Rolling condition setting method for cold rolling mill, cold rolling method, steel sheet manufacturing method, rolling condition setting device for cold rolling mill, and cold rolling mill

Abstract

In the rolling condition setting method of the cold rolling mill of the present invention, the prediction model is a model generated with a first multi-dimensional data as an explanatory variable and a post-rolling data of a rolled material at an exit side of the cold rolling mill as a target variable, the first multi-dimensional data being converted from past rolling performance data including pre-rolling data of a rolled material at an entry side of the cold rolling mill into a multi-dimensional data, the rolling condition setting method of the cold rolling mill including: a step of estimating a post-rolling shape of the rolled material at the exit side of the cold rolling mill by inputting a second multi-dimensional data into the prediction model, the second multi-dimensional data being generated from information including pre-rolling data of the rolled material at the entry side of the cold rolling mill and a target rolling condition of the cold rolling mill; and a step of changing the target rolling condition of the cold rolling mill in a manner that the estimated post-rolling shape satisfies a prescribed condition.

Description

Technical Field

[0001] This invention relates to a method for setting rolling conditions for a cold rolling mill, a cold rolling method, a method for manufacturing steel plates, a device for setting rolling conditions for a cold rolling mill, and a cold rolling mill. Background Technology

[0002] Generally, when cold rolling rolled materials such as thin cold-rolled steel sheets, it is preferable to perform cold rolling in a state that stabilizes the plate properties of the rolled material by maintaining good thickness accuracy in both the length and width directions and good shape (or flatness) of the rolled material. On the other hand, the demand for difficult-to-roll materials, such as thin, hard materials with high loads and relatively thin initial thickness, is increasing, with the aim of reducing fuel consumption through weight reduction. In the cold rolling of such difficult-to-roll materials, in order to suppress the rolling load, the difficult-to-roll material is thinned by hot rolling in the previous process and then sent to the cold rolling process.

[0003] In recent years, many control elements of cold rolling mills have been automatically controlled by actuators mounted on the mill, reducing the opportunities for operators to set these control elements. However, in the cold rolling of the aforementioned difficult-to-roll materials, sheet crown (thickness distribution in the width direction) sometimes varies significantly along the length direction. When sheet crown varies significantly along the length direction, variations in rolling load (and consequently calculated slip ratio and torque), relative to the mill's roll gap, work roll bending, intermediate roll displacement, and roll deflection correction represented by roll expansion due to thermal crown, often cannot be absorbed by automatic control.

[0004] Therefore, in this situation, operators set pass schedules and shape control actuators in a way that meets the equipment constraints of the cold rolling mill without hindering productivity. Consequently, in recent years, the operating speed and, consequently, productivity of cold rolling mills have become susceptible to operator experience and subjectivity. Against this backdrop, Patent Document 1 proposes a method for setting up the cold rolling mill using a neural network to learn past operating conditions and the learning results. Furthermore, Patent Document 2 proposes a method for feedforward control of edge drop using the plate thickness profile measured at the inlet side of the cold rolling mill.

[0005] Existing technical documents

[0006] Patent documents

[0007] Patent Document 1: Japanese Patent No. 6705519

[0008] Patent Document 2: Japanese Patent No. 4784320 Summary of the Invention

[0009] The problem that the invention aims to solve

[0010] However, in the method described in Patent Document 1, even when the cold rolling mill operates under optimal conditions at the time of mill setup, the shape of the rolled material at the exit side of the cold rolling mill varies considerably when the plate crown changes along the length direction. This can lead to limitations in rolling speed due to poor shape, and in the worst-case scenario, material breakage may occur. On the other hand, in the method described in Patent Document 2, since the plate thickness profile is only a cross-section along the length direction and a linear regression method is used to predict edge drop, it similarly cannot cope with the situation where the plate crown changes along the length direction.

[0011] The present invention addresses the aforementioned problems and aims to provide a method and apparatus for setting rolling conditions for a cold rolling mill that can ensure both stability and high productivity when rolling difficult-to-roll materials under high loads and with thin initial sheet thickness. Another objective of the present invention is to provide a cold rolling method and cold rolling mill that can ensure both stability and high productivity when cold rolling difficult-to-roll materials under high loads and with thin initial sheet thickness. Furthermore, another objective of the present invention is to provide a method for manufacturing steel sheets that can produce steel sheets with a good yield.

[0012] Methods for solving problems

[0013] The rolling condition setting method for a cold rolling mill of the present invention uses a predictive model that predicts the post-cold-rolled state of the material to be rolled to set the target rolling conditions of the cold rolling mill when cold-rolling the material. The predictive model is generated by using first multidimensional data as explanatory variables and post-cold-rolled data of the rolled material at the exit side of the cold rolling mill as the target variable. The first multidimensional data is generated by converting past rolling performance data, including pre-cold-rolled data of the rolled material at the inlet side of the cold rolling mill, into multidimensional data. The rolling condition setting method includes: a step of estimating the post-rolled shape of the material at the exit side of the cold rolling mill by inputting second multidimensional data into the predictive model; the second multidimensional data being generated based on information including pre-cold-rolled data of the material at the inlet side of the cold rolling mill and the target rolling conditions of the cold rolling mill; and a step of changing the target rolling conditions of the cold rolling mill in a manner that the estimated post-rolled shape meets specified conditions.

[0014] Alternatively, the data before cold rolling may include at least one of the thickness information and temperature information of the steel plate at the inlet side of the cold rolling mill.

[0015] Alternatively, the cold-rolled data may include shape parameters calculated based on the shape of the steel plate at the exit side of the cold rolling mill.

[0016] The cold rolling method of the present invention includes the step of cold rolling a material to be rolled using the target rolling conditions of a cold rolling mill, wherein the target rolling conditions of the cold rolling mill are obtained by modifying the rolling condition setting method of the cold rolling mill of the present invention.

[0017] The method for manufacturing the steel plate of the present invention includes the step of manufacturing the steel plate using the cold rolling method of the present invention.

