Plate camber control device

Abstract

The plate crown control device controls the tandem rolling mill based on the set calculation values of the exit side plate crown of each stand, the mechanical plate crown set calculation values of each stand, the set values of the bending force and displacement position of each stand calculated by the set calculation. The processor calculates a first learning current value based on the difference between the mechanical plate crown observation value and the actual calculation value of the mechanical plate crown. The processor updates the first learning value by proportionally distributing the first learning current value and the smoothing gain. In the set calculation after the next rolled material, the processor calculates the set values of the bending force and displacement position of each stand using the corrected mechanical plate crown set calculation value obtained by adding the first learning value to the mechanical plate crown set calculation value.

Description

Technical Field

[0001] This invention relates to a plate crown control device, and more particularly to the learning and control method of a plate crown control device in hot rolling. Background Technology

[0002] The requirements for the thickness accuracy of rolled materials in the width direction have become increasingly stringent each year. In rolling operations that use prediction models for plate crown, improving the accuracy of these prediction models is extremely important.

[0003] It is known that mechanisms such as work roll bending machines and CVC (Continuous Variable Crown) roll displacement are used to control the convexity of the plate in this way.

[0004] Furthermore, to achieve high-precision plate crown control, the mechanisms of each frame need to be correctly set. The plate crown control device calculates the setpoints for these mechanisms. Various plate crown prediction methods and setpoint calculation methods have been proposed and their accuracy improved.

[0005] In addition, the applicant believes that the documents described below are related to this invention.

[0006] Existing technical documents

[0007] Patent documents

[0008] Patent Document 1: Japanese Patent No. 3253013

[0009] Patent Document 2: Japanese Patent Publication No. 3-72364

[0010] Patent Document 3: Japanese Patent Application Publication No. 59-215205

[0011] Non-patent literature

[0012] Non-Patent Literature 1: Shusuke Yanagi, Masanori Ikeda, and Hiroshi Kunii, “Development of a Plate Crown Adjustment Model in Hot Rolling of Aluminum”, Kobe Steel Technical Bulletin, Vol. 58, No. 3 (Dec. 2008), p. 29

[0013] Non-Patent Literature 2: K.S. Hohet, Nathansend, “Roll bending methods of crown control in four-high plate mills”, J. The Iron and Steel Institute, (Nov. 1968) p1088

[0014] Non-patent literature 3: K.S. Hohet, Nathansend, “Flatness control in platerolling”, J. The Iron and Steel Institute, (Oct. 1971) p769 Summary of the Invention

[0015] The problem that the invention aims to solve

[0016] In recent years, plate crown gauges, which measure the thickness distribution of rolled material in the width direction on the mill exit side, and flatness gauges, which measure the shape of the plate, have been introduced. By actively utilizing these technologies and employing dynamic feedback control, high-quality rolled materials are manufactured.

[0017] However, dynamic feedback control based on plate crown gauges and flatness gauges cannot be applied to the very tip of the rolled material. Therefore, the plate crown of the product at the very tip depends on the accuracy of the plate crown control device's settings for each stand at the start of rolling.

[0018] Therefore, by using the learning capabilities of the plate crown gauge and flatness gauge, the setting accuracy of the plate crown control device for each stand can be improved. A widely used learning method is to correct the internal target value in the setting calculation of the next rolled material based on the measured values ​​of the plate crown gauge.

[0019] For example, if the measured convexity of the plate is 10 micrometers larger than the target value of 40 micrometers, the internal target value of 30 micrometers will be calculated as the setting value of the next rolled material after subtracting 10 micrometers from the target value.

[0020] However, in this method, the further back the rack is, the more the calculation is performed relative to a target that differs from the original target. Therefore, there are issues such as increased errors in various prediction models and a limit to the improvement in accuracy.

[0021] In addition, in Patent Document 1, the model error is corrected by allocating the deviation between the actual and predicted convex values ​​measured on the finished side to the predicted convex values ​​of the mechanical plates of each frame.

[0022] However, simply dividing the actual prediction error of the convexity of the finished side equally among all frames or allocating it according to the plate thickness ratio can lead to problems such as deterioration of the flatness of the finished side of each frame, and the bending force or displacement position of each frame reaching mechanical or operational limits, thus failing to achieve the target plate convexity.

[0023] Furthermore, Non-Patent Document 1 discloses a method that sets the actual value of the bending machine after operator correction to positive and corrects the various influence coefficients of the mechanical plate crown model. This method assumes that, as in hot aluminum rolling, there is almost no roll wear, and the roll's thermal expansion (thermal crown) is stable, allowing for prediction based on minute errors. It completely disregards correction for roll shape inference errors.

