Method for calculating rolling schedule and manufacturing method for thick steel plates
The method employs machine learning models to accurately calculate rolling schedules for thick steel plates, addressing impracticality and inaccuracies in existing methods, ensuring high-yield production by minimizing errors and constraints.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- JFE STEEL CORP
- Filing Date
- 2024-11-27
- Publication Date
- 2026-06-08
Smart Images
Figure 2026093105000001_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for calculating a rolling schedule and a manufacturing method of a thick steel plate.
Background Art
[0002] In the manufacturing process of thick steel plates, after heating the slab in a heating furnace, it is rolled in multiple passes by a rolling mill such as a reversing rolling mill to roll to a predetermined final plate thickness. The calculation of the reduction amount and rolling load for each pass of rolling is called rolling schedule calculation and is automatically calculated by a process computer or the like. However, in order to maximize the rolling capacity, it is necessary to ensure the maximum reduction amount under the constraints of equipment capacity and the like. On the other hand, in order to ensure the plate crown (width direction plate thickness deviation) and flat steel plate shape of the thick steel plate, it is also necessary to perform rolling schedule calculation with the main focus on ensuring flatness in the rolling pass close to the final plate thickness.
[0003] From such a background, in Patent Document 1, a plurality of appropriate pass schedules are stored in a computer in advance, and a pass schedule that meets the conditions is extracted based on the measurement results of the actual plate thickness and temperature during actual rolling, and the pass schedule is determined by making corrections in consideration of the actual values. Also, in Patent Document 2, a technique has been proposed in which a pass schedule for rolling under rolling conditions that are the upper limit of equipment constraints and a pass schedule for rolling aiming at the upper limit of the rolling load are combined within the range where the crown ratio change (a factor of flatness defect) caused by the rolling load error during rolling can be controlled by a shape control mechanism.
Prior Art Documents
Patent Documents
[0004]
Patent Document 1
Patent Document 2
Summary of the Invention
Problems to be Solved by the Invention
[0005] However, if the method described in Patent Document 1 is to be applied to the manufacturing processes of thick steel plates, which have become diverse, the number of required plate thickness reduction pass schedule patterns would be extremely large, making it impractical to apply the method described in Patent Document 1 to the manufacturing processes of thick steel plates in recent years. On the other hand, in the method described in Patent Document 2, if the calculation accuracy of the rolling load is poor, in order to suppress the risk of equipment damage etc. when errors occur in the rolling load, it is necessary to create a rolling schedule that has a certain margin from the original equipment constraint limits, taking into account the variation in the rolling load. For this reason, it is necessary to minimize the calculation error of the rolling load in the rolling schedule. Furthermore, even within the range of shape constraints, if the variation in the rolling load exceeds the capacity of the shape control mechanism, it will cause flatness defects, and similar to equipment constraints, it is necessary to create a schedule with a margin within the range in which shape control is possible.
[0006] The present invention was made to solve the above problems, and its objective is to provide a method for calculating the rolling schedule of thick steel plates that can accurately calculate the rolling load. Another objective of the present invention is to provide a method for manufacturing thick steel plates that can produce them with a good yield. [Means for solving the problem]
[0007] The present invention relates to a method for calculating a rolling schedule for a thick steel plate, which is manufactured by rolling a slab through multiple passes using a rolling mill. The method comprises: a first step of calculating a provisional rolling schedule using an existing rolling load calculation model; and a second step of calculating a rolling schedule by sequentially executing, using the provisional rolling schedule calculated in the first step as an initial value, an uphill calculation that determines the reduction amount from the target rolling load using a first machine learning model, and a downhill calculation that determines the rolling load from the reduction amount using a second machine learning model.
[0008] It is advisable to calculate the rolling schedule for the thickness-adding rolling of the aforementioned thick steel plate.
[0009] The first machine learning model and the second machine learning model are preferably deep learning models.