[0018] The rolling condition setting device for a cold rolling mill of the present invention uses a predictive model that predicts the post-cold-rolled state of the material to be rolled to set the target rolling conditions of the cold rolling mill when cold-rolling the material. The predictive model is generated by using first multidimensional data as explanatory variables and post-cold-rolled data of the rolled material at the exit side of the cold rolling mill as the target variable. The first multidimensional data is generated by converting past rolling performance data, including pre-cold-rolled data of the rolled material at the inlet side of the cold rolling mill, into multidimensional data. The rolling condition setting device includes: a mechanism for estimating the post-rolled shape of the material at the exit side of the cold rolling mill by inputting second multidimensional data into the predictive model; the second multidimensional data being generated based on information including pre-cold-rolled data of the material at the inlet side of the cold rolling mill and the target rolling conditions of the cold rolling mill; and a mechanism for changing the target rolling conditions of the cold rolling mill in a manner that satisfies predetermined conditions for the estimated post-rolled shape.

[0019] Alternatively, the data before cold rolling may include at least one of the thickness information and temperature information of the steel plate at the inlet side of the cold rolling mill.

[0020] Alternatively, the cold-rolled data may include shape parameters calculated based on the shape of the steel plate at the exit side of the cold rolling mill.

[0021] The cold rolling mill of the present invention includes the rolling condition setting device of the cold rolling mill of the present invention.

[0022] The effects of the invention

[0023] The rolling condition setting method and rolling condition setting device of the cold rolling mill according to the present invention can set rolling conditions that ensure both stability and good productivity during cold rolling, even when cold rolling difficult-to-roll materials with high loads and thin initial thicknesses. Furthermore, the cold rolling method and cold rolling mill according to the present invention can ensure both stability and good productivity during cold rolling, even when cold rolling difficult-to-roll materials with high loads and thin initial thicknesses. Additionally, the steel sheet manufacturing method according to the present invention enables the manufacture of steel sheets with good yield. Attached Figure Description

[0024] Figure 1 This is a schematic diagram showing the structure of a cold rolling mill as an embodiment of the present invention.

[0025] Figure 2 It is shown Figure 1 The diagram shows the structure of the arithmetic unit.

[0026] Figure 3 This is a diagram illustrating an example of multidimensional array information.

[0027] Figure 4 This is a diagram illustrating a structural example of a shape control prediction model.

[0028] Figure 5 This is a flowchart illustrating the process of converting multidimensional array information into one-dimensional information.

[0029] Figure 6 This is a flowchart illustrating the processing flow of the predictive model execution unit. Detailed Implementation

[0030] Hereinafter, with reference to the accompanying drawings, a method for setting rolling conditions for a cold rolling mill, a cold rolling method, a method for manufacturing steel plates, a device for setting rolling conditions for a cold rolling mill, and a cold rolling mill, as embodiments of the present invention, will be described. It should be noted that the embodiments shown below exemplify apparatus and methods for embodying the technical concept of the present invention, and do not limit the materials, shapes, structures, and arrangements of the constituent components to the embodiments shown below. Furthermore, the accompanying drawings are schematic. Therefore, it should be noted that the relationship between thickness and top view dimensions, ratios, etc., differ from reality, and the accompanying drawings also include portions with different dimensional relationships and ratios.

[0031] [Structure of a cold rolling mill]

[0032] First, refer to Figure 1 The structure of a cold rolling mill according to one embodiment of the present invention will be described. It should be noted that in this specification, "cold rolling" is sometimes referred to simply as "rolling," and both "cold rolling" and "rolling" are synonymous. Furthermore, in the following description, steel plate is used as an example of the material to be rolled using a cold rolling mill. However, the rolling material is not limited to steel plate; other metal plates such as aluminum plates can also be used.

[0033] Figure 1 This is a schematic diagram showing the structure of a cold rolling mill as an embodiment of the present invention. Figure 1 As shown, the cold rolling mill 1, as an embodiment of the present invention, is installed from the inlet side (towards) of the steel plate S. Figure 1 (The left side of the paper) facing outwards (towards) Figure 1The cold rolling mill 1 (as shown on the right side of the paper) has five rolling stands in sequence, namely the first rolling stand to the fifth rolling stand (#1STD to #5STD). In this cold rolling mill 1, tension rolls and differential rolls (not shown), a plate thickness gauge and a shape gauge are appropriately arranged between adjacent rolling stands. The structure of the rolling stands and the conveying device for the steel plate S are not particularly limited, and known technologies can be appropriately applied.

[0034] Emulsion rolling oil (hereinafter referred to as "rolling oil") OL is supplied to each rolling stand of the cold rolling mill 1. The cold rolling mill 1 has a dirty tank (recycling tank) 2 and a clean tank 3 as rolling oil storage tanks, and the rolling oil OL supplied from these tanks is supplied to each rolling stand through supply line 11.

[0035] The rolling oil recovered by the oil pan 5 located below the first to fifth rolling mill stands, i.e. the rolling oil used in cold rolling, flows into the dirty tank 2 through the return pipe 6.

[0036] The rolling oil OL stored in the clean tank 3 is produced by mixing warm water (diluent) and the original rolling oil (with added surfactant). The mixture of warm water and original rolling oil is adjusted by changing the rotation speed of the stirring blades of the mixer 7, that is, by adjusting the degree of stirring, to produce rolling oil OL with the desired average particle size and concentration range as a target.

[0037] As a base fluid for rolling, it can be used in common cold rolling processes, such as base fluids made from natural oils, fatty acid esters, and hydrocarbon-based synthetic lubricants. It should be noted that oiliness improvers, extreme pressure additives, and antioxidants, commonly used in cold rolling oils, can also be added to these rolling oils.

[0038] As the surfactant added to the rolling oil, any type, whether ionic or nonionic, can be used. Any surfactant commonly used in circulating coolant systems (circulating rolling oil supply methods) is acceptable. Then, the stock rolling oil is preferably diluted to a concentration of 2-8% by mass, more preferably 3-6.0% by mass, and an O / W emulsion rolling oil with oil dispersed in water is formed using the surfactant. It should be noted that the average particle size of the rolling oil is preferably 15 μm or less, more preferably 3-10 μm.

[0039] After operation begins, the rolling oil recovered to the dirty tank 2 flows into the clean tank 3 via an iron powder removal device 8, which consists of an iron powder quantity control device and the like. The rolling oil recovered to the dirty tank 2 contains wear powder (iron powder) generated by the friction between the rolls and the steel plate S. Therefore, the iron powder removal device 8 removes the wear powder in a manner that reduces the soluble iron content of the recovered rolling oil to a level permissible for the rolling oil OL accumulated in the clean tank 3.