[0024] However, steel rolling involves high rolling temperatures and the presence of various alloy components, resulting in roll wear typically reaching hundreds of micrometers. Furthermore, the various alloys at different rolling temperatures are rolled in a mixed sequence, leading to complex thermal crown behavior. Therefore, even parameter adjustments via roll temperature measurement are insufficient to adequately reduce prediction errors. Consequently, even when applying the method of Non-Patent Document 1 to steel rolling, the inference error related to roll shape prevents the correction of influence coefficients with sufficient accuracy.

[0025] Furthermore, in the methods disclosed in Patent Documents 2 and 3, the plate convexity error detected by the plate convexity gauge is ultimately distributed according to a distribution method determined solely by the genetic coefficient or shape change coefficient, or learned by distributing it to each rack. That is, in this distribution, the differences in the mechanical convexity variable capability of each rack due to the presence or absence of CVC and mating cross mechanisms are completely ignored. Therefore, the following problem exists: the upper or lower limit of mechanical convexity is reached in racks with the largest margin for flatness between racks but small mechanical convexity variable capability, which makes it highly likely that flatness between racks cannot be achieved, and ultimately the rack will not produce lateral convexity.

[0026] This invention was made to solve the aforementioned problems, and its purpose is to provide a plate convexity control device that can improve the setting accuracy of work roll bending, curve roll displacement, etc., and achieve stable plate passing and improved yield.

[0027] Methods for solving problems

[0028] The first point concerns the plate crown control device used in tandem rolling mills.

[0029] Tandem rolling mills continuously roll the material being rolled through multiple stands.

[0030] The plate crown control device controls the tandem rolling equipment based on the set calculation values ​​of the exit plate crown of each stand, the set calculation values ​​of the mechanical plate crown of each stand, and the set values ​​of the bending force and displacement position of each stand.

[0031] The plate convexity control device has at least one processor and memory.

[0032] The memory stores the measured values ​​of the plate convexity and flatness of the output plate, which are determined at the measurement point along the length direction of the rolled material and measured by the plate convexity gauge and flatness gauge installed on the output side of the final stand of the plurality of stands.

[0033] The memory stores the actual values ​​of the measuring point, including rolling load, bending force, and displacement position, when the measuring point passes through each frame.

[0034] The memory stores a learning table containing the first learning values ​​of each rack.

[0035] The processor is configured to perform the setting calculation, actual value calculation, observed value calculation, first learning value calculation, first learning, and setting calculation value correction.

[0036] The actual value is calculated by calculating the actual calculated value of the mechanical plate convexity of each frame based on each measured value and each actual value.

[0037] The observation value is calculated by using the first learning weighting coefficient of each rack to calculate the mechanical plate convexity observation value of each rack, so that it is consistent with the measured value of the output side plate convexity of the final rack.

[0038] The first learning value is calculated by calculating the first current learning value of each rack based on the difference between the observed value of the mechanical plate convexity of each rack and the actual calculated value of the mechanical plate convexity of each rack.

[0039] The first learning is to update the first learning values ​​of each rack stored in the learning table by proportionally allocating them based on the current first learning value and smoothing gain of each rack.

[0040] The setting calculation value correction is performed by using the corrected mechanical plate convexity setting calculation value of each stand obtained by adding the first learning value of each stand read from the learning table to the setting calculation value of the mechanical plate convexity of each stand in the setting calculation after the next rolled material, and then calculating the setting values ​​of the bending force and displacement position of each stand.

[0041] The second viewpoint, in addition to the first viewpoint, also has the following characteristics.

[0042] The learning table stores the second learning value for each rack.

[0043] The actual calculation of the calculated value includes calculating the actual calculated value of the convexity of the side plate of each frame based on each measured value and each actual value.

[0044] The processor is configured to also perform a second learning value calculation and a second learning process.

[0045] The second learning value is calculated by using the difference between the actual calculated value of the output side panel convexity of the final rack and the set calculated value of the output side panel convexity of the final rack, the second learning weighting coefficient of each rack, and the transfer rate of each rack to calculate the second learning current value of each rack.

[0046] The second learning is to update the second learning values ​​of each rack stored in the learning table by proportionally allocating them based on the current second learning value of each rack and the smoothing gain.

[0047] The setting calculation value correction is as follows: in the setting calculation after the next rolled material, the corrected mechanical plate convexity setting value of each stand is obtained by adding the mechanical plate convexity setting calculation value of each stand to the first learning value and the second learning value of each stand read from the learning table, and then calculating the setting values ​​of bending force and displacement position of each stand.

[0048] The third viewpoint, in addition to the first or second viewpoint, also has the following characteristics.

[0049] The actual calculation of the calculated value includes calculating the actual calculated value of the convexity of the side plate of each frame based on each measured value and each actual value.

[0050] The observation value is calculated by multiplying the deviation between the measured value of the side panel convexity of the final frame and the actual calculated value of the side panel convexity of the final frame by the first learning weighting coefficient and the plate thickness ratio of each frame, and then adding the deviation to the actual calculated value of the side panel convexity of each frame to calculate the observed value of the side panel convexity of each frame.