[0010] The method for manufacturing thick steel plates according to the present invention includes the steps of calculating a rolling schedule for thick steel plates using the method for calculating a rolling schedule for thick steel plates according to the present invention, and manufacturing thick steel plates according to the calculated rolling schedule. [Effects of the Invention]
[0011] The rolling schedule calculation method for thick steel plates according to the present invention allows for accurate calculation of the rolling load. Furthermore, the manufacturing method for thick steel plates according to the present invention allows for the production of thick steel plates with a high yield. [Brief explanation of the drawing]
[0012] [Figure 1] Figure 1 is a block diagram showing the configuration of a rolling control system, which is one embodiment of the present invention. [Figure 2] Figure 2 is a flowchart showing the flow of the rolling schedule calculation process, which is one embodiment of the present invention. [Figure 3] Figure 3 is a flowchart showing the flow of the first step as shown in Figure 2. [Figure 4] Figure 4 is a conceptual diagram illustrating the second step shown in Figure 2. [Figure 5] Figure 5 is a schematic diagram showing the configuration of a deep learning model. [Modes for carrying out the invention]
[0013] A rolling control system according to one embodiment of the present invention will be described below with reference to the drawings.
[0014] 〔composition〕 FIG. 1 is a block diagram showing the configuration of a rolling control system according to an embodiment of the present invention. As shown in FIG. 1, a rolling control system 1 according to an embodiment of the present invention is a system that controls the operation of a reversible rolling mill A that manufactures a thick steel plate by rolling a slab in a plurality of passes, and includes a rolling schedule calculation device 2 and a rolling control device 3.
[0015] The rolling schedule calculation device 2 is constituted by an information device such as a process computer and is connected to the rolling control device 3 via a telecommunication line. By executing a rolling schedule calculation process described later, the rolling schedule calculation device 2 calculates the reduction amount, rolling load, etc. of each pass of a rolling process including thickness reduction rolling executed in the reversible rolling mill A. Then, the rolling schedule calculation device 2 transmits information regarding the calculation result to the rolling control device 3 as rolling schedule data.
[0016] The rolling control device 3 is constituted by an information processing device such as a personal computer and is connected to the rolling schedule calculation device 2 and the reversible rolling mill A via a telecommunication line. The rolling control device 3 controls the reversible rolling mill A according to the rolling schedule data transmitted from the rolling schedule calculation device 2, thereby controlling the rolling process in the reversible rolling mill A.
[0017] In the rolling control system 1 having such a configuration, the rolling schedule calculation device 2 executes a rolling schedule calculation process shown below to accurately calculate the rolling load in the rolling schedule. Hereinafter, the operation of the rolling schedule calculation device 2 when executing the rolling schedule calculation process will be described.
[0018] 〔Rolling Schedule Calculation Process〕 [[ID=IS]]FIG. 2 is a flowchart showing the flow of a rolling schedule calculation process according to an embodiment of the present invention. The flowchart shown in FIG. 2 starts at the timing when an execution command for the rolling schedule calculation process is input to the rolling schedule calculation device 2, and the rolling schedule calculation process proceeds to the process of step S1.
[0019] In the process of step S1, the rolling schedule calculation device 2 tentatively determines an initial value of the rolling schedule for the thick steel plate using an existing rolling load calculation model (the first step). Here, examples of the existing rolling load calculation model can include a deformation resistance model, a roll force function model, a roll flattening model, and a torque model. Details of this first step will be described later by referring to the flowchart shown in FIG. 3. Thereby, the process of step S1 is completed, and the rolling schedule calculation process proceeds to the process of step S2.
[0020] In the process of step S2, the rolling schedule calculation device 2 calculates the rolling schedule for the thick steel plate by combining and continuously performing upward machine learning for obtaining the reduction amount from the target rolling load and downward machine learning for obtaining the rolling load from the reduction amount using the initial value of the rolling schedule tentatively determined in the process of step S1 (the second step). Details of this second step will be described later by referring to FIG. 4. Thereby, the process of step S2 is completed, and the series of rolling schedule calculation processes ends.
[0021] 〔The first step〕 FIG. 3 is a flowchart showing the flow of the first step shown in FIG. 2. The flowchart shown in FIG. 3 starts at the timing when an execution instruction for the rolling schedule calculation process is input to the rolling schedule calculation device 2, and the first step proceeds to the process of step S1a.
[0022] In step S1a, the rolling schedule calculation device 2 first provisionally determines the temperature of the steel sheet in each pass from the first pass to the last pass, pass n, of the rolling process. Next, the rolling schedule calculation device 2 calculates the rolling load (final rolling load) of the final pass from the predetermined final sheet width and final sheet thickness. Then, the rolling schedule calculation device 2 calculates the reduction amount of the final pass from the final rolling load, and calculates the sheet crown (sheet Cr) and sheet crown ratio (sheet Cr ratio) of the final pass from the reduction amount of the final pass. Here, sheet crown refers to the difference between the sheet thickness at the center in the width direction of the steel sheet and the sheet thickness at both ends in the width direction. The sheet crown ratio refers to the ratio of the size of the sheet crown to the sheet thickness of the steel sheet. With this, the processing of step S1a is completed, and the first step proceeds to the processing of step S1b.
