Slab preheating determination device, hot scarf control device, slab preheating determination method, slab manufacturing method, and method for generating trained model for determining slab preheating

Abstract

To provide a technique capable of determining preheating completion of a slab accurately in hot scarf processing of the slab, improving the slab quality and saving the labor of an operator in the hot scarf processing.SOLUTION: A slab preheating determination device for determining preheating completion of a slab in hot scarf processing that performs surface scarfing of the slab after preheating the slab comprises: a data acquisition unit which acquires image data showing a preheating state of the slab captured by a camera unit; a primary determination unit which performs primary determination of preheating completion by estimating the preheating state with deep learning using a learned model to the acquired image data; and a secondary determination unit which performs final preheating completion determination with luminance of the image acquired by the camera unit at a stage where the preheating completion state is determined in the primary determination.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a slab preheating determination device, a hot scarf control device, a slab preheating determination method, a slab manufacturing method, and a method for generating a trained model for determining slab preheating in hot scarf processing of slabs such as bloomed rolled slabs and continuous cast slabs. [Background technology]

[0002] In the steel slab preparation process, impurities and defects on the surface of the slab that occurred during the steelmaking process are removed by hot scarfing the surface of the slab using hot scarf equipment. The hot scarf equipment has a nozzle unit equipped with a scarf nozzle that sprays oxygen and propane gas, and the nozzle unit is fitted closely to the slab according to its size. The slab is thermally cut by igniting a propane burner and continuously moving the slab in the longitudinal direction after preheating for a specific period of time.

[0003] This slag cutting operation is generally performed semi-automatically with the intervention of an operator, and the preheating time is also determined by the operator, who monitors the billet of steel with a camera while it is being preheated, judging its condition. In actual operation, the operator starts conveying the billet of steel when the operator judges that preheating is complete based on the billet condition monitoring, and then slag cutting is performed on the billet. If the preheating time is insufficient, there will be residual molten material on the billet surface during the slag cutting operation, which will result in an extended time for grinding the billet surface in the next process, the grinding process, leading to reduced operational efficiency. Furthermore, if the preheating time is excessive, it will lead to a deterioration in the quality of the billet of steel and melting damage to the nozzle unit. Therefore, the preheating time before slag cutting must be determined appropriately.

[0004] Currently, the preheating time is determined largely at the discretion of the operator, which creates the problem of individual differences. In addition, since the operation relies on the operator's visual judgment, it is difficult to promote automation and reduce labor. Therefore, for the purpose of automatically determining the completion of preheating in the hot scarf process, for example, Patent Document 1 proposes a method of capturing an image of the preheated part with a CCD camera, binarizing the image data to calculate the area of ​​the molten steel piece (pocket), and determining the completion of preheating based on the calculated area. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 05-269577 Summary of the Invention [Problem to be solved by the invention]

[0006] However, in the conventional technology described in Patent Document 1, there is a problem that the accuracy of judging the completion of preheating based only on the area of ​​the basin is not high because actual steel billets have different steel types and shapes. In other words, since the judgment is made only on the basis of the area of ​​the basin without considering the change in the state of the billet during preheating, it is difficult to deal with variations in the steel type and shape of the billet, the surrounding environment, and other variations in actual operation.

[0007] Therefore, the present invention provides a technology that can accurately determine the completion of preheating of a steel piece during hot scarf processing of the steel piece, improve the quality of the steel piece, and reduce the operator's labor in the hot scarf processing. [Means for solving the problem]

[0008] In order to solve the above problems, the present invention provides the following [1] to

[11] .

[0009] [1] A steel piece preheating judgment device for judging completion of preheating of a steel piece in a hot scarf process in which the steel piece is preheated and then surface-scraped, comprising: a data acquisition unit that acquires image data showing a preheating state of the steel piece captured by a camera unit; A primary determination unit that estimates a preheating state by deep learning using a learned model for the acquired image data and performs a primary determination of preheating completion; a secondary determination unit that performs a final determination of completion of preheating based on the brightness of an image captured by the camera unit when the primary determination determines that the preheating is complete; The steel piece preheating determination device has the above-mentioned features.