[0051] The observation value is calculated by using the observation value of the convexity of the inlet side plate of each rack, the observation value of the convexity of the outlet side plate of each rack, the genetic coefficient, the transfer rate, and the ratio of the plate thickness of the inlet side to the outlet side of each rack to calculate the observation value of the convexity of the mechanical plate of each rack.

[0052] The fourth viewpoint, in addition to the first or second viewpoint, also has the following characteristics.

[0053] The observations are calculated by determining the constraints, design variables, and objective function, and by calculating the design variables in a manner that minimizes the objective function.

[0054] The constraint condition is the plate convexity prediction formula for each frame.

[0055] The design variables are the observed convexity of the mechanical plate of each rack and the observed convexity of the exit side plate of each rack other than the final rack.

[0056] The objective function is the sum or square of the absolute values ​​of the deviations between the observed convexity values ​​of the mechanical plates of each rack and the actual calculated convexity values ​​of the mechanical plates of each rack, or the values ​​obtained by multiplying the deviations by the first learning weighting coefficients of each rack.

[0057] The fifth viewpoint, in addition to the third or fourth viewpoint, also has the following characteristics.

[0058] The observed values ​​are calculated by determining the variable range of the mechanical plate convexity of each frame based on the upper and lower limits of the bending force and displacement position of each frame.

[0059] The observed value is calculated by multiplying the variable range of the mechanical plate convexity of each frame by the transfer rate and dividing by the thickness of the side plate of the frame to calculate the variable range of the first plate convexity ratio of each frame.

[0060] The observed values ​​are calculated by determining the variable range of the second plate convexity ratio of each rack based on the flatness limit of each rack.

[0061] The observation value is calculated by multiplying the smaller of the first plate convexity ratio variable range and the second plate convexity ratio variable range of each rack by a correction factor to calculate the first learning weighting coefficient of each rack.

[0062] The sixth point, in addition to the fourth point, has the following characteristics.

[0063] The design variables include the plate convexity ratio genetic coefficient or transfer rate of each rack.

[0064] The constraints include inequalities such as the rear rack having a larger plate convexity ratio genetic coefficient compared to the front rack, or the rear rack having a smaller transfer rate compared to the front rack.

[0065] The seventh point, in addition to the characteristics of the second point, also has the following features.

[0066] The second learning weighting coefficient for each rack is the difference between the actual calculated value of the mechanical plate convexity of each rack and the set calculated value of the mechanical plate convexity of each rack, divided by the thickness of the side plate of each rack.

[0067] The effects of the invention

[0068] According to the present invention, the observed mechanical plate crown value can be inferred based on the measured value determined by a plate crown gauge disposed on the exit side of a tandem rolling mill, and the difference between the observed mechanical plate crown value and the actual calculated mechanical plate crown value can be learned. Using the learning result, the predicted mechanical plate crown value (set calculation value) can be corrected in the setting calculation after the next rolled material. Accordingly, the problem of flatness deterioration on the exit side of each stand, and the problem of bending force and displacement position of each stand reaching mechanical and operational limits and failing to achieve the target plate crown, can be suppressed. As a result, the setting accuracy of work roll bending, curve roll displacement, etc., can be improved, and the plate crown on the exit side of the final stand can be made closer to the target value, achieving stable plate passage and improving yield. Attached Figure Description

[0069] Figure 1 This is a diagram illustrating an example of the configuration of a rolling mill according to an embodiment of the present invention.

[0070] Figure 2 This is a diagram illustrating the definition of plate convexity and the structure of a plate convexity meter.

[0071] Figure 3 This is a diagram used to illustrate the working roller displacement mechanism.

[0072] Figure 4 This is a diagram used to illustrate the bending mechanism of the work roll.

[0073] Figure 5 This is a block diagram illustrating the general functions of the plate convexity control device according to an embodiment of the present invention.

[0074] Figure 6 This is a diagram representing an example of a design variable.

[0075] Figure 7 This is a diagram representing an example of a design variable.

[0076] Figure 8 It is a graph showing the relationship between aspect ratio (plate width / plate thickness) and the heritability coefficient of convexity ratio.

[0077] Figure 9 This is a block diagram illustrating an example of the hardware configuration of a plate convexity control device according to an embodiment of the present invention. Detailed Implementation

[0078] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. Furthermore, in the various drawings, common elements are labeled with the same reference numerals, and repeated descriptions are omitted.

[0079] Implementation method.

[0080] 1. Overview of Rolling Equipment

[0081] Figure 1 This is a diagram illustrating an example configuration of a rolling mill according to an embodiment of the present invention. The rolling mill 1 is a series rolling mill that continuously rolls the workpiece 2 using N stands (N being a natural number of 2 or more). As an example, the rolling mill 1 is a finishing mill with 7 stands (N=7), numbered F1 to F7. The rolling mill 1 is configured in a hot rolling production line.

