Cold rolling condition setting method, cold rolling method, cold rolled steel sheet manufacturing method, cold rolling condition calculation device, and cold rolling machine

Abstract

The cold rolling conditions are calculated so as to ensure the stability of rolling and not impair productivity even when rolling a material that is difficult to roll under high load. The cold rolling condition setting method of the present invention is a method for setting conditions for cold rolling a steel sheet having a total content of Si and Al of 3.0% or more, which has been successively subjected to hot rolling, annealing, and pickling, by using a prediction model, the prediction model being a prediction model trained using cold rolling conditions including information on the material to be rolled, annealing conditions, pickling conditions, and the amount of work rolls used at the first rolling stand in cold rolling as explanatory variables from past operational results, and an asymmetric component from the rolling results at the first rolling stand as a target variable, and includes a step of predicting the asymmetric component at the first rolling stand by inputting the annealing conditions, pickling conditions, and set cold rolling conditions at the first rolling stand for the material to be rolled into the prediction model, and a step of changing the set cold rolling conditions so that the predicted asymmetric component satisfies a predetermined condition.

Description

[Technical field]

[0001] The present invention relates to a cold rolling condition setting method, a cold rolling method, a cold rolled steel sheet manufacturing method, a cold rolling condition calculation device, and a cold rolling mill. [Background technology]

[0002] In general, when rolling cold-rolled thin steel sheets, the shape (or flatness) of the steel sheet is improved while maintaining good thickness accuracy in the longitudinal and transverse directions, thereby stabilizing the threadability of the steel sheet during rolling. Many of the control factors of the cold rolling mill are automatically controlled by actuators mounted on the cold rolling mill, and opportunities for operators to set the control factors of the cold rolling mill are decreasing.

[0003] On the other hand, there is an increasing need for thin, high-strength, thin-gauge hard materials for the purpose of reducing fuel consumption through weight reduction. In cold rolling of such difficult-to-roll materials, annealing is performed on the hot-rolled thin steel sheet to improve the non-uniformity of the structure and reduce the load during cold rolling, since the material contains a large amount of alloying elements. The oxide scale formed by annealing is then removed in a pickling process before being sent to the cold rolling process. During cold rolling of such difficult-to-roll materials, which are subject to high loads and contain a large amount of alloying elements, the distribution of remaining oxide scale can vary greatly in the longitudinal direction of the coil. In areas where oxide scale remains partially, oil adhesion to the steel sheet is inhibited, which causes the friction state in the roll bite to vary, and thus the rolling load (and the forward slip and torque calculated accordingly) to vary. Therefore, in many cases, the variations in the roll gap of the cold rolling mill, the work roll bender, intermediate roll shift, and roll expansion due to thermal crown, which are representative of roll deflection compensation, cannot be absorbed by automatic control. In this case, the operator sets the pass schedule and the shape control actuators so as to satisfy the equipment constraints of the cold rolling mill and not to impede productivity.

[0004] For this reason, in recent years, the operating speed of a cold rolling mill, and therefore the productivity, has become more and more dependent on the experience of the operator. In light of this background, Patent Document 1 proposes a method of learning past operating conditions using a neural network and performing mill setup of a rolling mill using the learning results.

[0005] In addition, there is a problem that the surface of the work roll is roughened because the oxide scale and the work roll are in metal contact without an oil film in the roll bite. In view of this, Patent Document 2 proposes a method in which the steel sheet before cold rolling is heated, bent and stretched using a tension leveler, and pickled. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] JP 2020-134967 A [Patent Document 2] JP 2020-059050 A Summary of the Invention [Problem to be solved by the invention]

[0007] However, the method proposed in Patent Document 1 has the following problems. That is, in the method, past operating conditions are learned and the results are used to perform mill setup of the rolling mill. However, when there is a variation in the residual scale in the coil longitudinal direction, particularly in difficult-to-roll materials with a total Si and Al content of 3.0 mass% or more, the shape of the rolling mill outlet side varies greatly due to the variation in the friction coefficient. Therefore, even if the rolling mill is in the optimal operating condition at the time of mill setup, the rolling speed may be suppressed due to poor shape, and in the worst case, the sheet may break.

[0008] In addition, the method shown in Patent Document 2 is expected to enable mechanical scale removal using a tension leveler. However, since the elongation rate cannot be changed continuously along the length of the coil, it cannot handle cases where the scale thickness or scale composition varies along the length of the coil, such as when only one section of the coil is operated at a low speed during hot-rolled sheet annealing.

[0009] The present invention has been made in consideration of the above problems, and an object of the present invention is to ensure rolling stability without impairing productivity, even when cold rolling a difficult-to-roll material (steel plate) that is subjected to high load and contains a large amount of additive elements. [Means for solving the problem]

[0010] The present invention has been completed based on the above findings, and the gist of the present invention is as follows.

