25 (horizontal direction) × 10 (vertical direction) cells as shown in FIG. 3a.

In process P6, the position is normalized. That is, for example, as is clear from the difference between FIGS. 2a and 2b,
The position where the abnormal temperature change appears is not constant, and this distribution pattern occurs in an arbitrary partial area on the developed mold plane. Generally, in pattern recognition, if the reference pattern and the recognition target pattern are misaligned, the correlation between the two becomes small and recognition is difficult. Therefore, in this example, as shown in FIG. 9, for example, the lateral position of the pattern is aligned with a predetermined reference position (in FIG. 9, the center of the pattern is located in the center of the data area) to facilitate recognition. ing.
Further, since this alignment is performed, it is not necessary to input the data of the entire mold, and the data except for the place where the abnormal temperature change pattern appears is ignored, and the data is partially extracted and processed. Therefore, the amount of data is small and the number of neurons in the input layer of the neural network is small.

In process P7 of FIG. 1, data is written in each cell of the input data area generated in process P6. That is, the isotherm obtained in the process P4 and each cell position are calculated, and all the cells at the positions where the isotherm passes are set to the ignition state (black part in FIG. 3a). Figure 3a shows the cell array of the input data area resulting from processing only the isotherm C5 shown in Figure 2a, and the results processed for all the isotherms of Figure 2a are shown in Figure 3b. There is.

The data obtained by the process P7 is input to the neural network in this embodiment. In this example, the neural network is composed of three layers, an input layer, an intermediate layer, and an output layer, as shown in FIG. The result of the process P7 is applied to the input layer of the neural network. The output layer is provided with two cells that indicate whether it is normal and whether there is a breakout.

In order to use the neural net, it is necessary to train it beforehand. In this example, the past data that has already been known (whether or not a breakout has occurred) is input from the external storage device, and the data shown in Fig. 3b is generated from that data. Is applied to the input layer of the neural network, and the learning process is performed so that the correct result appears in the output layer at that time, that is, the weighting of each cell in each layer and the adjustment of the coupling strength are performed. Known back propagation is used as an algorithm for this learning process. As the neural network device, a personal neurocomputer NEURO-07 manufactured by NEC Corporation was used in this embodiment.

In the actual learning, as input data, occurrence data (actually breakout occurred) as shown in FIGS. 2a and 2b and false alarm data (as shown in FIGS. 2c and 2d) Although the breakout did not occur, the temperature distribution pattern was similar to that of the generated data, so that a false alarm is likely to occur).

After the above learning process is completed, the temperature data is input during the actual continuous casting process, and the first
By executing the processing of P1 ~ P7 shown in the figure, by applying the obtained data as prediction target data to each corresponding cell of the input layer of the neural network, the signal appearing in the output layer can be obtained as the result of the prediction. it can.

Although the neural net is used as the pattern recognition means in the above-described embodiment, the present invention can be implemented by using a general pattern recognition method such as pattern matching. Further, in the example, as the pattern to be recognized, a plurality of isotherms obtained at different points in time were used, but as shown in, for example, FIGS. 3a and 4b, the isotherms at a certain time. The input pattern may be generated by using only.

[Effect] As a result of predicting the restraint breakout by the method described above, the accuracy of prediction is significantly improved and the occurrence of false alarms is reduced compared with the conventional method.

[Brief description of drawings]

FIG. 1 is a block diagram showing the contents of processing in the embodiment of the present invention. FIG. 2a, FIG. 2b, FIG. 2c, FIG. 2d and FIG. 4a are schematic diagrams showing isotherm patterns generated from mutually different input temperature data. 3a and 3b are schematic diagrams of a data array showing the result of converting the data of FIG. 2a into the data of each cell of the input data area. 4b and 4c are schematic diagrams of a data array showing the result of converting the data of FIG. 4a into the data of each cell of the input data area. 5a, 5b, and 5c are a perspective view, a front view, and a side view, respectively, showing the outer appearance of the mold of the embodiment. FIG. 6 is a vertical cross-sectional view showing a process of occurrence of restraint breakout. FIG. 7 is a perspective view showing the appearance of a slab in which a constrained breakout has occurred. FIG. 8 is a timing chart showing a temperature change at each measurement point when a restrictive breakout occurs. FIG. 9 is a schematic diagram showing a change in the position of a pattern by normalizing the position. TC: Thermocouple X1: Measurement point C1 to C5: Isothermal line

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Yuko Tosa, 1 Kimitsu, Kimitsu-shi, Chiba Shin Nippon Steel Co., Ltd.Kaisha Kimitsu Steel Co., Ltd. In-house (72) Inventor Shuichi Endo 1015 Kamiodanaka, Nakahara-ku, Kawasaki-shi, Kanagawa Fujitsu Limited

Claims (4)

[Claims] 1. A method for predicting a constrained breakout by monitoring the temperature at each position detected by a plurality of temperature detecting means embedded in different positions of a mold wall of a continuous casting facility: From the temperature information obtained at the same timing by each, to detect the past and present temperature deviation of each position, to detect the peak point of abnormal temperature rise from the temperature deviation,
By forming a position distribution pattern of information indicating a curve connecting each detected peak point and recognizing the similarity between this position distribution pattern and a reference pattern in which the position distribution patterns actually measured in the past are stored in advance. A method for predicting the constrained breakout in continuous casting, characterized by predicting the presence or absence of breakout. 2. The restraint breakout prediction method for continuous casting according to claim 1, wherein the position distribution pattern of information indicating the presence or absence of an abnormal temperature rise is constructed by combining instantaneous detection patterns obtained at different times. 3. The position of each element in the position distribution pattern and the reference pattern is adjusted to the position of the element in which the abnormal temperature rise is first detected at a predetermined reference point, and the position distribution pattern and the reference pattern. 2. The method for predicting constrained breakout in continuous casting according to claim 1, wherein the positions of and are matched. 4. A position distribution pattern when a breakout occurs and a position distribution pattern when a breakout does not occur, which have been actually measured in the past, are input to a neural network and learned as teacher data. After that, the restraint breakout prediction method for continuous casting according to claim 1, wherein the position distribution pattern of the detection target is input to predict the presence or absence of breakout.