[0082] An X-ray-based plate thickness gauge 3, an X-ray-based plate convexity gauge 4 (profile gauge), and an optical flatness gauge 5 are installed on the output side of the final rack (Nth rack). In addition, these measuring instruments are relatively large and expensive, so in most cases only one is installed on the output side of the final rack.

[0083] However, regarding the inlet and outlet plate thicknesses of each stand (the i-th stand (i = 1, ..., N)), even when measurement by a plate thickness gauge is not possible, relatively accurate estimations can be made using instrument-based methods based on roll gaps, rolling loads, and mill rigidity in each stand, as well as mass flow methods utilizing the assumption of identical volumetric velocity between stands. By using these methods of plate thickness control, the plate thickness is controlled to match a predetermined target value; therefore, these conditions are treated as known in this invention.

[0084] Figure 2 This diagram illustrates the definition of plate convexity and the configuration of the plate convexity meter 4. (See diagram below.) Figure 2 As shown, when the central plate thickness in the width direction of the rolled material is set as h... c The plate thickness at a specified distance (40mm in this example) from the edge is defined as h. e At that time, the plate convexity was defined as its deviation.

[0085] Figure 1 Each stand of the rolling mill is a roll displacement rolling mill with paired upper and lower work rolls 6, the upper and lower work rolls 6 having a prescribed crown curve for cross-sectional shape control. In addition, each rolling mill is equipped with a work roll displacement mechanism 7 and a work roll bending mechanism 8 for axially moving the upper and lower work rolls 6.

[0086] The plate crown control device 10 is connected to the plate thickness gauge 3, plate crown gauge 4, flatness gauge 5, work roll displacement mechanism 7, work roll bending mechanism 8, etc. The plate crown control device 10 determines the displacement and bending force of each stand in order to obtain the specified cross-sectional shape of the rolled material.

[0087] Figure 3This diagram illustrates the work roll displacement mechanism 7. The work rolls 6 are ground to have a concave-convex shape represented by a curve of degree 3 or higher and are arranged in opposite vertical directions. The work roll displacement mechanism 7 changes the convexity of the rolled material by moving the upper and lower work rolls 6 axially in opposite directions by the same distance, thereby altering the distribution of the roll gap. Figure 3 The diagram shows the work roll displacement states corresponding to three displacement positions. This work roll 6 is called a CVC (Continuous Variable Crown) roll, but here it is referred to as a curve roll. Furthermore, the axial movement distance of the work roll 6 is called the displacement amount. The average diameter of the upper and lower work rolls 6 at a certain axial position is called the average roll diameter. The difference between the average roll diameter at a certain displacement position and between the roll center and the roll end is called the equivalent roll crown.

[0088] Figure 4 This diagram illustrates the work roll bending mechanism 8. The work roll bending mechanism 8 is a mechanism that, by applying a bending force in the vertical direction to the shaft end of the work roll 6, causes the work roll 6 to flex, thereby changing the contact stress distribution between the work roll 6 and the workpiece 2, and thus changing the convexity of the workpiece 2. The bending force is generated by a hydraulic cylinder disposed between the upper and lower work rolls 6, and acts on the work roll 6 via a shaft box mounted at the shaft end of the work roll and its bearings. The hydraulic cylinders are located at both ends of the work roll 6, and the sum of the forces generated by both is called the bending force. Figure 4 As shown, increasing the bending force decreases the plate crown, while decreasing the bending force increases the plate crown. Generally, increasing the rolling load increases roll deflection and plate crown. Therefore, correspondingly, the bending force needs to be appropriately increased to increase the rolling load.

[0089] 2. Plate crown control device

[0090] Figure 5 This is a block diagram illustrating the general functions of the plate crown control device 10. The plate crown control device 10 includes a setting calculation unit 51, an actual calculation value calculation unit 52, an observation value calculation unit 53, a first learning value calculation unit 54, a first learning unit 55, a second learning value calculation unit 56, and a second learning unit 57.

[0091] 2-1. Setting up the calculation unit

[0092] The setting calculation unit 51 performs setting calculations before the rolled material 2 is bitten into the finishing mill. It calculates the displacement position settings and bending force settings of the upper and lower work rolls 6 of each stand, ensuring the desired plate crown is achieved on the finishing exit side and the rolled material is flat between stands. Generally, setting calculations refer to numerically calculating the rolling phenomena using a numerical model and determining the rolling schedule. Specifically, this means calculating the target dimensions and temperatures for each process from when the billet (base material) is removed from the furnace to when the hot-rolled coil (product) is wound, and calculating the initial settings (roll gap, initial displacement position) of each actuator used to achieve these.

[0093] The plate crown control device 10 operates the work roll displacement mechanism 7 to move the work roll 6 to the displacement position set value. Furthermore, after the workpiece 2 is bitten in, the work roll bending mechanism 8 is operated to make the bending force match the bending force set value. Generally, from a mechanical protection perspective, the displacement positions of each stand are set to be fixed during rolling.