[0011] 1. A method for setting cold rolling conditions, which uses a prediction model to set cold rolling conditions when cold rolling a steel sheet having a total Si and Al content of 3.0 mass% or more that has been successively subjected to hot rolling, annealing, and pickling, The prediction model is a prediction model trained using cold rolling conditions, including information on the material to be rolled, annealing conditions, pickling conditions, and the amount of work rolls used in the first rolling stand in cold rolling, as explanatory variables, and using asymmetric components, among the rolling results in the first rolling stand, as objective variables, among the past operational results; A step of predicting an asymmetric component in the first rolling stand by inputting annealing conditions, pickling conditions, and cold rolling conditions set in the first rolling stand in cold rolling for a rolling target material to be rolled into the prediction model; and changing the set cold rolling conditions so that the predicted asymmetric component satisfies a predetermined condition.

[0012] 2. A cold rolling method comprising the steps of cold rolling a material under set cold rolling conditions changed by the cold rolling condition setting method described in 1 above.

[0013] 3. A method for producing cold-rolled steel sheet, comprising the steps of hot rolling, annealing and pickling in that order on a material to be rolled, and then cold rolling using the cold rolling method described in 2 above to produce a cold-rolled steel sheet.

[0014] 4. A cold rolling condition calculation device that uses a prediction model to set cold rolling conditions when cold rolling a steel plate having a total Si and Al content of 3.0 mass% or more that has been successively subjected to hot rolling, annealing, and pickling, The prediction model is a prediction model trained using cold rolling conditions, including information on the material to be rolled, annealing conditions, pickling conditions, and the amount of work rolls used in the first rolling stand in cold rolling, as explanatory variables, and using asymmetric components, among the rolling results in the first rolling stand, as objective variables, among the past operational results; A prediction means for predicting an asymmetric component in the first rolling stand by inputting annealing conditions, pickling conditions, and cold rolling conditions set in the first rolling stand in cold rolling for a material to be rolled into the prediction model; and changing means for changing the set cold rolling conditions so that the predicted asymmetric component satisfies a predetermined condition.

[0015] 5. A cold rolling machine equipped with the cold rolling condition calculation device described in 4 above. Effect of the Invention

[0016] According to the method for setting cold rolling conditions of the present invention, it is possible to calculate cold rolling conditions that ensure rolling stability and do not impair productivity, even when rolling a difficult-to-roll material that is subjected to a high load and contains many additive elements and has a total content of Si and Al of 3.0 mass% or more. Therefore, according to the present invention, it is possible to stably and productively roll a difficult-to-roll material that is subjected to a high load and contains many additive elements. [Brief description of the drawings]

[0017] [Figure 1] FIG. 1 is a schematic diagram showing the configuration of a production facility according to one embodiment of the present invention. [Diagram 2]FIG. 2 is a block diagram showing a configuration of the arithmetic unit shown in FIG. [Diagram 3] FIG. 3 is a graph showing the relationship between the pickling rate and the standard deviation of the differential load in the first rolling stand. [Figure 4] FIG. 4 is a flowchart showing a flow of processing by the prediction model executing unit. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] Hereinafter, a cold rolling method, a cold rolling condition calculation device for cold rolling, and a cold rolling mill according to an embodiment of the present invention will be described with reference to the drawings. Note that the embodiments shown below are merely examples of devices and methods for embodying the technical ideas of the present invention, and the materials, shapes, structures, arrangements, etc. of the components are not limited to the embodiments shown below. Also, the drawings are schematic. For this reason, it should be noted that the relationship, ratio, etc. between thickness and planar dimensions differ from the actual ones, and the drawings also include parts in which the relationship and ratio of dimensions differ from each other.

[0019] [Configuration of cold rolling mill] First, the configuration of a manufacturing facility according to an embodiment of the present invention will be described with reference to Fig. 1. In this specification, "cold rolling" may be simply referred to as "rolling", and in this specification, "cold rolling" and "rolling" have the same meaning.

[0020] In the present invention, a steel sheet having a total content of Si and Al of 3.0% by mass or more is used as the rolled material (rolling target material) to be rolled by a cold rolling mill. When the total content of Si and Al is 3.0% by mass or more, a strong oxide scale is likely to be formed during continuous annealing. When the total content of Si and Al is less than 3.0% by mass, a strong oxide scale is not formed, and scale removal by pickling is possible. On the other hand, when the total content of Si and Al is more than 6.0% by mass, the effect of scale removal by pickling is reduced. Therefore, the total content of Si and Al is preferably 6.0% by mass or less.

[0021] Fig. 1 is a schematic diagram showing the configuration of a manufacturing facility in one embodiment of the present invention, in which (i) shows a continuous annealing line and (ii) shows a cold rolling line connected to the continuous annealing line.

[0022] In the continuous annealing line (i), 2 is a payoff rule, 3 is an annealing furnace, and 4 is a pickling tank. The annealing furnace 3 and the pickling tank 4 are appropriately provided with transport rolls (not shown).

[0023] In the cold rolling line (ii), 5 is a cold tandem rolling mill, and 6 is a coiler. In the example shown in FIG. 1, the cold tandem rolling mill 5 is equipped with five rolling stands. Specifically, the cold tandem rolling mill 5 is equipped with five rolling stands, namely, the first rolling stand to the fifth rolling stand (#1STD to #5STD), in order from the entry side (left side of the paper in FIG. 1) to the exit side (right side of the paper in FIG. 1). In the cold tandem rolling mill 5, tension rolls and diff rolls, thickness gauges, and shape gauges (not shown) are appropriately installed between adjacent rolling stands. The configuration of the rolling stands and the conveying device for the steel sheet 1 are not particularly limited, and publicly known techniques may be applied as appropriate.