[0040] The movement of rolling oil from the dirty tank 2 side to the clean tank 3 side via the iron powder removal device 8 can be continuous or intermittent. As the iron powder removal device 8, a magnetic filter such as an electromagnetic filter or a magnetic separator is preferred to adsorb and remove iron powder, but it is not limited to this. The iron powder removal device 8 can also be a known device using methods such as centrifugal separation.

[0041] However, some of the rolling oil supplied to the rolling mill stand is carried out of the system by the steel plate S or lost due to evaporation. Therefore, the clean tank 3 is designed to appropriately replenish (supply) the raw rolling oil from the raw oil tank (not shown) so that the accumulation level and concentration of the rolling oil OL in the clean tank 3 are within a specified range. In addition, warm water for diluting the rolling oil is also appropriately replenished (supplyed) to the clean tank 3. It should be noted that the accumulation level and concentration of the emulsion rolling oil OL in the clean tank 3 can be measured using a sensor (not shown).

[0042] Next, the details of the rolling oil supply system for the cold rolling mill 1 will be explained. The rolling oil supply system for the cold rolling mill 1 includes a dirty tank 2, an iron powder removal device 8, a clean tank 3, and a pump 9 that pumps rolling oil OL from the clean tank 3. It should be noted that a filter for removing foreign matter can also be installed between the clean tank 3 and the pump 9.

[0043] The rolling oil supply system of the cold rolling mill 1 has a supply line 11 with one end connected to the clean tank 3 and five sets of lubricating coolant manifolds 12 and five sets of cooling coolant manifolds 13 that branch off at the other end of the supply line 11 (on the mill side) and are respectively arranged at positions corresponding to each rolling mill stand.

[0044] Each lubrication coolant manifold 12 is located on the inlet side of the rolling mill stand, supplying lubricating oil to the roll gap and work rolls by spraying rolling oil OL, which is used as lubricating oil, from separately provided nozzles toward the roll gap. The cooling coolant manifold 13 is located on the outlet side of the rolling mill stand, cooling the rolls by spraying rolling oil OL from separately provided nozzles toward the rolls.

[0045] With this structure, the emulsion rolling oil OL in the clean tank 3 is pumped by pump 9 to supply line 11, and supplied to the lubrication coolant manifold 12 and cooling coolant manifold 13 located on each rolling mill stand, and supplied to the spraying parts from the separately provided nozzles. In addition, the emulsion rolling oil OL supplied to the rolls, except for the portion carried out of the system by the steel plate S or lost due to evaporation, is recovered by oil pan 5 and returned to the dirty tank 2 via return piping 6. Thereafter, a portion of the emulsion rolling oil accumulated in the dirty tank 2 is returned to the clean tank 3 after a certain amount of soluble iron components generated by cold rolling are removed by the iron powder removal device 8.

[0046] The rolling oil supply system described above circulates rolling oil to the rolls after the wear parts have been removed. In other words, the supplied emulsion rolling oil is recycled. It should be noted that the clean tank 3 corresponds to the circulating rolling oil tank in conventional circulating oil supply methods. As described above, the clean tank 3 is appropriately replenished (supplied) with the original rolling oil.

[0047] [Shape Control Prediction Model]

[0048] Next, refer to Figures 1-6 This describes a shape control prediction model as one embodiment of the present invention.

[0049] The functions associated with the shape control prediction model, which is an embodiment of the present invention, are provided by Figure 1 The rolling control device 100, the calculation unit 200, and the steel plate information measuring device 300 shown are implemented.

[0050] The rolling control device 100 controls the rolling conditions of the cold rolling mill 1 based on the control signals from the computing unit 200.

[0051] Figure 2 It is shown Figure 1 The diagram shows the structure of the arithmetic unit 200. (As shown...) Figure 2 As shown, the arithmetic unit 200 includes an arithmetic device 210, an input device 220, a storage device 230, and an output device 240.

[0052] The arithmetic unit 210 is wired to the input device 220, the storage device 230, and the output device 240 via the bus 250. However, the arithmetic unit 210, the input device 220, the storage device 230, and the output device 240 are not limited to this connection method; they can also be connected wirelessly, or a combination of wired and wireless connections.

[0053] The input device 220 functions as an input port for the control information of the cold rolling mill 1 generated by the rolling control device 100, the information of the steel plate entering the rolling mill measured by the steel plate information measuring device 300 (information related to the steel plate S entering the cold rolling mill 1 (e.g., steel grade, plate thickness before rolling, plate width, etc.)), and information from the operation monitoring device 400. The information from the operation monitoring device 400 includes execution instruction information of the shape control prediction model, information related to the steel plate S that is the object of rolling (previous process conditions, steel grade, dimensions), and cold rolling condition information (numerical information, text information, and image information) set by the process control computer or operator before cold rolling.

[0054] The storage device 230, for example, is composed of a hard disk drive, a semiconductor drive, an optical drive, etc., and is a device for storing the information required in this system (the information required for the implementation of the functions of the prediction model generation unit 214 and the prediction model execution unit 215, which will be described later).

[0055] Information required for the realization of the function of the prediction model generation unit 214 includes, for example, information related to cold rolling, such as the information of the rolled-in side steel plate measured by the steel plate information measuring device 300, the required characteristics of the steel plate S (steel grade, product thickness, plate width, etc.), the equipment constraints of the cold rolling mill 1, the rolling information after the weld point of the steel plate S passes through (including coil information, shape actuator position), the properties of the coolant used in the rolling mill stand, rolling conditions (including target rolling speed), and other explanatory variables related to cold rolling, as well as information of the rolled-out side steel plate (including shape parameters such as the first to fourth order composition of the shape of the rolled-out side steel plate, steepness, edge reduction ratio (thickness reduction rate at the end of the steel plate), etc.).

[0056] It should be noted that the 1st to 4th order components of the steel plate shape can be calculated using the mathematical formulas (1) to (4) shown below. That is, the shape parameters Λ2 and Λ4 representing the symmetric components are calculated using the mathematical formulas (1) and (2) shown below, and the shape parameters Λ1 and Λ3 representing the asymmetric components are calculated using the mathematical formulas (3) and (4) shown below. Among them, the parameters λ1 to λ4 in mathematical formulas (1) to (4) show the coefficients when the steel plate shape Y is approximated by the 4th order function shown in the mathematical formula (5) below, taking the elongation as the steel plate shape Y, taking the dimensionless coordinate x (-1≤x≤1) of the plate width in the width direction. In addition, the steepness refers to the value of the wave height δ and its spacing P of the rolled steel plate S defined by λ=δ / P.