[0094] Here, the calculation method for the set value in the setting calculation is shown. It is known that the change of the plate convexity on the entry and exit side of each rack is represented by the following formula (1). The plate convexity prediction formula on the entry and exit side of each rack shown in formula (1) is disclosed, for example, by Patent Document 2.

[0095]

Formula 1

[0096]

[0097] i represents the stand number, for example, 1 to 7 in the case of a 7-stand rolling mill. Here, H... i h represents the plate thickness on the inlet side of each rack. i C represents the thickness of the plate on the side facing outwards. Hi C represents the convexity of the rolled material 2 on the infeed side (infeed side convexity). hi C represents the convexity of the plate on the outward side (outward side plate convexity). Mi This refers to the convexity of the mechanical plate.

[0098] Mechanical plate crown, also known as roll gap crown, is the plate crown determined by the elastic deformation of the rolling mill. Mechanical plate crown is expressed as the difference between the plate thickness at the center position and the plate thickness at the end of the plate in the width direction, i.e., the plate crown, and represents the gap distribution between the upper and lower work rolls 6, i.e., the plate thickness distribution, assuming that the rolling load acting between the rolled material 2 and the work rolls 6 is uniformly distributed in the width direction.

[0099] The crown of the rolling mill is theoretically calculated based on rolling conditions including the roll size and crown, the width of the rolled material 2, the thickness of the inlet side, the thickness of the outlet side, the rolling load, the displacement position, and the bending force, thereby determining the elastic deformation of the rolling mill. For example, Patent Document 2 discloses an approximate formula obtained through analysis. Patent Document 3 discloses a solution that divides the roll axis direction into multiple small regions.

[0100] Based on these, the convexity of the mechanical plate is expressed by the following formula (2). In addition, the diameter, length, elastic coefficient, Poisson's ratio, etc. of the roller remain unchanged, so they are omitted from the description.

[0101]

Formula 2

[0102] C Mi =f(H i h i P i F i C wi C bi (2)

[0103] Here, P is the rolling load, F is the bending force, H is the thickness of the inlet plate, and h is the thickness of the outlet plate. C w C is the equivalent roll crown of work roll 6. b Let be the equivalent roll crown of the support roll. Besides displacement, the equivalent roll crown is also affected by thermal expansion and wear, and is represented by a function as follows.

[0104]

Formula 3

[0105] C wi =f w (L wrsi C thrmi c weari (3A)

[0106] C bi =f b (C bthrmi c bweari (3B)

[0107] Here, L wrsi Indicates the displacement position of the curve roller. C thrmi This indicates the change in crown caused by the thermal expansion of the work roll. (C) weari This indicates the change in crown caused by wear of the work roll. Additionally, C bthrmi This indicates the change in crown caused by the thermal expansion of the support roller. (C) bweari This indicates the change in crown caused by wear of the support roller.

[0108] (Heritage coefficient of plate convexity ratio)

[0109] In addition, η in equation (1) i The coefficient representing the convexity ratio of the plate is heritable.

[0110] Here, the plate crown ratio inheritance coefficient refers to the plate crown ratio (entry-side plate crown ratio) on the infeasibility side of each stand, where all other rolling conditions are kept the same. hi / H i When the convexity ratio of the plate on the output side of the frame changes (output side plate convexity ratio) (C) hi / h i The influence coefficient generated.

[0111] It is known that the heritability coefficient of the plate convexity ratio is expressed as a function value with geometric conditions such as the plate width / thickness ratio and roll diameter as parameters (e.g., Patent Document 2).

[0112] Thus, the reason why the exit side plate crown ratio is affected by the inlet side plate crown ratio is due to the following mechanism: due to the change in the inlet side plate crown, a slight difference is generated in the width distribution of the plate during rolling in this stand. As a result, the width distribution of tension in the roll gap changes, the contact stress distribution between the roll and the rolled material changes, and the elastic deformation of the roll changes accordingly.

[0113] (Transfer rate)

[0114] In addition, ζ in equation (1) represents the transfer rate.

[0115] Here, the transfer rate refers to the value obtained by dividing the machine plate crown of each stand by the thickness of the side plate of that stand, while keeping all other rolling conditions the same (C). Mi / h i When the convexity ratio of the plate on the output side of the frame changes (output side plate convexity ratio) (C) hi / h i The influence coefficient generated.

[0116] It is known that when the convexity ratio of the inlet side plate and the convexity ratio of the outlet side plate do not change significantly, there is a relationship between the heritability coefficient of the plate convexity ratio and the transfer rate as follows (e.g., Patent Document 2).