[0024] In addition, in the embodiment of FIG. 1, the continuous annealing line and the cold rolling line are connected to form one line, but the continuous annealing line and the cold rolling line may each be an independent line.

[0025] In the embodiment shown in Fig. 1, a steel sheet 1 rolled in a hot rolling line (not shown) is discharged from a payoff reel 2, and then passes through an annealing furnace 3 and a pickling tank 4. Next, the steel sheet 1 is cold rolled in a cold tandem rolling mill 5, and then coiled around a coiler 6.

[0026] The annealing method in the annealing furnace 3 is not particularly limited, and a vertical furnace, horizontal furnace, or batch furnace can be used. The oxides generated in the annealing furnace 3 are pickled (chemically ground) by passing through the pickling tank 4. In this pickling process, acid penetrates into the boundary between the base material and the surface oxide, and the surface oxide is peeled off and removed. The pickling solution used in the pickling is not particularly limited, but it is preferable to use a hydrochloric acid or sulfuric acid solution with a concentration of 5% by volume or more, or a nitric acid solution (a solution in which nitric acid at 5% by volume or more and hydrochloric acid at 0.5% by volume or more are mixed).

[0027] [Rolling control prediction model] Next, a rolling control prediction model in one embodiment of the present invention will be described.

[0028] Functions related to a rolling control prediction model according to one embodiment of the present invention are realized by a rolling control device 7, a calculation unit 8, and an operation information measuring device 9 shown in FIG.

[0029] The rolling control device 7 controls the cold rolling conditions of the tandem cold mill 5 based on a control signal from the arithmetic unit 8 .

[0030] Fig. 2 is a block diagram showing the configuration of the arithmetic unit 8 shown in Fig. 1. As shown in Fig. 2, the arithmetic unit 8 includes a calculation device 71, an input device 88, a storage device 89, and an output device 90.

[0031] The arithmetic device 71 is connected by wire to an input device 88, a storage device 89, and an output device 90 via a bus 87. However, the arithmetic device 71, the input device 88, the storage device 89, and the output device 90 are not limited to this connection mode, and may be connected wirelessly or in a combination of wired and wireless connections.

[0032] The input device 88 functions as an input port for inputting rolling results (plate thickness, deformation resistance, rolling load, tension, roll information (roll diameter, roll material, roll roughness, tonnage used since the roll was inserted (amount of work roll used)), rolling speed, etc.) from the rolling control device 7, annealing / pickling results (plate thickness, temperature, gas, speed, acid results, etc.) from the operation information measuring device 9, and information from the operation monitoring device 91. The information from the operation monitoring device 91 includes execution command information for a rolling control prediction model, information about the steel plate 1 to be rolled (previous process conditions, steel type, size), and cold rolling condition information (numeric information, character information, and image information) set by the process computer or an operator before cold rolling.

[0033] The storage device 89 is composed of, for example, a hard disk drive, a semiconductor drive, an optical drive, etc., and is a device that stores information necessary in this system (information necessary to realize the functions of the prediction model creation unit 77 and the prediction model execution unit 78 described later).

[0034] Information required for realizing the function of the prediction model creation unit 77 includes, for example, annealing, pickling, and rolling results in the first rolling stand from the operation information measurement device 9, required characteristics of the steel sheet 1 (steel type, plate thickness, plate width, etc.), mill equipment constraints, roll information used in rolling (roll diameter, roll material, roll roughness, tonnage used after the roll is inserted), properties of coolant used in the rolling stand, explanatory variables related to cold rolling such as target rolling speed, and objective variables. It is important that the roll information includes the tonnage used after the roll is inserted as the amount of work roll used in the first rolling stand in cold rolling. The surface of the work roll becomes rough depending on the rolled material and rolling tonnage, and roughness information in the barrel direction remains as an asymmetric component, but it is difficult to measure the roughness of the work roll online while lubricant is constantly being supplied. On the other hand, roll wear progresses due to the accumulation of rolling load and slip speed (speed difference between material and roll speed), so the correlation with roll roughness can be indirectly evaluated by the tonnage used after the roll is inserted. In the present invention, the asymmetric component of the rolling performance in the first rolling stand is used as the objective variable. Here, the "asymmetric component" refers to the operating conditions that are independently controlled on the operator side (OP side) and the drive side (DR side) during cold rolling, that is, the cold rolling conditions in which a difference occurs between the OP side and the DR side. In the following description, information on the asymmetric component may be referred to as asymmetric rolling information or asymmetric rolling data. The "OP side" refers to the operator side in the steel plate width direction, and the "DR side" refers to the equipment side opposite to the OP side.

[0035] Examples of information required to realize the functions of the prediction model executing unit 78 include a rolling control prediction model for each rolling state of the steel plate 1 created by the prediction model creation unit 77 and various information input to the rolling control prediction model.