[0010] [2] The steel piece preheating judgment device described in [1], wherein the trained model is deep-learned by associating multiple images captured and stored by the camera unit with multiple preheating states at the start of preheating, during preheating, and completion of preheating.

[0011] [3] The steel piece preheating judgment device described in [1] or [2], wherein the secondary judgment unit calculates the brightness for each pixel of the image captured by the camera unit, counts the number of pixels whose calculated brightness exceeds a first threshold, and judges that preheating is complete when the number of pixels exceeds a second threshold.

[0012] [4] The steel piece preheating judgment device according to any one of [1] to [3], wherein deep learning in the primary judgment unit is performed using a convolutional neural network.

[0013] [5] A hot scarf control device in a hot scarf process in which a steel piece is preheated and then surface-scraped, comprising: A camera unit that captures an image of the preheated state of the steel piece; a slab preheating determination device that determines completion of preheating of the slab based on image data captured by the camera unit; a conveyance control unit that controls conveyance of the steel piece for surface machining of the steel piece when the steel piece preheating determination device determines that preheating is complete; Equipped with A hot scarf control device that uses a steel slab preheating determination device described in any one of [1] to [4] as the steel slab preheating determination device.

[0014] [6] A method for determining whether or not a steel piece is preheated in a hot scarf process in which the steel piece is preheated and then surface-scraped, the method comprising the steps of: acquiring image data showing a preheating state of the steel piece captured by a camera unit; A step of estimating a preheating state by deep learning using a trained model for the acquired image data and making a primary determination of preheating completion; a step of making a final determination of completion of preheating based on the brightness of an image captured by the camera unit when the first determination determines that the preheating is completed;

[0015] [7] The steel piece preheating judgment method described in [6], wherein the trained model is deep-learned by associating multiple images captured and stored by the camera unit with multiple preheating states at the start of preheating, during preheating, and completion of preheating.

[0016] [8] The steel piece preheating judgment method described in [6] or [7], wherein the step of making the final preheating completion judgment comprises calculating the brightness for each pixel of the image captured by the camera unit, counting the number of pixels whose calculated brightness exceeds a first threshold value, and further judging that preheating is complete when the number of pixels exceeds a second threshold value.

[0017] [9] A method for determining whether or not a steel piece is preheated, comprising the steps of: (a) determining whether or not a steel piece is preheated; and (b) determining whether or not a steel piece is preheated.

[0018]

[10] A method for manufacturing a steel piece, comprising subjecting the surface of the steel piece to surface machining using any one of the steel piece preheating determination methods [6] to [9] above.

[0019]

[11] A method for generating a trained model for determining preheating of a steel piece, comprising: A method for generating a trained model for determining the preheating of a piece of steel, using multiple images showing the preheating state of the piece of steel captured by a camera unit as input, and performing deep learning using one of the preheating states at the start of preheating, during preheating, or completion of preheating as output. Effect of the Invention

[0020] According to the present invention, the completion of preheating of a slab can be accurately determined in the hot scarf process of the slab, thereby improving the quality of the slab and reducing the labor required for the operator in the hot scarf process. In addition, the quality of the slab can be improved by reducing the variation in preheating time at the discretion of the operator. [Brief description of the drawings]

[0021] [Figure 1] 1 is a diagram showing a schematic configuration of a hot skirt facility for performing hot scarf treatment having a steel slab preheating determination device according to one embodiment of the present invention. FIG. [Diagram 2] FIG. 13 shows images of a steel slab during an actual preheating process. [Diagram 3] 1 is a flowchart showing processing operations in a hot scarf equipment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0022] Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

[0023] FIG. 1 is a diagram showing a schematic configuration of a hot scarf facility for performing a hot scarf treatment having a steel slab pre-heating determination device according to one embodiment of the present invention. The hot scarf equipment has a nozzle unit 11 for scarifying the surface of a steel piece 1 transported on a transport roll 2, and a hot scarf control device 12.