[0117]

Formula 4

[0118] ζ i =1-η i (4)

[0119] (flatness limit)

[0120] Furthermore, on the exit side of each stand, due to variations in plate crown, differences in elongation strain arise depending on the position in the width direction. If the plate crown ratio (the value obtained by dividing plate crown by plate thickness) on the stand entry side is the same as the plate crown ratio on the stand exit side, no difference in elongation strain occurs, and the rolled material 2 remains flat. Conversely, when the difference between the plate crown ratio on the stand entry side and the plate crown ratio on the stand exit side exceeds the allowable value (called the flatness limit), the rolled material is warped, resulting in edge waviness or poor flatness extending into the center.

[0121] The flatness limit is determined experimentally, for example, by the following formula (5) (Non-Patent Literature 3).

[0122]

Formula 5

[0123]

[0124] Here, B i For each stand, the side plate width is given. In hot rolling of thin plates, even if the value that converts the product width to a thermal dimension (which takes into account thermal expansion) is used, the result is almost unchanged.

[0125] The plate crown ratio on the inlet side of the first stand of the finishing mill is determined by the conditions on the roughing mill side and differs from the target plate crown ratio on the outlet side of the final stand of the finishing mill. Therefore, it is necessary to vary the plate crown ratio across several stands of the finishing mill to ensure that the plate crown on the outlet side of the final stand matches the target plate crown. Thus, according to equation (5), the plate crown ratio is primarily varied in the front-side stands where the flatness limit is larger, while the variation in the plate crown ratio is smaller in the rear-side stands.

[0126] (Control based on preset calculations)

[0127] In the setup calculations, the projection convexity C of each rack's side plate is determined based on this perspective. hi Substitute it into equation (1) and calculate to determine the mechanical plate convexity C of each frame. Mi Then, substitute it into equations (2) and (3) to solve in reverse, thereby determining the bending force and displacement position.

[0128] In addition, there is a priority order in the operation of bending force and displacement position. First, the displacement position is operated with the bending force set to an arbitrary value. When the displacement position reaches the mechanical or operational mobility limit, the bending force is operated while fixing the displacement position at the mobility limit.

[0129] However, when a large amount of rolled material 2 is continuously rolled, the thermal expansion and wear of the work roll 6 develop, making it difficult to maintain the accuracy of the set calculation.

[0130] Therefore, in this embodiment, the plate crown control device 10, after the start of finishing rolling, collects the measured values ​​of the plate crown and flatness of the final stand at a measurement point determined along the length direction of the rolled material 2, as well as the actual values ​​including rolling load, bending force, and displacement position as the measurement point passes through each stand. Based on these measured values ​​and actual values, the plate crown control device 10 learns the correction amount (learning value) for the set calculation, and corrects the set calculation according to the learned correction value in the set calculation of the next rolled material. Through this learning, the accuracy of the set calculation after the next rolled material can be improved, and the plate crown accuracy of the rolled material can be improved.

[0131] The learning method of the present invention is shown below.

[0132] 2-2. Actual Calculation Value Calculation Section

[0133] First, the actual calculation unit 52 calculates the actual calculated value of the mechanical plate convexity of each frame.

[0134] The actual calculated value refers to the value calculated by assigning the actual values ​​collected during rolling as parameters to the same model formula used in the set calculation. The suffix ACAL is added below for distinction. The actual calculated value of the machine plate crown is calculated according to equations (2) and (3) as follows.

[0135]

Formula 6

[0136]

[0137]

[0138]

[0139] Here, P i ACT This represents the measured value of the rolling load. F i ACT This is the measured value of the bending force. L wrsi ACT This represents the actual displacement position. C thrmi ACAL C represents the change in roll crown caused by the thermal expansion of the work roll 6. weari ACAL C represents the change in roll crown due to wear of the work roll 6. bthrmi ACAL C represents the change in roll crown caused by the thermal expansion of the support roll. bweari ACAL This refers to the change in roll crown caused by wear of the support roll.

[0140] When these values ​​are substituted into equation (1) and calculated sequentially from the upstream frame to the downstream frame, the actual calculated value of the convexity of the side plate of each frame can be obtained.

[0141]

Formula 7

[0142]

[0143] Thus, based on the aforementioned measured values ​​and actual values, the actual calculation unit 52 calculates the actual calculated value C of the mechanical plate convexity of each frame. Mi ACAL And the actual calculated value C of the convexity of the side plate of each frame. hi ACAL .

[0144] 2-3. Observation Calculation Department

[0145] The observation calculation unit 53 uses the first learning weighting coefficient for each rack to calculate the observed mechanical plate convexity of each rack, so that it is consistent with the final measured value of the exit side plate convexity of the rack. The observed value refers to the convexity C of the first rack's inlet side plate. H1 SUP Starting with the pre-rolling calculation value (adjustment value), and ensuring that the exit side plate convexity of the final stand matches the measured value, the exit side plate convexity C of each stand other than the final stand is adjusted. Mi Side plate convexity C hi The values ​​derived through inference are distinguished by adding the suffix OBS below.

[0146] At observation value C Mi OBS There are multiple ideas involved in the inference, and two calculation methods are shown here.