[0036] The output device 90 functions as an output port for outputting a control signal from the arithmetic device 71 to the rolling control device 7 .

[0037] The operation monitoring device 91 is equipped with any display such as a liquid crystal display, an organic display, etc. The operation monitoring device 91 receives various information indicating the operation status of the cold tandem rolling mill 5 from the rolling control device 7, and displays this information on an operation screen (operation screen) for an operator to monitor the operation status of the cold tandem rolling mill 5.

[0038] The arithmetic device 71 includes a RAM 72, a ROM 73, and an arithmetic processing unit .

[0039] The ROM 73 stores a prediction model creating program 74 and a prediction model executing program 75.

[0040] The arithmetic processing unit 76 has an arithmetic processing function, and is connected to the RAM 72 and the ROM 73 via a bus 87 .

[0041] The RAM 72 , the ROM 73 , and the arithmetic processing unit 76 are connected to an input device 88 , a storage device 89 , and an output device 90 via a bus 87 .

[0042] The calculation processing unit 76 includes a prediction model creation unit 77 and a prediction model execution unit 78 as functional blocks.

[0043] The prediction model creation unit 77 is a processing unit that creates a rolling control prediction model by a machine learning technique that links past annealing / pickling / rolling results and rolling constraint conditions corresponding to the past rolling results in the cold tandem rolling mill 5. As the rolling control prediction model by the machine learning technique, a neural network model such as deep learning, a regression tree model such as random forest, a gradient boosting model such as XGBOOST, etc. can be used, but there is no particular limitation, and other known machine learning techniques may be adopted.

[0044] The prediction model creation unit 77 includes a learning data acquisition unit 77A, a first data pre-processing unit 77B, a model creation unit 77C, and a result storage unit 77D. When the prediction model creation unit 77 receives an instruction to create a rolling control prediction model from the operation monitoring device 91, the prediction model creation unit 77 executes the prediction model creation program 74 stored in the ROM 73, thereby functioning as the learning data acquisition unit 77A, the first data pre-processing unit 77B, the model creation unit 77C, and the result storage unit 77D. The rolling control prediction model is updated every time the prediction model creation unit 77 is executed.

[0045] The learning data acquisition unit 77A acquires a plurality of learning data as a pre-processing for generating a rolling control prediction model. In acquiring the learning data, among past rolling performance data, information on the material to be rolled, annealing conditions, pickling conditions from the operation information measuring device 9, and rolling performance data at the first rolling stand are set as input performance data (explanatory variables). In addition, the asymmetric rolling data at the first rolling stand is set as output performance data (objective variable).

[0046] Examples of the annealing conditions include line speed, furnace temperature, gas flow rate, dew point, cooling amount, sheet temperature, etc. Examples of the pickling conditions include acid concentration, additive concentration, acid temperature, line speed, etc. Examples of the rolling performance data include the inlet and outlet thicknesses, roll diameter, roll material, roll roughness, tonnage used after the roll is inserted, reduction, deformation resistance, rolling speed, rolling load, inlet and outlet tension, coolant flow rate, etc., in the rolling of the coil in the first rolling stand.

[0047] The asymmetric rolling data includes the asymmetric rolling state in the first rolling stand using the input performance data, such as the rolling differential load, the differential tension at the inlet and outlet, the difference in the roll reduction position, the lubricating coolant settings (flow rate difference) in the upper and lower work rolls, and the roll surface roughness (roughness difference between the upper and lower work rolls due to the remaining scale state on the front and back surfaces of the steel sheet).

[0048] The learning data acquisition unit 77A acquires the above-mentioned input actual data and output actual data from the storage device 89 to create learning data. Each piece of learning data consists of a pair of input actual data and output actual data. The learning data is stored in the storage device 89. The learning data acquisition unit 77A may supply the learning data to the first data pre-processing unit 77B or the model creation unit 77C without storing the learning data in the storage device 89. It is not necessary to input all of the data contained in each piece of information described above, and only a portion of the data may be input.

[0049] When no past rolling performance data is stored in the storage device 89 (for example, when the cold rolling conditions or steel type conditions have no past performance) or when the sample amount is small, the learning data acquisition unit 77A requests the operator to perform cold rolling once or multiple times without using the rolling control prediction model. Also, the greater the number of learning data stored in the storage device 89, the higher the prediction accuracy by the rolling control prediction model. For this reason, when the number of learning data is less than a preset threshold, the learning data acquisition unit 77A may request the operator to perform cold rolling without using the rolling control prediction model until the number of data reaches the threshold.

[0050] The first data pre-processing unit 77B processes the learning data acquired by the learning data acquisition unit 77A for creating a rolling control prediction model. Specifically, the first data pre-processing unit 77B standardizes (normalizes) the value range of the input actual data between 0 and 1 as necessary in order to read the rolling actual data constituting the learning data into the machine learning model. In addition, when abnormal data or unacquired data is included, a process of deleting or filling in the relevant data is performed.

[0051] The model creation unit 77C generates a rolling control prediction model that includes explanatory variables (the coil information and past annealing, pickling, and rolling results) as input actual data and has the asymmetric rolling state during cold rolling as output actual data, by machine learning using a plurality of learning data acquired by the first data pre-processing unit 77B.