[0024] The nozzle unit 11 is equipped with a scarf nozzle that sprays oxygen and propane gas, and shaves the surface of the steel piece 1. The nozzle unit 11 fits closely to the steel piece 1 according to the size of the steel piece 1. The shave of the steel piece 1 is performed by igniting a propane burner, preheating the steel piece 1 for a specific period of time, and then moving the steel piece 1 continuously in the longitudinal direction along the transport rolls 2.

[0025] The hot scarf control device 12 has a camera unit 21 that captures images of the steel piece melt-cutting state, a steel piece preheating determination device 22 that determines whether preheating has been completed based on the captured image, and a drive control device 23 that controls the roll drive when the steel piece preheating determination device 22 determines that preheating has been completed.

[0026] The camera unit 21 has an imaging camera installed to monitor the upper surface of the steel piece. The imaging camera is configured to capture images at 30 frames per second, for example, and the captured images are transmitted to the steel piece pre-heating determination device 22 as video data.

[0027] The billet preheating determination device 22 acquires video data at regular intervals and determines whether preheating of the billet 1 is complete, and includes a data acquisition unit 31, a primary determination unit 32, a secondary determination unit 33, and a storage medium 34. As described below, the billet preheating determination device 22 detects not only the final state of the billet 1 but also state changes during preheating, thereby accurately determining whether preheating is complete.

[0028] Regarding the change in the state of billet 1 during preheating, the billet before preheating is black overall, the billet during preheating has a striped pattern due to a mixture of white areas where preheating is partially complete and black areas where preheating is not complete, and the billet that is preheated completely appears uniformly white throughout. Figure 2 shows images of a billet during the actual preheating process. This change in the appearance of the billet is also used as a judgment indicator when an operator visually judges whether preheating is complete, and is a common characteristic even when the steel type and billet shape are different.

[0029] In the slab preheating judgment device 22 of this embodiment, in order to capture characteristic changes in the appearance of the slab during the preheating process as described above for the video data acquired by the data acquisition unit 31, the primary judgment unit 32 performs image classification by deep learning using a convolutional neural network (hereinafter, CNN) to estimate the preheating state of the slab 1. A trained model created in advance is implemented in the primary judgment unit 32, which detects the feature amounts of the slab surface in each of the preheating states described above, and makes a primary judgment of preheating completion when the output of the trained model indicates a preheating completion state.

[0030] The trained model is a deep learning model in which a plurality of images captured and stored by the camera unit 21 are associated with a plurality of preheating states at the start of preheating, during preheating, and completion of preheating. That is, when generating a trained model, first, a plurality of images captured by the photographing camera of the camera unit 21 are stored as input in the storage medium 34 in the steel piece preheating judgment device 22. Then, deep learning is performed with one of the preheating states at the start of preheating, during preheating, and completion of preheating of the plurality of stored images as output. By implementing the trained model generated in this manner in the primary judgment unit 32, the model can be used to extract features of the captured images and estimate the state of the steel piece 1.

[0031] However, deep learning using CNN is vulnerable to image noise and distortion. In addition, when preheating is complete, the entire piece of steel is heated sufficiently and its brightness increases, but it is difficult for CNN to capture changes in the brightness of the piece of steel. For this reason, there is a risk that the completion of preheating cannot be determined with high accuracy.

[0032] Therefore, in order to improve the accuracy of judging whether preheating is complete, the steel piece preheating judgment device 22 further includes a secondary judgment unit 33. After the primary judgment unit 32 judges whether preheating is complete using deep learning, the secondary judgment unit 33 judges whether preheating is complete based on the brightness of the image captured by the camera unit. That is, the primary judgment unit 32 estimates the characteristic changes in the steel piece appearance during the preheating process using a learned model to make a primary judgment on whether preheating is complete, and at the stage where the primary judgment judges that the preheating is complete, the secondary judgment unit 33 separately makes a secondary judgment, which is a final judgment on whether preheating is complete, based on the brightness of the image.