[0057] [Mathematical Expression 1]

[0058]

[0059] [Mathematical Expression 2]

[0060]

[0061] [Mathematical Expression 3]

[0062]

[0063] [Mathematical Expression 4]

[0064]

[0065] [Mathematical Expression 5]

[0066]

[0067] Information required for the implementation of the function of the prediction model execution unit 215 includes, for example, the shape control prediction model for each rolling state of the steel plate S generated by the prediction model generation unit 214, and various information and shape constraints input into the shape control prediction model. Here, shape constraints refer to the conditions that serve as the basis for determining whether the shape of the steel plate at the exit side of the cold rolling mill 1 is acceptable. For example, a range is appropriately set in advance for each of the first to fourth order composition, steepness, and edge drop ratio of the steel plate shape at the exit side as acceptable.

[0068] The output device 240 functions as an output port for outputting control signals from the arithmetic unit 210 to the rolling control device 100.

[0069] The operation monitoring device 400 is equipped with any display device such as an LCD or an OLED. The operation monitoring device 400 receives various information indicating the operating status of the cold rolling mill 1 from the rolling control device 100 and displays the received information on an operation screen (operation screen) for the operator to monitor the operating status of the cold rolling mill 1.

[0070] The arithmetic unit 210 includes RAM (Random Access Memory) 211, ROM (Read Only Memory) 212 and arithmetic processing unit 213.

[0071] ROM 212 stores a prediction model generation program 212a and a prediction model execution program 212b, which are computer programs.

[0072] The arithmetic processing unit 213 has arithmetic processing functions and is connected to RAM 211 and ROM 212 via bus 250.

[0073] RAM 211, ROM 212 and arithmetic processing unit 213 are connected to input device 220, storage device 230 and output device 240 via bus 250.

[0074] The computation processing unit 213 includes a prediction model generation unit 214 and a prediction model execution unit 215 as function blocks.

[0075] The prediction model generation unit 214 is a processing unit that generates a shape control prediction model based on a machine learning method. This machine learning method involves associating pre-rolling data and rolling conditions of steel plates S from past rolling records in the cold rolling mill 1 with post-rolling data of steel plates S corresponding to each pre-rolling data from past rolling records. In this embodiment, a neural network model is used as the shape control prediction model based on the machine learning method. However, the machine learning method is not limited to neural networks, and other known machine learning methods may also be employed.

[0076] The prediction model generation unit 214 includes a learning data acquisition unit 214a, a preprocessing unit 214b, a first data conversion unit 214c, a model generation unit 214d, and a result storage unit 214e. When the prediction model generation unit 214 receives an instruction from the operation monitoring device 400 to generate a shape control prediction model, it executes the prediction model generation program 212a stored in the ROM 212, thereby functioning as the learning data acquisition unit 214a, the preprocessing unit 214b, the first data conversion unit 214c, the model generation unit 214d, and the result storage unit 214e. The shape control prediction model is updated each time the prediction model generation unit 214 executes.

[0077] The learning data acquisition unit 214a acquires multiple learning data sets as pre-processing for generating a shape control prediction model. These multiple learning data sets use past rolling performance data, including steel plate information from the steel plate information measuring device 300 at the mill inlet, and rolling conditions, as input performance data (explanatory variables). They also use steel plate information at the mill outlet as output performance data (target variables). Specifically, the learning data acquisition unit 214a acquires multiple learning data sets. These multiple learning data sets use at least one of the thickness information and temperature information in the width and length directions of the steel plate S measured at the mill inlet, along with past rolling performance data of the coil, as input performance data. They also use the shape parameters calculated based on the shape of the steel plate at the outlet of the cold rolling mill 1 during cold rolling using this input performance data as output performance data. The learning data acquisition unit 214a acquires the aforementioned input and output performance data from the storage device 230 to create learning data. Each learning data set consists of a group of input and output performance data. The learning data is stored in the storage device 230. Alternatively, the learning data acquisition unit 214a may not store the learning data in the storage device 230, but instead supply the learning data to the preprocessing unit 214b and the model generation unit 214d.

[0078] The input performance data includes multidimensional array information obtained by linking explanatory variables along the time axis. In this embodiment, the multidimensional array information is provided by, for example... Figure 3 The information shown in (a) to (c).

[0079] Figure 3 (a) An example is shown where the steel plate information measuring device 300 has only one measuring point. In this case, the learning data acquisition unit 214a copies the data along the width direction of the steel plate S for the measuring points that are continuously measured relative to the length direction of the steel plate S. This data is then used to create an array where the vertical columns form the width direction and the horizontal columns form the acquisition intervals. A multidimensional array of information is created, linking explanatory variables, and used as input performance data. These explanatory variables are selected based on the information of the coil and past rolling performance. The number of vertical columns, horizontal columns, and explanatory variables is not particularly limited.

[0080] Figure 3 (b) An example is shown where the measurement points of the steel plate information measuring device 300 are scanned relative to the width direction of the steel plate S. In this case, the learning data acquisition unit 214a copies the data along the length direction of the steel plate S for the measurement points that are measured continuously and in a wavy pattern relative to the length direction of the steel plate S. Figure 3 Similarly, the example shown in (a) is created as a multidimensional array of information that links the explanatory variables, and becomes the input performance data.

[0081] Figure 3 (c) An example is shown where the steel plate information measuring device 300 has multiple measuring points in the width direction of the steel plate S. In this case, the learning data acquisition unit 214a continuously measures the set of measuring points relative to the length direction of the steel plate S, and... Figure 3 Similarly, in the example shown in (a), multidimensional array information is created by linking the explanatory variables and used as input performance data.

[0082] It should be noted that the information measured by the steel plate information measuring device 300 is at least one of plate thickness and temperature information. The measurement method for the plate thickness gauge is not particularly limited; it can be contact or non-contact (gamma rays, X-rays, etc.). Similarly, the thermometer is not limited; it can be contact or a non-contact type such as a radiation thermometer. Furthermore, if the steel plate information measuring device 300 is a thermometer, a steel plate heating device for imparting temperature to the steel plate S can be installed on the upstream side.