[0052] In this embodiment, a neural network is adopted as the machine learning technique, and therefore the model creation unit 77C creates a neural network model as the rolling control prediction model. That is, the model creation unit 77C creates a neural network model as the rolling control prediction model that links input result data (past annealing, pickling, and rolling result data) and output result data (asymmetric rolling state during cold rolling) in the learning data processed for creating the rolling control prediction model. The neural network model is expressed, for example, by a function formula.

[0053] Specifically, the model creation unit 77C sets hyperparameters used in the machine learning model and performs learning by a neural network model using the hyperparameters. As an optimization calculation of the hyperparameters, the model creation unit 77C first creates a neural network model in which some of the hyperparameters are changed stepwise for the learning data, and selects the hyperparameters that provide the highest prediction accuracy for the validation data.

[0054] As the hyperparameters, the number of hidden layers, the number of neurons in each hidden layer, the dropout rate in each hidden layer (blocking the transmission of neurons with a certain probability), the activation function in each hidden layer, and the number of outputs are usually set, but are not limited to these. In addition, the optimization method of the hyperparameters is not particularly limited, but a grid search that changes parameters stepwise, a random search that randomly selects parameters, or a search by Bayesian optimization can be used.

[0055] Incidentally, the model creation unit 77C is incorporated as a part of the arithmetic device 71, but the configuration is not limited to this. For example, the rolling control prediction models may be created and stored in advance, and may be read out as necessary.

[0056] The model creation unit 77C outputs asymmetric rolling state values ​​for the final cold rolling. The number of configured outputs is not particularly limited. Learning is performed by gradually optimizing the weight coefficients in the machine learning model based on the output results, the past operation results of the steel sheet 1 (annealing conditions, pickling conditions, and rolling results), and the asymmetric rolling results at the first rolling stand of the cold tandem rolling mill 5 at that time (rolling differential load, differential tension at the inlet and outlet, roll reduction position difference, lubricating coolant settings at the upper and lower work rolls, roll surface roughness, etc.).

[0057] Fig. 3 is a diagram showing the relationship between the pickling speed (mpm, meter per minute) and the standard deviation of the differential load in the first rolling stand (tonf). It can be seen that in steel sheets with a total Si and Al content of 3.0 mass% or more, the standard deviation of the differential load in the first rolling stand increases as the pickling speed increases. This is thought to be because scale formed during annealing was not completely removed by pickling and remained, causing localized uneven lubrication or surface roughness of the work roll due to the remaining scale.

[0058] After the weighting coefficients of the machine learning model have been learned, the model creation unit 77C inputs evaluation data (actual operating conditions of the steel plate 1 to be rolled using the rolling control prediction model) into the machine learning model in which the weighting coefficients have been learned, and obtains an estimation result for the evaluation data.

[0059] The result storage unit 77D stores in the memory device 89 the training data, the evaluation data, parameters of the machine learning model, the output results of the machine learning model for the training data, and the output results of the machine learning model for the evaluation data.

[0060] During the cold rolling of the steel sheet 1, the prediction model executing unit 78 predicts the asymmetric rolling state of the steel sheet 1 during cold rolling corresponding to the cold rolling conditions of the steel sheet 1 to be rolled, using the rolling control prediction model created by the prediction model creating unit 77. Then, the prediction model executing unit 78 determines the cold rolling conditions for the steel sheet 1 to be rolled.

[0061] In order to carry out the above processing, the prediction model execution unit 78 includes an information reading unit 78A, a second data preprocessing unit 78B, a rolling state prediction unit 78C, a cold rolling condition determination unit 78D, and a result output unit 78E. Here, when the prediction model execution unit 78 receives a signal from the rolling control device 7 via the input device 88 notifying that cold rolling is being performed, the prediction model execution unit 78 executes the prediction model execution program 75 stored in the ROM 73, thereby functioning as the information reading unit 78A, the rolling state prediction unit 78C, the cold rolling condition determination unit 78D, and the result output unit 78E.

[0062] The information reading unit 78A reads from the storage device 89 the cold rolling conditions of the steel sheet 1 to be rolled, which are set by the process computer and the operator in the operation monitoring device 91.

[0063] The second data pre-processing unit 78B performs data creation processing to be input to the rolling state prediction unit 78C. The processing of the second data pre-processing unit 78B is the same as the processing of the first data pre-processing unit 77B, so detailed description of the processing is omitted. The first data pre-processing unit 77B and the second data pre-processing unit 78B may be made into a subroutine as one processing unit.

[0064] The rolling state prediction unit 78C inputs the input data created by the second data pre-processing unit 78B into a rolling state prediction model, and predicts the asymmetric rolling state of the steel sheet 1 to be rolled.

[0065] The cold rolling condition determination unit 78D performs processing to change the setting of the cold rolling conditions in the explanatory variables until the asymmetric rolling state of the steel sheet 1 satisfies a predetermined condition, and to repeatedly return to the execution of the processing of the information reading unit 78A, the second data pre-processing unit 78B, and the rolling state prediction unit 78C. In one embodiment, it is sufficient to perform processing to change the setting of the cold rolling conditions in the explanatory variables and to repeatedly return to the execution of the processing of the information reading unit 78A, the second data pre-processing unit 78B, and the rolling state prediction unit 78C until the asymmetric rolling state of the steel sheet 1 becomes equal to or less than a preset threshold value.