[0033] In the secondary judgment unit 33, first, the luminance of each pixel of the image captured by the camera unit 21 and acquired by the data acquisition unit 31 is calculated using, for example, the NTSC weighted average method that is generally used when calculating luminance from a color image. Next, the number of pixels whose calculated luminance exceeds a first threshold is counted, and when this number of pixels exceeds a second threshold, preheating is judged to be complete. Since the number of pixels varies depending on the size (width) of the billet, multiple second thresholds are set by dividing the size (width) of the billet. By appropriately setting these thresholds, it is possible to accurately detect whether the entire billet has been sufficiently heated in the preheating process, i.e., whether preheating has been completed.

[0034] When the piece of steel preheating determination device 22 determines that preheating is complete, the drive control device 23 receives that signal and controls the driving of the drive roll of the transport rolls 2, thereby controlling the transport of the piece of steel 1 in the longitudinal direction. That is, the drive control device 23 functions as a transport control unit that controls the transport of the piece of steel. The transport control of the piece of steel 1 at this time is performed for the purpose of surface machining of the piece of steel 1 using the nozzle unit 11. That is, the surface machining of the piece of steel 1 is performed using the nozzle unit 11 while the transport of the piece of steel 1 is controlled by the drive control device 23. Then, when the transport of a predetermined piece of steel 1 has been completed by the drive control device 23, the surface machining of the piece of steel 1 is automatically completed.

[0035] Next, the operation of the hot scarf equipment configured as above will be described. 3 is a flow chart showing the processing operation in the hot scarf equipment. First, a command to start preheating is issued from a higher-level control unit (ST1), and imaging by the imaging camera in the camera unit 21 is started (ST2). The captured image data is acquired by the data acquisition unit 31 of the steel piece preheating determination device 22.

[0036] Next, the primary judgment unit 32 of the steel piece preheating judgment device 22 performs image classification using deep learning with CNN to estimate the preheating state of the steel piece 1, thereby making a primary judgment of preheating completion (ST3). After the primary judgment of preheating completion is made by CNN, the secondary judgment unit 33 makes a secondary judgment of preheating completion based on brightness (ST4). By repeating these steps at a constant cycle, it is possible to detect the timing of preheating completion, and finally a judgment of preheating completion is made (ST5).

[0037] In other words, the primary judgment by CNN plays the role of monitoring the condition of the steel piece, and when it is determined through condition monitoring using this CNN that the preheating has been completed, a secondary judgment (final judgment) based on brightness is made.

[0038] When the completion of preheating is determined by the billet preheating determination device 22, a signal is transmitted to the drive control device 23, which starts the transportation of the billet (ST6), and starts the surface machining process of the billet 1. Then, when the transportation of a predetermined number of billets 1 has been completed, the transportation of the billet is stopped (ST7), and the surface machining of the billet 1 is automatically completed (ST8).

[0039] Conventional preheating judgment methods make judgments only based on the result of the basin area without considering changes in the state of the slab during preheating, and therefore have the problem of being unable to deal with variations in actual operation such as the steel type and shape of the slab, the surrounding environment, etc. In this embodiment, therefore, the completion of preheating is judged with high accuracy by detecting state changes in addition to the final state of the slab.

[0040] Regarding the change in the state of the slab during preheating, as described above, image classification is performed by deep learning using CNN, and the implemented CNN judgment model is used to extract features from the captured image, making it possible to classify the state of the slab 1.

[0041] However, deep learning using CNN is vulnerable to image noise and distortion, and it is difficult for deep learning using CNN to capture the characteristic that the brightness increases when the entire steel piece is sufficiently heated at the completion of preheating. For this reason, in this embodiment, the primary judgment unit 32 estimates the characteristic changes in the steel piece's appearance during the preheating process using a learned model to make a primary judgment of preheating completion, and when the primary judgment indicates that preheating is complete, the secondary judgment unit 33 separately judges the brightness of the image to make a secondary judgment, which is a final judgment of preheating completion. In this way, by combining the primary judgment by deep learning using CNN and the secondary judgment by brightness, it is possible to accurately judge the completion of preheating.

[0042] <Other applications> Although the embodiments of the present invention have been described above, these are merely examples and should not be considered as limiting. The above embodiments may be omitted, substituted, or modified in various ways without departing from the gist of the present invention.