[0023] In step S1b, the rolling schedule calculation device 2 sets the value of the program counter i, which counts the number of passes in the thickness-adding rolling process, to the last pass number n. With this, the process of step S1b is completed, and the first step proceeds to the process of step S1c.
[0024] In step S1c, the rolling schedule calculation device 2 decrements the value of the program counter i by 1. This completes step S1c, and the first step proceeds to step S1d.
[0025] In step S1d, the rolling schedule calculation device 2 uses an existing reduction amount calculation model to calculate the reduction amount and entry plate thickness for the pass corresponding to the value of program counter i. Specifically, the rolling schedule calculation device 2 calculates the reduction amount and entry plate thickness for the pass corresponding to the value of program counter i based on the target rolling load determined by equipment constraints such as rolling load and rolling torque, or shape constraints corresponding to the allowable change in plate crown ratio. Hereinafter, the process of calculating the reduction amount from the target rolling load until the entry plate thickness exceeds the thickness start thickness (width-out rolling end thickness) is referred to as the up-roll calculation. With this, the processing of step S1d is completed, and the first step proceeds to the processing of step S1e.
[0026] In step S1e, the rolling schedule calculation device 2 determines whether the entry plate thickness of the pass corresponding to the program counter i calculated in step S1d is less than or equal to the shape control plate thickness. If the result of the determination is that the entry plate thickness is less than or equal to the shape control plate thickness (step S1e: Yes), the rolling schedule calculation device 2 proceeds to step S1f as the first step. On the other hand, if the entry plate thickness is greater than the shape control plate thickness (step S1e: No), the rolling schedule calculation device 2 proceeds to step S1g as the first step.
[0027] In step S1f, the rolling schedule calculation device 2 sets the rolling load for the pass corresponding to the value of the program counter i to a value that allows the change in the plate crown ratio to be within the acceptable range. With this, the process of step S1f is completed, and the first step returns to the process of step S1c.
[0028] In step S1g, the rolling schedule calculation device 2 sets the rolling load for the path corresponding to the value of the program counter i to a value that satisfies the equipment constraints such as rolling load, rolling torque, and bite angle. With this, the process of step S1g is completed, and the first step proceeds to the process of step S1h.
[0029] In step S1h, the rolling schedule calculation device 2 determines whether the reduction amount for the pass corresponding to the program counter i calculated in step S1d exceeds the thickness start thickness. If the reduction amount exceeds the thickness start thickness (width-out rolling end thickness) (step S1h: Yes), the rolling schedule calculation device 2 proceeds to step S1i as the first step. On the other hand, if the reduction amount does not exceed the thickness start thickness (step S1h: No), the rolling schedule calculation device 2 returns to step S1c as the first step.
[0030] In step S1i, the rolling schedule calculation device 2 recalculates the rolling load for each pass from the reduction amount of each pass using the existing rolling load calculation model. Specifically, the rolling schedule calculation device 2 calculates the reduction amount (round-off amount) for the pass corresponding to the program counter i value that exceeds the thickness start thickness, and performs a round-off operation to allocate the round-off amount to the passes upstream of the shape control plate thickness. Then, the rolling schedule calculation device 2 calculates the temperature and rolling load of the steel plate in each pass based on the reduction amount after the round-off operation. Hereafter, the process of calculating the temperature and rolling load of the steel plate in each pass from the reduction amount of each pass will be referred to as the downhill calculation. With this, the processing of step S1i is completed, and the first step proceeds to the processing of step S1j.
[0031] In step S1j, the rolling schedule calculation device 2 determines, starting from the first pass, whether the finished temperature of the steel sheet is within the acceptable range. If the determination shows that the finished temperature of the steel sheet is within the acceptable range (step S1j: Yes), the rolling schedule calculation device 2 terminates the first step. On the other hand, if the finished temperature of the steel sheet is not within the acceptable range (step S1j: No), the rolling schedule calculation device 2 uses the temperature of the steel sheet calculated in step S1i to repeat the processes from step S1a onward.