[0066] The result output unit 78E operates when the asymmetric rolling state of the steel sheet 1 satisfies a predetermined condition, and outputs the determined rolling operation conditions for the steel sheet 1 to be rolled. In one embodiment, the result output unit 78E operates when the asymmetric rolling state of the steel sheet 1 becomes equal to or less than a predetermined threshold value, and outputs the determined rolling operation conditions for the steel sheet 1 to be rolled.

[0067] Next, the process of the prediction model executing unit 78 will be described with reference to FIG.

[0068] Fig. 4 is a flowchart showing the flow of processing by the prediction model executing unit 78. As shown in Fig. 4, when executing a rolling state prediction model, first, in the processing of step S41, the information reading unit 78A of the prediction model executing unit 78 reads from the storage device 89 a machine learning model as a rolling state prediction model corresponding to the required characteristics of the steel sheet 1 to be rolled.

[0069] Next, in the process of step S42, the information reading unit 78A reads the judgment threshold value for the requested asymmetric rolling state stored in the storage device 89 from the host computer via the input device 88. Next, in the process of step S43, the information reading unit 78A reads the cold rolling conditions for the steel sheet 1 to be rolled stored in the storage device 89 from the host computer via the input device 88.

[0070] Next, in the process of step S44, the rolling state prediction unit 78C of the prediction model execution unit 78 uses the machine learning model as the rolling state prediction model loaded in the process of step S41 to determine an asymmetric rolling state value for the corresponding steel plate 1 during cold rolling as input actual data including the annealing, pickling, and cold rolling conditions for the steel plate 1 to be rolled loaded in the process of step S43.

[0071] Next, in the process of step S45, the cold rolling condition determination unit 78D of the prediction model execution unit 78 determines whether or not the asymmetric rolling state value of the steel sheet 1 obtained in the process of step S44 is within the judgment threshold value of the asymmetric rolling state read in the process of step S42. If the convergence of the calculation is not sufficient, an upper limit may be set on the number of iterations of convergence within the range of the calculation time that can actually be executed in the process of step S45. Incidentally, the fact that the asymmetric rolling state value is within the judgment threshold value corresponds to the satisfaction of the predetermined condition in the present invention.

[0072] If it is determined that the asymmetric rolling state value is within the judgment threshold (if the judgment result in step S45 is YES), the prediction model executing unit 78 ends the process. On the other hand, if it is determined that the asymmetric rolling state value is not within the judgment threshold (if the judgment result in step S45 is NO), the prediction model executing unit 78 proceeds to the process in step S46.

[0073] In the process of step S46, the cold rolling condition determination unit 78D changes some of the cold rolling conditions (such as the shape control actuator in the first rolling stand, the lubricating coolant flow rate, and the rolling speed) of the steel sheet 1 to be rolled that were read in the process of step S43, and the process proceeds to step S47. In the process of step S47, the result output unit 78E of the prediction model execution unit 78 transmits information related to the part of the determined cold rolling conditions to the rolling control device 7 via the output device 90.

[0074] When some of the cold rolling conditions have been changed in the process of step S46, in the process of step S47, the cold rolling condition determination unit 78D determines the cold rolling conditions of the steel sheet 1 to be rolled, in which some of the cold rolling conditions, specifically the bender amount and shift amount of the work rolls and intermediate rolls, the roll down position (leveling) in the rolling stand, the amount of lubricating coolant at the entry side of the rolling stand, and the operation amount of the rolling speed have been changed in the process of step S47, as the optimized cold rolling conditions of the steel sheet 1. Then, the cold rolling condition determination unit 78D determines the operation amount of the rolling state based on the cold rolling conditions at that time. The rolling control device 7 changes the cold rolling conditions based on the information on the rolling state transmitted from the result output unit 78E in the cold rolling stage.

[0075] As a method of calculating the change amount of the cold rolling conditions, the cold rolling condition determination unit 78D calculates appropriate cold rolling conditions for the steel sheet 1 to be rolled based on the difference between the asymmetric rolling state value obtained in the process of step S44 and the judgment threshold value read in the process of step S42. Then, the cold rolling condition determination unit 78D compares the calculated cold rolling conditions with the cold rolling conditions for the steel sheet 1 to be rolled read in the process of step S43, and changes the cold rolling conditions in the process of step S47.

[0076] Returning to the process of step S43, the rolling state prediction unit 78C reads the cold rolling conditions of the steel sheet 1 to be rolled, some of which have been changed. In addition, in the process of step S44, the rolling state prediction unit 78C obtains an asymmetric rolling state value of the steel sheet 1 during cold rolling, which corresponds to the cold rolling conditions of the steel sheet 1 to be rolled, some of which have been changed and read in the process of step S43, by a machine learning model as a rolling state prediction model. In addition, in the process of step S45, the cold rolling condition determination unit 78D determines whether the constraint judgment value of the asymmetric rolling state obtained in the process of step S44 is within the judgment threshold value read in the process of step S42. Then, a series of processes of steps S43, S44, S45, S46, and S47 are repeatedly executed until the judgment result becomes YES. This ends the process by the prediction model execution unit 78 (rolling state control determination step).