[0083] It should be noted that when the storage device 230 does not store past rolling performance data (e.g., rolling conditions or steel grade conditions for which there are no past performance records) or the sample size is small, the learning data acquisition unit 214a may request the operator to perform cold rolling without using the shape control prediction model once or multiple times. Furthermore, the more learning data stored in the storage device 230, the higher the prediction accuracy based on the shape control prediction model. Therefore, even if the amount of learning data is less than a preset threshold, the learning data acquisition unit 214a may still request the operator to perform cold rolling without using the shape control prediction model until the data amount reaches the threshold.

[0084] The preprocessing unit 214b processes the learning data acquired by the learning data acquisition unit 214a into a shape control prediction model for generation. Specifically, in order to enable the neural network model to read the rolling performance data constituting the learning data, the preprocessing unit 214b standardizes (normalizes) the value range of the input performance data between 0 and 1 as needed.

[0085] The input performance data is multidimensional information. Therefore, the first data conversion unit 214c uses a convolutional neural network to compress the dimensionality of the input performance data while retaining the feature quantities, making it one-dimensional information (see reference). Figure 4 The input performance data has been transformed into one-dimensional information. Figure 4 The input layer 501 of the neural network model shown is combined.

[0086] Here, refer to Figure 5 This section describes a processing example of the first data conversion unit 214c. Figure 5 This is a flowchart illustrating the process of converting multidimensional array information into one-dimensional information. For example... Figure 5 As shown, the process of converting multidimensional array information into one-dimensional information, i.e., the method for storing multidimensional array information, has a structure with multiple filters connected in multiple stages from input to output. That is, the process of converting multidimensional array information into one-dimensional information includes, from the input side, a first convolution step S1, a first pooling step S2, a second convolution step S3, a second pooling step S4, and a fully connected step S5.

[0087] In the first convolution step S1, the first data conversion unit 214c takes a 64×64 multidimensional array as input and outputs a 64×64 first feature map through convolution. The first feature map indicates what kind of local features exist in which part of the input array. In the convolution operation, for example, a 3×3 pixel, 32-channel filter is used, the filter application interval is set to 1, and the length of the surrounding zero-filling is set to 1.

[0088] In the first pooling step S2, the first data conversion unit 214c takes the first feature map output from the first convolution step S1 as input and sets the maximum value within a 3×3 pixel area of ​​the first feature map as a new pixel. The first data conversion unit 214c performs this operation across the entire map while shifting pixels. As a result, in the first pooling step S2, the first data conversion unit 214c outputs a second feature map obtained by compressing the first feature map.

[0089] In the second convolution step S3, the first data conversion unit 214c takes the second feature map as input and outputs the third feature map through convolution operation. In the convolution operation, for example, a 32-channel filter with a horizontal × vertical pixel size is set, the filter application interval is set to 1, and the length of the surrounding zero-filled area is set to 1.

[0090] In the second pooling step S4, the first data conversion unit 214c takes the third feature map output from the second convolution step S3 as input and sets the maximum value within the horizontal × vertical 3 pixels of the third feature map as a new pixel. The first data conversion unit 214c performs this operation across the entire map while shifting pixels. As a result, in the second pooling step S4, the first data conversion unit 214c outputs a fourth feature map obtained by compressing the third feature map.

[0091] In the fully connected step S5, the first data transformation unit 214c arranges the information of the fourth feature map output from the second pooling step S4 into a column. Then, the 100 neurons output from the fully connected step S5 become... Figure 4 The neural network model shown has an input layer of 501. It should be noted that the convolution method and the number of output neurons are not limited to those described above. Furthermore, known models such as GoogleNet, VGG16, MOBILENET, and EFFICIENTNET can also be used as convolutional neural network methods.

[0092] return Figure 2 The model generation unit 214d uses machine learning (including information converted by the first data conversion unit 214c) to obtain multiple learning data from the preprocessing unit 214b to generate a shape control prediction model. The shape control prediction model includes rolling in side steel plate information and explanatory variables (the coil information and past rolling performance) as input performance data, and rolling out side steel plate information as output performance data.

[0093] In this embodiment, since a neural network is used as a machine learning method, the model generation unit 214d generates a neural network model as a shape control prediction model. That is, the model generation unit 214d generates a neural network model as a shape control prediction model, and this model is processed into a correlation between input performance data (including rolling performance data of the rolled-in side steel plate information) and output performance data (rolled-out side steel plate information) from the learning data used for generating the shape control prediction model. The neural network model is expressed, for example, using a functional expression.

[0094] Specifically, the model generation unit 214d sets the hyperparameters used in the neural network model and performs learning based on the neural network model using the hyperparameters. As an optimization calculation of the hyperparameters, the model generation unit 214d first generates a neural network model by changing some of the hyperparameters in stages on the learning data, and selects the hyperparameters that have the highest prediction accuracy relative to the validation data.

[0095] As hyperparameters, the number of hidden layers, the number of neurons in each hidden layer, the dropout rate in each hidden layer (cutting off the transmission of neurons with a certain probability), the activation function in each hidden layer, and the number of outputs are typically set, but are not limited to these. Furthermore, the optimization method for hyperparameters is not particularly limited, but grid search with phased parameter changes, random search with randomly selected parameters, or Bayesian optimization-based search can be used.

[0096] It should be noted that the model generation unit 214d is incorporated as part of the computing unit 210, but its structure is not limited thereto. For example, shape control prediction models can also be generated and saved in advance, and then read out appropriately.

[0097] like Figure 4 As shown, the neural network model that serves as the shape control prediction model in this embodiment includes an input layer 501, an intermediate layer 502, and an output layer 503 in sequence from the input side.

[0098] exist Figure 3 The multidimensional array information generated in the learning data acquisition unit 214a is compressed in dimension by using a convolutional neural network to retain the state of feature quantities, and then saved as a state of one-dimensional information to the input layer 501.

[0099] The intermediate layer 502 consists of multiple hidden layers, each containing multiple neurons. The number of hidden layers within the intermediate layer 502 and the number of neurons in each hidden layer are not particularly limited. In the intermediate layer 502, the transmission from a neuron to neurons in the next hidden layer, along with the weighting of variables by weighting coefficients, is performed via an activation function. For the activation function, a sigmoid function, a hyperbolic tangent function, or a ramp function can be used.