[0032] [Second Step] Figure 4 is a conceptual diagram illustrating the second step. As shown in Figure 4, the second step, like the first step, performs uphill and downhill calculations, but uses machine learning models for each calculation. Specifically, in the uphill calculation, the first machine learning model is used to calculate the reduction amount and inlet plate thickness from the target rolling load determined by equipment constraints or shape constraints. Here, the first machine learning model is a machine learning model that uses the target rolling load, temperature (temperature obtained in the first step or calculated in the downhill calculation), outlet plate thickness, dimensions, and component conditions as explanatory variables, and the reduction amount and inlet plate thickness as the objective variables. An example of a machine learning model is the deep learning model shown in Figure 5. In the downhill calculation, the second machine learning model is used to calculate the rolling load from the reduction amount. Here, the second machine learning model is a machine learning model that uses the inlet and outlet plate thicknesses (values calculated in the uphill calculation are used), temperature, dimensions, and component conditions as explanatory variables, and the rolling load as the objective variable. Then, the temperature is recalculated in the same way as in the first step, and the upward and downward calculations are repeated until the temperature converges, just as in the first step. Once the temperature converges, the rolling schedule calculation is terminated.
[0033] Since the first and second machine learning models are created based on actual rolling conditions and historical values, if values that deviate from historical values are used as explanatory variables during the rolling schedule calculation, there is a high probability that the reduction amount and rolling load will not show values within the normal range. In particular, the initial temperature values for each pass in the first loop of the calculation often deviate from historical values. For this reason, if the rolling schedule is calculated using the machine learning model without going through the first step, there is a high probability that the normal reduction amount and rolling load will not be output, and the calculation will not converge. In particular, in the second step, the output results of the uphill and downhill calculations are used as explanatory variables for the other calculation, so if the calculation result shows an abnormal value even once, the output of abnormal values will continue from that point onward, and normal calculation results will not be output. For this reason, in this embodiment, in the first step, a provisional rolling schedule is calculated using an existing rolling schedule calculation method, and the calculated provisional rolling schedule is used as the initial value for the calculation in the second step. This allows the rolling schedule to be calculated with high accuracy. Furthermore, by manufacturing thick steel plates according to the calculated rolling schedule, thick steel plates can be manufactured with a high yield.
[0034] [Examples] To evaluate the effectiveness of the present invention, the ratio (actual value ÷ calculated value) of the rolling load at the rolling schedule calculation stage in the rolling of thick steel plates was compared. The evaluation subjects were materials with a final plate thickness of 15 to 40 mm from controlled rolling (22 target materials, 97 target passes). For the same material, the rolling load of the provisional schedule calculated in the first step of the present invention was used as the calculated value in the comparative example, and the rolling load calculated in the second step was used as the calculated value in the example, and both were compared with the actual value. The comparison results are shown in Table 1 below. As shown in Table 1, the standard deviation of the ratio was 9.5% in the comparative example, while the standard deviation of the ratio was 8.2% in the example, confirming an improvement of 1.3%. From the above, it was confirmed that the rolling load can be calculated with high accuracy according to the present invention.
[0035] [Table 1]
[0036] Although embodiments applying the invention made by the present inventors have been described above, the present invention is not limited by the descriptions and drawings that constitute part of the disclosure of the present invention in this embodiment. That is, all other embodiments, examples, and operational techniques made by those skilled in the art based on this embodiment are included in the scope of the present invention. [Explanation of Symbols]
[0037] 1. Rolling control system 2. Rolling schedule calculation device 3. Rolling control device A Reversible Rolling Mill
Claims
1. A method for calculating the rolling schedule of a thick steel plate, which is manufactured by rolling a slab through multiple passes using a rolling mill, The first step is to calculate a provisional rolling schedule using an existing rolling load calculation model, The second step involves calculating the rolling schedule by sequentially executing, using the provisional rolling schedule calculated in the first step as an initial value, an uphill calculation that determines the reduction amount from the target rolling load using a first machine learning model, and a downhill calculation that determines the rolling load from the reduction amount using a second machine learning model. A method for calculating the rolling schedule of thick steel plates, including the method described above.
2. A method for calculating the rolling schedule for increasing the thickness of the thick steel plate, according to claim 1.
3. The method for calculating the rolling schedule of a thick steel plate according to claim 1, wherein the first machine learning model and the second machine learning model are deep learning models.
4. A method for manufacturing a thick steel plate, comprising the steps of calculating a rolling schedule for a thick steel plate using the method for calculating a rolling schedule for a thick steel plate described in any one of claims 1 to 3, and manufacturing a thick steel plate according to the calculated rolling schedule.