[0077] As is clear from the above description, in this embodiment, the prediction model creation unit 77 creates a rolling state prediction model by a machine learning technique that links past operational results including annealing, pickling, and rolling of the steel sheet 1 with past asymmetric rolling results corresponding to the operational results. In addition, the prediction model execution unit 78 obtains an asymmetric rolling state value of the steel sheet 1 to be rolled using the created rolling state prediction model during cold rolling of the steel sheet 1. Then, the prediction model execution unit 78 determines the cold rolling conditions of the steel sheet 1 to be rolled such that the obtained asymmetric rolling state value is within a threshold value. As a result, rolling control that satisfies various constraints in rolling operation is performed without depending on the experience or subjectivity of the operator, and productivity can be maintained while preventing troubles such as shape defects and breakage during cold rolling.

[0078] [Modifications] Although the embodiment of the present invention has been described above, the present invention is not limited to this and various modifications and improvements can be made. For example, in this embodiment, the repetition of the rolling state prediction of the steel sheet 1 using the rolling state prediction model and the determination of the cold rolling conditions are performed over the entire length of the coil, but they may be performed only partially. In addition, the cold tandem rolling mill 5 is not limited to a four-high type, and may be a multiple rolling mill such as a two-high (2Hi) or six-high (6Hi) type, and there is no particular limit to the number of rolling stands. In addition, it may be a cluster rolling mill or a Sendzimir rolling mill.

[0079] Furthermore, when the calculation unit 8 calculates an abnormal control amount exceeding the upper or lower limit value for changing the cold rolling conditions or when the control amount cannot be calculated, the rolling control device 7 cannot execute control based on a command from the calculation unit 8. Therefore, when the rolling control device 7 judges that the control amount from the calculation unit 8 is abnormal or when the control amount is not supplied from the calculation unit 8, it is preferable that the rolling control device 7 does not execute this implementation.

[0080] 2, the output device 90 and the operation monitoring device 91 are not connected, but they may be connected to each other so as to be able to communicate with each other. This allows the processing results of the prediction model executing unit 78 (particularly, the rolling state prediction information of the steel sheet 1 being rolled by the rolling state prediction unit 78C, and the changed cold rolling conditions determined by the cold rolling condition determination unit 78D) to be displayed on the operation screen of the operation monitoring device 91. EXAMPLES

[0081] The present invention will now be described with reference to examples.

[0082] Using a cold tandem rolling mill consisting of a total of five rolling stands according to the embodiment shown in FIG. 1, an experiment was conducted in which a base steel sheet containing Si and Al, with a base material thickness of 2.0 mm and a sheet width of 1000 mm, was used as the rolling material and cold rolled to a finishing thickness of 0.300 mm. The stock solution of rolling oil was prepared by adding 1% by mass of an oiliness agent and an antioxidant to a base oil in which vegetable oil was added to a synthetic ester oil, and by adding a nonionic surfactant to the oil at a concentration of 3% by mass. The emulsion rolling oil to be recycled was prepared to have a rolling oil concentration of 3.5% by mass, an average particle size of 5 μm, and a temperature of 55° C.

[0083] As part of the pre-learning process, learning was first carried out using a machine learning model (gradient boosting method using LightGBM (Gradient Boosting Machine)) using learning data (approximately 3,000 pieces of data on past steel plate annealing, pickling, and rolling performance), and past steel plate operating performance and past asymmetric rolling performance at the first rolling stand were linked to create a machine learning model to be used to predict the rolling condition of steel plate.

[0084] In the example of the invention, the input data consisted of annealing conditions (line speed, furnace temperature, gas flow rate, dew point, cooling amount, sheet temperature), pickling conditions (acid concentration, additive concentration, acid temperature, line speed) and cold rolling conditions (roll diameter, roll material, roll initial roughness, tonnage used after roll insertion, deformation resistance of steel sheet, rolling load, rolling tension, emulsion properties, work roll dimensions, crown, roughness information, bender amount, and work roll shift amount) as past operational data of steel sheet. As past non-rolling actual data of steel sheet, the differential load, inlet and outlet differential tension, and roll reduction position difference judgment results in the first rolling stand were learned. The roll gap was adjusted in the cold tandem rolling mill 5, and after the welding point of the steel sheet passed, the rolling control device 7 was turned on, and the asymmetric rolling state during cold rolling was predicted by the created machine learning model. Then, the cold rolling conditions were set by successively changing the cold rolling conditions so that the predicted asymmetric rolling state was equal to or less than a predetermined threshold value. When changing a single cold rolling condition, any of the inlet coolant flow rate, the rolling position difference, and the rolling speed may be changed, but when the rolling position difference was changed, a break occurred with a probability of 1%, although it was within the allowable range, so it is more preferable to change the inlet coolant flow rate or the rolling speed.