[0100] The output layer 503 combines the information from the neurons transmitted from the intermediate layer 502 and outputs it as a shape constraint determination value relative to the final cold rolling. The number of outputs constructed within the output layer 503 is not particularly limited. Based on the output result, the past rolling performance of the steel plate S during cold rolling (information on the steel plate at the rolling entry point and operating conditions), and the current rolling constraint performance (plate shape determination), the model learns by gradually optimizing the weighting coefficients within the neural network model.

[0101] After the weighting coefficients of the neural network model are learned, the model generation unit 214d inputs the evaluation data (the actual rolling conditions of the steel plate S to be rolled, which uses the shape control prediction model) into the neural network model after the weighting coefficients have been learned, and obtains the estimation result relative to the evaluation data.

[0102] return Figure 2 The result storage unit 214e stores the learning data, evaluation data, parameters (weighting coefficients) of the neural network model, the output of the neural network model relative to the learning data, and the output of the neural network model relative to the evaluation data in the storage device 230.

[0103] In the cold rolling of the steel sheet S, the prediction model execution unit 215 uses a shape control prediction model generated by the prediction model generation unit 214 to predict the shape parameters of the cold-rolled steel sheet S corresponding to the rolling conditions of the steel sheet S to be rolled. Then, the prediction model execution unit 215 determines the target rolling conditions of the steel sheet S to be rolled.

[0104] To perform the above processing, the prediction model execution unit 215 includes an information reading unit 215a, a second data conversion unit 215b, a rolling shape prediction unit 215c, a rolling condition determination unit 215d, and a result output unit 215e. Here, when the prediction model execution unit 215 receives a signal from the rolling control device 100 via the input device 220 indicating that cold rolling is being performed, it executes the prediction model execution program 212b stored in the ROM 212, thereby functioning as the information reading unit 215a, the second data conversion unit 215b, the rolling shape prediction unit 215c, the rolling condition determination unit 215d, and the result output unit 215e.

[0105] The information reading unit 215a reads from the storage device 230 the rolling conditions of the steel plate S to be rolled, which are set by the process control computer and the operator through the operation monitoring device 400.

[0106] The second data conversion unit 215b performs convolution of the multidimensional array information that will become the input data to the shape control prediction model into one-dimensional information. The processing of the second data conversion unit 215b is the same as that of the first data conversion unit 214c, so a detailed description of the processing is omitted. Alternatively, the first data conversion unit 214c and the second data conversion unit 215b can be subroutineted as a single processing unit.

[0107] The rolling shape prediction unit 215c inputs the one-dimensional information convolved by the second data conversion unit 215b into the shape control prediction model to predict the shape parameters of the cold rolling mill exit side of the steel plate S, which is the object of rolling.

[0108] The rolling condition determination unit 215d processes the steel plate S such that its shape parameters are within a separately set shape constraint determination threshold as follows: it sets the target rolling condition in the change explanatory variables and repeats the processing of the above-mentioned information reading unit 215a, second data conversion unit 215b and rolling shape prediction unit 215c.

[0109] When the shape parameters of the rolled steel plate S are within the preset shape constraint determination threshold, the result output unit 215e operates and outputs the determined rolling conditions (shape control actuator quantity) of the steel plate S as the rolling object.

[0110] Next, refer to Figure 6 Explain the processing of the prediction model execution unit 215.

[0111] Figure 6 This is a flowchart illustrating the processing flow of the prediction model execution unit 215. For example... Figure 6 As shown, when executing the shape control prediction model, firstly, as part of step S11, the information reading unit 215a of the prediction model execution unit 215 reads from the storage device 230 a neural network model that corresponds to the required characteristics of the steel plate S, which is the object to be rolled, as the neural network model for shape control prediction.

[0112] Next, as part of step S12, the information reading unit 215a reads the required shape constraint determination threshold stored in the storage device 230 from the host computer via the input device 220. Next, as part of step S13, the information reading unit 215a reads the rolling conditions of the steel plate S to be rolled, stored in the storage device 230 from the host computer via the input device 220.

[0113] Next, as part of step S14, the rolling shape prediction unit 215c uses the neural network model, which is the shape control prediction model read in step S11, as input data obtained by multi-dimensionalizing the rolling conditions of the steel plate S to be rolled, read in step S13, to calculate the shape parameters of the corresponding cold-rolled steel plate S. It should be noted that the prediction results based on the neural network model are output to... Figure 4 The output layer 503 of the neural network model shown.

[0114] Next, as part of step S15, the rolling condition determination unit 215d determines whether the shape parameters of the steel plate S obtained in step S14 are within the shape constraint determination threshold read in step S12. It should be noted that if the calculation convergence is insufficient, an upper limit can be set on the number of convergence iterations within the actual calculation time that can be performed in step S15. The shape parameters being within the shape constraint determination threshold is equivalent to satisfying the conditions specified in this invention.

[0115] Then, if the shape parameters are within the shape constraint determination threshold (step S15: Yes), the prediction model execution unit 215 ends a series of processes. On the other hand, if the shape parameters are not within the shape constraint determination threshold (step S15: No), the prediction model execution unit 215 proceeds to the processing in step S16.

[0116] In step S16, the rolling condition determination unit 215d modifies a portion of the rolling conditions (e.g., shape control actuator operation amount) of the steel plate S to be rolled, which was read in step S13, and proceeds to step S17. In step S17, the result output unit 215e transmits information related to the modified rolling conditions to the rolling control device 100 via the output device 240.

[0117] When a portion of the rolling conditions is changed during the processing in step S16, the rolling condition determination unit 215d, during the processing in step S17, determines the rolling conditions of the steel plate S to be rolled as optimized rolling conditions, specifically, the modified bending or shifting amount of the work roll or intermediate roll. Then, the rolling condition determination unit 215d determines the operating amount of the shape control actuator based on the current rolling conditions. During the cold rolling stage, the rolling control device 100 changes the rolling conditions based on information related to the shape control actuator transmitted from the result output unit 215e.