[0147] 2-3-1. The First Method for Calculating Observations

[0148] The first calculation method for the observed values ​​(corresponding to technical solution 3) is based on the following concept: In the calculation, considering the above-mentioned constraints, the convexity of the side plate of each frame is calculated in a way that the flatness between each frame is within the flatness limit. Therefore, the distribution of the convexity of the side plate of each frame should be maintained as much as possible.

[0149] Therefore, the plate convexity error measured on the final rack exit side is allocated to each rack using the plate thickness ratio and a learning weighting coefficient. As shown in Equation (10) below, the observation calculation unit 53 calculates the measured plate convexity C on the final rack exit side. hN ACT The actual calculated value C of the side panel convexity of the final frame hN ACAL The deviation multiplied by the first learning weighting coefficient w of each racki and plate thickness ratio (h) i / H i Add the actual calculated value C of the convexity of the side plate of each frame. hi ACAL The observed value C of the side plate convexity of each frame was calculated. hi OBS .

[0150]

Formula 8

[0151]

[0152]

[0153] Additionally, the first learning weighting coefficient w i It is a learning weighting coefficient used to adjust the distribution ratio of the plate convexity ratio error measured on the final rack exit side to the learning values ​​of each rack. Regarding w i The decision-making process will be described later.

[0154] When equation (1) is modified, and the observed value C of the inlet side plate convexity of each frame is substituted into equation (10), equation (10) is transformed. Hi OBS (=C hi-1 OBS ), the observed value C of the convexity of the side plate of each frame hi OBS and the observed values ​​C of the mechanical plate convexity of each frame. Mi OBS When, we can obtain the following formula (12).

[0155]

Formula 9

[0156]

[0157] As shown in equation (12), the observation calculation unit 53 uses the observation value C of the convexity of the inlet plate of each frame. Hi OBS And the observed value C of the convexity of the side plate of each frame. hi OBS Genetic coefficient ηi, transfer rate ζi, and the ratio of the thickness of the plate on the inlet side to the thickness of the plate on the outlet side of each rack (h) i / H i ), calculate the observed value C of the mechanical plate convexity of each frame. Mi OBS .

[0158] 2-3-2. The Second Method for Calculating Observations

[0159] The second method for calculating the observed values ​​(corresponding to technical solutions 4 and 6) is based on the following concept: In the calculation, the optimal combination of the side plate convexity and the mechanical plate convexity of each frame is calculated and determined based on the above-mentioned constraints, so that their distribution should be maintained as much as possible.

[0160] In this calculation method, the observation calculation unit 53 calculates the mechanical plate convexity observation value C for each frame. Mi OBS And the observed values ​​C of the side panel convexity of each frame other than the final frame. hi OBS When determining constraints, design variables, and objective function, calculate the design variables in a manner that minimizes the objective function.

[0161] The constraint is the plate convexity prediction formula for the entry and exit sides of each rack (i = 1, ..., N) as shown in equation (1). Substitute the observed value C into the plate convexity between each rack in equation (1). hi OBS Substitute the observed value C into the convexity of the mechanical plate of each frame. Mi OBS Equation (13A) is the constraint condition for the first rack case, Equation (13B) is the constraint condition for a rack that is neither the first rack nor the final rack case, and Equation (13C) is the constraint condition for the final rack case. Figure 6 As shown, the observed values ​​C of the mechanical plate convexity of each frame are... Mi OBS And the observed values ​​C of the side plate convexity of each frame other than the final frame. hi OBS As a design variable, if the number of racks is set to N, then the number of design variables is (N×2-1). Additionally, the convexity of the inlet side plate of the first rack is calculated using a set value C. H1 SUP The final side panel convexity of the frame was determined using the measured value C. hN ACT .

[0162]

Formula 10

[0163]

[0164]

[0165]

[0166] In addition to the convexity of each rack plate and the convexity of the mechanical plate, such as Figure 7 As shown, the design variables also include the convexity ratio inheritance coefficient η of each rack. i Or the transfer rate ζ iIn this case, one of the inequalities in equation (13D) or equation (13E) is added to the constraint conditions (corresponding to technical solution 6). Equation (13D) represents the heritability coefficient η of the plate convexity ratio of the rear rack compared to the front rack. i Larger. Equation (13E) represents the transfer rate ζ of the back rack compared to the front rack. i Smaller.

[0167]

Formula 11

[0168] η i <η i+1 (i = 1, ..., N-1) (13D)

[0169] ζ i+1 <ζ i (i = 1, ..., N-1) (13E)

[0170] The reason is that, for example Figure 8 As shown, the genetic coefficient η of the convexity ratio is known. i It is a function value with the horizontal axis as the aspect ratio (plate width / plate thickness), which increases as the plate thickness of the subsequent frame becomes thinner.