[0085] In the comparative example, similarly to the invention example, an experiment was conducted in which a base steel sheet (rolling target) for electrical steel sheet containing Si and Al, with a base material thickness of 2.0 mm and a sheet width of 1000 mm, was cold rolled to a sheet thickness of 0.3 mm. In the comparative example, the rolling performance (rolling sum load, sum tension) without asymmetric components in the first rolling stand was linked to the operational performance data of the past steel sheets as the objective variable, and a neural network model used for predicting the rolling state was created.

[0086] The number of fractures that occurred in the steel sheets after rolling 100 coils in the invention examples and comparative examples is shown in Table 1. As shown in Table 1, in the comparative example, sufficient learning was not performed regarding the asymmetric rolling state in the rolling stand, so when the remaining scale state fluctuated significantly along the coil length, operational constraints were exceeded, and problems such as shrinkage fracture occurred.

[0087] From the above, it was confirmed that it is preferable to appropriately predict the asymmetric rolling state during rolling of a steel sheet using the cold rolling method and cold rolling mill according to the present invention, and to determine the rolling state by successively changing the cold rolling conditions so that the predicted asymmetric rolling state value is equal to or less than a preset threshold value. It was also confirmed that the application of the present invention not only prevents product troubles such as defective shape and sheet breakage during cold rolling, but also greatly contributes to improving the productivity and quality of the rolling process and subsequent processes.

[0088] Although the embodiments of the invention made by the present inventors have been described above, the present invention is not limited by the descriptions and drawings that form part of the disclosure of the present invention according to the present embodiments. In other words, other embodiments, examples, operation techniques, etc. made by those skilled in the art based on the present embodiments are all included in the scope of the present invention.

[0089] [Table 1] [Explanation of symbols]

[0090] 1 steel plate 2 Payoff Reel 3 Annealing furnace 4 Pickling tank 5. Tandem cold rolling mill 6. Coiler 7. Rolling control device 8 Computing Units 9 Operational information measuring device 71 Arithmetic unit 74 Predictive model creation program 75 Prediction model execution program 76 Processing Unit 77 Prediction Model Creation Department 77A Learning data acquisition section 77B First data preprocessing section 77C Model Creation Department 77D Result storage section 78 Prediction Model Execution Unit 78A Information reading unit 78B Second Data Conversion Unit 78C Rolling Condition Prediction Department 78D Cold Rolling Condition Determination Department 78E Result output section 88 Input Devices 89 Storage device 90 Output Device 91 Operation monitoring device

Claims

1. A method for setting cold rolling conditions for cold rolling a steel sheet having a total content of Si and Al of 3.0 mass% or more that has been successively subjected to hot rolling, annealing, and pickling, using a prediction model, comprising: The prediction model is a prediction model trained using cold rolling conditions, including annealing conditions, pickling conditions, and a work roll usage amount in the first rolling stand in cold rolling, as explanatory variables, among past operational results, and using asymmetric components, among rolling results in the first rolling stand, as objective variables; A step of predicting an asymmetric component in the first rolling stand by inputting set cold rolling conditions, including annealing conditions, pickling conditions, and a work roll usage amount in the first rolling stand in cold rolling, for a material to be rolled, into the prediction model; and repeatedly executing prediction of the non-target components while changing a part of the set cold rolling conditions until the predicted asymmetric components are within a judgment threshold value; The cold rolling condition setting method, wherein the non-target components include at least one of a rolling differential load, a differential tension, a rolling position difference of a work roll, and a lubricating coolant flow rate difference between an operator side and a drive side in cold rolling.

2. A cold rolling method comprising the steps of: cold rolling a material to be rolled under set cold rolling conditions changed by the method for setting cold rolling conditions according to claim 1 .

3. A method for producing a cold-rolled steel sheet, comprising the steps of: hot rolling, annealing, and pickling in that order on a material to be rolled; and then cold rolling by the cold rolling method according to claim 2 to produce a cold-rolled steel sheet.

4. A cold rolling condition calculation device that uses a prediction model to set cold rolling conditions when cold rolling a steel plate having a total content of Si and Al of 3.0 mass% or more that has been successively subjected to hot rolling, annealing, and pickling, The prediction model is a prediction model trained using cold rolling conditions, including annealing conditions, pickling conditions, and a work roll usage amount in the first rolling stand in cold rolling, as explanatory variables, among past operational results, and using asymmetric components, among rolling results in the first rolling stand, as objective variables; A prediction means for predicting an asymmetric component in the first rolling stand by inputting set cold rolling conditions, including annealing conditions, pickling conditions, and a usage amount of work rolls in the first rolling stand in cold rolling, for a material to be rolled, into the prediction model; A change means for changing a part of the set cold rolling conditions, The prediction means repeatedly predicts the non-target components while changing a part of the set cold rolling conditions by the change means until the predicted asymmetric components are within a judgment threshold value; The cold rolling condition calculation device, wherein the non-target components include at least one of a rolling differential load, a differential tension, a rolling position difference of a work roll, and a lubricating coolant flow rate difference between an operator side and a drive side in cold rolling.

5. A cold rolling mill comprising the cold rolling condition calculation device according to claim 4.

Images