[0118] As a method for calculating the amount of change in rolling conditions, the rolling condition determination unit 215d calculates appropriate rolling conditions for the steel plate S to be rolled based on the difference between the shape parameters obtained in step S14 and the shape constraint determination threshold read in step S12. Then, the rolling condition determination unit 215d compares the calculated rolling conditions with the rolling conditions of the steel plate S to be rolled read in step S13, and changes the rolling conditions in step S17.

[0119] When returning to step S13, the rolling shape prediction unit 215c reads the rolling conditions of the steel plate S to be rolled, which have been modified by a portion of the rolling conditions. Furthermore, in step S14, the rolling shape prediction unit 215c uses a neural network model as a shape control prediction model to determine the shape parameters of the cold-rolled steel plate S corresponding to the modified rolling conditions of the steel plate S to be rolled, which were read in step S13. In step S15, the rolling condition determination unit 215d determines whether the shape parameters determined in step S14 are within the shape constraint determination threshold read in step S12. Then, steps S13, S14, S15, S16, and S17 are repeatedly executed until the determination result is YES. Thus, the processing of the prediction model execution unit 215 (shape control determination step) ends.

[0120] As can be seen from the above description, in this embodiment, the prediction model generation unit 214 generates a shape control prediction model based on a machine learning method, which establishes a correlation between past rolling performance of the steel plate S and the corresponding past shape control performance. Furthermore, the prediction model execution unit 215 uses the generated shape control prediction model to determine the shape parameters of the steel plate S to be rolled during cold rolling. Then, the prediction model execution unit 215 determines the rolling conditions of the steel plate S to be rolled in such a way that the determined shape parameters are within the shape constraint determination threshold. Thus, by implementing shape control that is independent of operator experience and subjectivity, and satisfies various constraints in the rolling operation, it is possible to suppress the occurrence of shape defects, fractures, and other faults in cold rolling and maintain productivity. Moreover, according to this embodiment, the explanatory variables used in the shape prediction of the steel plate S in cold rolling are numerical information collected from rolling performance data and multidimensional array information is used as input data. Therefore, it is possible to identify factors that contribute significantly to constraints generated in cold rolling on a neural network model.

[0121] [Variation Example]

[0122] While embodiments of the present invention have been described above, the present invention is not limited thereto and various modifications and improvements are possible. For example, in this embodiment, the shape prediction of the steel plate S based on the shape control prediction model and the determination of rolling conditions are repeated along the entire length of the coil, but it can also be performed in a portion of the coil. Furthermore, the cold rolling mill 1 is not limited to a four-stage type; it can also be a two-stage (2Hi), a six-stage (6Hi), or other multi-stage mills, and the number of rolling stands is not particularly limited. Alternatively, it can be a multi-roll mill or a Sendzimir mill.

[0123] Furthermore, if the calculation unit 200 calculates an abnormal control quantity exceeding the upper and lower limits of the shape control actuator, or if the control quantity cannot be calculated, the rolling control device 100 cannot execute control based on the instructions from the calculation unit 200. Therefore, it is best not to perform this embodiment if the rolling control device 100 determines that the control quantity from the calculation unit 200 is abnormal or that no control quantity is supplied from the calculation unit 200.

[0124] In addition, Figure 2 In the structural example shown, the output device 240 and the operation monitoring device 400 are not connected, but they can be connected in a communicative manner. As a result, the processing results of the prediction model execution unit 215 (especially the shape prediction information of the rolled steel plate S obtained by the rolling shape prediction unit 215c and the changed rolling conditions determined by the rolling condition determination unit 215d) can be displayed on the operation screen of the operation monitoring device 400.

[0125] Example

[0126] The present invention will now be described based on embodiments.

[0127] use Figure 1 The cold rolling mill of the embodiment shown, consisting of a total of five rolling mills, was used to conduct experiments on cold rolling of a base steel sheet containing 2.5% by mass Si for electromagnetic steel sheets with a base material thickness of 2.0 mm and a plate width of 1000 mm to a final thickness of 0.300 mm. The following stock solution was used as the rolling oil: a base oil obtained by adding vegetable oil to synthetic ester oil, with 1% by mass of an oiliness agent and an antioxidant added respectively; and a nonionic surfactant added at a concentration relative to the oil of 3% by mass. Additionally, recycled emulsion rolling oil was prepared to a rolling oil concentration of 3.5% by mass, an average particle size of 5 μm, and a temperature of 55°C. As a prior learning process, learning was first performed using training data (rolling performance data of approximately 3000 past steel sheets) to implement a neural network model. This established a correlation between past steel sheet rolling performance data and created a neural network model used for predicting steel sheet shape.

[0128] In the invention example, the rolling performance data for conventional steel plates includes, in addition to the strip steel plate information measured on the width direction of the steel plate at the rolling entry side, information including the deformation resistance of the steel plate, rolling pass planning (rolling load / tension / steel plate shape / thickness accuracy), emulsion properties, work roll size / camber / roughness information, bending amount, and work roll displacement amount. Furthermore, a multidimensional array information formed by copying and linking the above rolling performance data is used as input performance data. The rolling performance data for conventional steel plates was used to learn the shape performance of the rolled-out side steel plate. Using a cold rolling mill, the roll gap is adjusted, and after the steel plate passes through the weld point, during the stage when the rolling control device 100 is turned on, the shape of the cold-rolled steel plate is predicted based on a generated neural network model. Then, the rolling conditions are changed sequentially in such a way that the predicted shape is within a specified shape constraint judgment threshold, and the rolling conditions are set.

[0129] In the comparative examples, similar to the inventive examples, an experiment was conducted to cold roll a raw material steel sheet (the object to be rolled) containing 2.8% by mass Si and with a base material thickness of 1.8 mm and a plate width of 1000 mm to a plate thickness of 0.3 mm. In the comparative examples numbered 1, 3, 5, 7, 9, and 11 shown in Table 1, past steel sheet rolling performance data were used as input data for a one-dimensional array without copying them in the time direction to establish a correlation between past steel sheet shape performance data, thereby generating a neural network model used for predicting steel sheet shape.

[0130] Table 1 shows the number of fractures in the steel sheets after rolling 100 coils of the invention example and the comparative example. As shown in Table 1, in the comparative example, due to insufficient training, failures such as deep drawing fracture occurred when the convexity of the infeasor plate changed too much, exceeding the operational constraints.