[0043] For example, in the above embodiment, an example has been shown in which a primary determination is performed by deep learning using CNN, but deep learning is not limited to CNN, and various methods can be used. [Explanation of symbols]

[0044] 1 piece of steel 2 Transport roll 11 Crater mouth unit 12 Hot Scarf Control Device 21 Camera unit (imaging camera) 22 Steel billet preheating judgment device 23 Drive control device 31 Data Acquisition Section 32 Primary judgment part 33 Secondary judgment section 34 Storage medium

Claims

1. A slab preheating determination device for determining completion of preheating of a slab in a hot scarf process in which the slab is preheated and then surface-scaling is performed, comprising: a data acquisition unit that acquires image data showing a preheating state of the steel piece captured by a camera unit; A primary determination unit that estimates a preheating state by deep learning using a learned model for the acquired image data and performs a primary determination of preheating completion; a secondary determination unit that performs a final determination of completion of preheating based on the brightness of an image captured by the camera unit when the primary determination determines that the preheating is complete; The steel piece preheating determination device has the above-mentioned features.

2. 2. The steel piece preheating judgment device according to claim 1, wherein the learned model is deep-learned by associating a plurality of images captured and stored by the camera unit with a plurality of preheating states at the start of preheating, during preheating, and completion of preheating.

3. 3. The steel piece preheating determination device according to claim 1 or claim 2, wherein the secondary determination unit calculates a luminance for each pixel of the image captured by the camera unit, counts the number of pixels whose calculated luminance exceeds a first threshold, and determines that preheating is completed when the number of pixels exceeds a second threshold.

4. The steel slab preheating determination device according to claim 1 , wherein deep learning in the primary determination unit is performed by a convolutional neural network.

5. A hot scarf control device in a hot scarf process in which a steel piece is preheated and then surface-scalded, comprising: A camera unit that captures an image of the preheated state of the steel piece; a slab preheating determination device that determines completion of preheating of the slab based on image data captured by the camera unit; a conveyance control unit that controls conveyance of the steel piece for surface machining of the steel piece when the steel piece preheating determination device determines that preheating is complete; Equipped with A hot scarf control device using the slab pre-heating determination device according to any one of claims 1 to 4 as the slab pre-heating determination device.

6. A method for determining whether or not preheating of a steel piece is completed in a hot scarf process in which the steel piece is preheated and then surface-scaled, comprising: acquiring image data showing a preheating state of the steel piece captured by a camera unit; A step of estimating a preheating state by deep learning using a trained model for the acquired image data and making a primary determination of preheating completion; a step of making a final determination of completion of preheating based on the brightness of an image captured by the camera unit when the first determination determines that the preheating is completed; The method for determining preheating of a steel piece comprises the steps of:

7. The steel piece preheating judgment method according to claim 6, wherein the learned model is deep-learned by associating a plurality of images captured and stored by the camera unit with a plurality of preheating states at the start of preheating, during preheating, and completion of preheating.

8. The steel piece preheating judgment method according to claim 6 or claim 7, wherein the step of making the final preheating completion judgment comprises calculating a luminance for each pixel of the image captured by the camera unit, counting the number of pixels whose calculated luminance exceeds a first threshold value, and judging that preheating is completed when the number of pixels exceeds a second threshold value.

9. The steel slab preheating judgment method according to claim 6 , wherein deep learning in the step of making the primary judgment is performed by a convolutional neural network.

10. A method for manufacturing a steel slab, comprising the steps of: subjecting a surface of the steel slab to spalling using the method for determining preheating of the steel slab according to any one of claims 6 to 9.

11. In the steel piece preheating judgment device according to claim 1, a method for generating a trained model to be used in the primary judgment unit, comprising: A method for generating a trained model, comprising: inputting a plurality of images showing the preheating state of the steel piece captured by the camera unit; associating the images with any one of the preheating states at the start of preheating, during preheating, and at the completion of preheating as an output; performing deep learning to generate the trained model.

Images