[0131] From the above, it is confirmed that, preferably, by using the cold rolling method and cold rolling mill of the present invention, the shape of the steel sheet during rolling is appropriately predicted, and the rolling conditions are changed successively in such a way that the predicted shape parameters are within a pre-set shape constraint determination threshold to determine the shape of the rolled steel sheet. Furthermore, it is confirmed that by applying the present invention, not only can defects such as shape defects and sheet breakage be suppressed during cold rolling, but the productivity and quality of the rolling process and subsequent processes can also be greatly improved.

[0132] [Table 1]

[0133]

[0134] The embodiments of the invention implemented by those using the present invention have been described above. However, the present invention is not limited to the description and drawings that constitute a part of the disclosure of the present invention based on these embodiments. That is, all other embodiments, examples, and techniques applied by those skilled in the art based on these embodiments are included within the scope of the present invention.

[0135] Industrial availability

[0136] According to the present invention, a method and apparatus for setting rolling conditions for a cold rolling mill can be provided, which can ensure the stability of cold rolling while maintaining good productivity when rolling difficult-to-roll materials with high loads and thin sheet thickness before rolling. Furthermore, according to the present invention, a cold rolling method and a cold rolling mill can be provided, which can ensure the stability of cold rolling while maintaining good productivity when cold rolling difficult-to-roll materials with high loads and thin sheet thickness before rolling. Additionally, according to the present invention, a method for manufacturing steel sheets that can produce steel sheets with good yield can be provided.

[0137] Explanation of reference numerals in the attached figures

[0138] 1. Cold rolling mill

[0139] 2. Dirty container (recycling container)

[0140] 3 Clean tanks

[0141] 5. Oil pan

[0142] 6 Return to piping

[0143] 7. Mixer

[0144] 8. Iron powder removal device

[0145] 9 pumps

[0146] 11 Supply Lines

[0147] 12. Lubrication coolant manifold

[0148] 13 Coolant manifold for cooling

[0149] 100 Rolling Control Device

[0150] 200 arithmetic units

[0151] 210 Computing Device

[0152] 211 RAM (Random Access Memory)

[0153] 212 ROM (Read Only Memory)

[0154] 212a Predictive Model Generation Program

[0155] 212b Prediction Model Execution Procedure

[0156] 213 Computational Processing Unit

[0157] 214 Prediction Model Generation Department

[0158] 214a Learning Data Acquisition Department

[0159] 214b Pre-processing unit

[0160] 214c First Data Conversion Unit

[0161] 214d Model Generation Department

[0162] 214e Results Preservation Department

[0163] 215 Predictive Model Execution Department

[0164] 215a Information Reading Department

[0165] 215b Second Data Conversion Unit

[0166] 215c Rolling Shape Prediction Section

[0167] 215d Rolling Conditions Determination Department

[0168] 215e Result Output Section

[0169] 220 Input Device

[0170] 230 storage device

[0171] 240 Output Device

[0172] 300 Steel Plate Information Measuring Device

[0173] 400 Operation monitoring device

[0174] 501 Input Layer

[0175] 502 Intermediate Layer

[0176] 503 Output Layer S steel plate

Claims

1. A method for setting rolling conditions for a cold rolling mill, which uses a predictive model to predict the post-cold-rolled state of the material to be rolled to set the target rolling conditions for the cold rolling mill when cold rolling the material. The prediction model is generated by using the first multidimensional data as the explanatory variable and the post-cold-rolled data of the rolled material at the exit side of the cold rolling mill as the target variable. The first multidimensional data is generated by linking past rolling performance data, including the pre-cold-rolled data of the rolled material at the inlet side of the cold rolling mill, in the time direction. The method for setting rolling conditions for the cold rolling mill includes: estimating the post-rolled shape of the material to be rolled at the exit side of the cold rolling mill by inputting second multi-dimensional data into the prediction model, wherein the second multi-dimensional data is generated based on information including pre-cold rolling data of the material to be rolled at the inlet side of the cold rolling mill and the target rolling conditions of the cold rolling mill; and The step of changing the target rolling conditions of the cold rolling mill in a manner that presumes the post-rolled shape meets the specified conditions.

2. The method for setting rolling conditions for a cold rolling mill according to claim 1, wherein, The data prior to cold rolling includes at least one of the thickness and temperature information of the steel plate at the inlet side of the cold rolling mill.

3. The method for setting rolling conditions for a cold rolling mill according to claim 1 or 2, wherein, The post-cold rolling data includes shape parameters calculated based on the shape of the steel plate at the exit side of the cold rolling mill.

4. A cold rolling method comprising the step of cold rolling a material to be rolled using target rolling conditions of a cold rolling mill, wherein the target rolling conditions of the cold rolling mill are obtained by modifying the rolling condition setting method of the cold rolling mill according to any one of claims 1 to 3.

5. A method for manufacturing a steel plate, comprising the step of manufacturing the steel plate using the cold rolling method of claim 4.

6. A rolling condition setting device for a cold rolling mill, which uses a predictive model to predict the post-cold-rolled state of the material to be rolled to set the target rolling conditions of the cold rolling mill when cold-rolling the material. The prediction model is generated by using the first multidimensional data as the explanatory variable and the post-cold-rolled data of the rolled material at the exit side of the cold rolling mill as the target variable. The first multidimensional data is generated by linking past rolling performance data, including the pre-cold-rolled data of the rolled material at the inlet side of the cold rolling mill, in the time direction. The rolling condition setting device for the cold rolling mill includes: a mechanism for estimating the post-rolled shape of the material to be rolled at the exit side of the cold rolling mill by inputting second multi-dimensional data into the prediction model, wherein the second multi-dimensional data is generated based on information including pre-cold rolling data of the material to be rolled at the inlet side of the cold rolling mill and the target rolling conditions of the cold rolling mill; and An organization that changes the target rolling conditions of the cold rolling mill in a manner that presumes the rolled shape meets specified conditions.

7. The rolling condition setting device for a cold rolling mill according to claim 6, wherein, The data prior to cold rolling includes at least one of the thickness and temperature information of the steel plate at the inlet side of the cold rolling mill.

8. The rolling condition setting device for a cold rolling mill according to claim 6 or 7, wherein, The post-cold rolling data includes shape parameters calculated based on the shape of the steel plate at the exit side of the cold rolling mill.

9. A cold rolling mill, comprising a rolling condition setting device for a cold rolling mill as described in any one of claims 6 to 8.

Images