[0171] The objective function Φ is one of equations (14) or (15) shown below. That is, the observed values ​​C of the mechanical plate convexity of each frame. Mi OBS The actual calculated value C of the mechanical plate convexity of each frame Mi ACAL The deviation, or the deviation multiplied by the first learning weighting coefficient w for each rack. i The sum of the absolute values ​​or the sum of squares of the obtained values ​​is used as the objective function.

[0172]

Formula 12

[0173]

[0174]

[0175] Additionally, the first learning weighting coefficient w i It is a learning weighting coefficient used to adjust the distribution ratio of the plate convexity ratio error measured on the final rack exit side to the learning value of each rack. It can be set to any value or calculated by the determination method described later.

[0176] The observation calculation unit 53 calculates each design variable in a manner that minimizes the objective function Φ. From this, the observed mechanical plate convexity C of each frame is obtained. Mi OBS Equations (13A) to (13C) represent the observed value C. Mi OBSThe linear form allows for the use of linear planning methods such as the Simplex method in the calculation when the objective function is Equation (11). Furthermore, when the objective function is Equation (12), a quadratic planning method can be used. Equation (11) allows for lower computational load and faster calculation. On the other hand, Equation (12) can sometimes be used to calculate the mechanical plate convexity observation value C. Mi OBS A more suitable allocation of each rack.

[0177] 2-4. First Learning Value Calculation Unit and First Learning Unit

[0178] The mechanical plate convexity observation value C of each frame is obtained using either the first calculation method (2-3-1) or the second calculation method (2-3-2) based on the above observations. Mi OBS The first learning value calculation unit 54 is based on the mechanical plate convexity observation value C of each rack. Mi OBS The actual calculated value C of the mechanical plate convexity of each frame Mi ACAL The difference is used to calculate the first learning current value Z for each rack. MOi CUR (Current value: CUR).

[0179]

Formula 13

[0180]

[0181] The first learning department 55 stores the first learning value Z of each rack in the learning table (steel type × plate thickness × plate width). MOi OLD (OLD value) read out, using the first learned current value Z of each rack. MOi CUR The update is performed as shown in equation (17). The new first learning value Z MOi NEW The (NEW value) is written back to the learning table.

[0182]

Formula 14

[0183]

[0184] Here, α MOi It is a smoothing gain that can be adjusted for each rack (0≤α). MOi ≤1).

[0185] First learning value Z MOi The main components include corrections for errors in inferring the crown of the mechanical plate due to rolling loads and bending forces, as well as errors in inferring the thermal expansion and wear of the rolls.

[0186] 2-5. First, learn the method for determining the weighting coefficients.

[0187] The mechanical plate convexity observation value C of each frame is obtained by using either the first calculation method (2-3-1) or the second calculation method (2-3-2) of the above observation values. Mi OBS At that time, the first learning weighting coefficient w is used to adjust the allocation of the first learning value of the plate convexity ratio error measured in the final rack to each rack. i There are different decision-making methods, such as the one described below.

[0188] The first is the first learning weighting coefficient w for all racks (i = 1, ..., N). i Set to the same value (e.g., w) i =1.0) is a determination method. According to this determination method, the distribution of the measured plate convexity ratio error in the final frame to the learning value of each frame is nearly equal. Therefore, it has the advantage that the changes in bending force and displacement position based on learning are easy for operators to understand, and thus intervention operations are easy to perform.

[0189] Secondly, the larger the variable range of the side panel convexity ratio of each rack, the more the first learning weighting coefficient w... i The larger the value, the better (corresponding to technical solution 5). Due to mechanical specifications and operational reasons, upper and lower limits are set for the bending force and displacement position of each frame. When these upper and lower limits are substituted into equations (2) and (3), the upper limit value C of the mechanical plate convexity can be obtained. Mi MAX and the lower limit C Mi MIN When these upper and lower limits are substituted into each equation (1) and the difference between the two (the variable range of the machine plate convexity) is taken, multiplied by the transfer rate and divided by the thickness of the side plate of the frame, the variable range of the first plate convexity ratio Δγ can be obtained. hi CHG1 Equation (18).

[0190]

Formula 15

[0191]

[0192] On the other hand, according to equation (5), the maximum and minimum values ​​of the convexity ratio change within the flatness limit can be obtained. Therefore, when the difference between the two is taken, the variable range Δγ of the convexity ratio of the second plate can be obtained. hi CHG2 Equation (19).

[0193]

Formula 16

[0194]

[0195] By taking the smaller of equations (18) and (19) as the variable range of the plate convexity ratio, the variable range of the convexity ratio of the exit side plate for each frame can be obtained. The larger the variable range of the convexity ratio of the exit side plate for each frame, the more the weighting coefficient w will be learned. i The larger the value, the better. It allows for the calculation of the variable range of plate convexity ratios under typical operating conditions, and uses these as a reference to determine the first learning weighting coefficient w. i .

[0196] Alternatively, if the first learning weighting coefficient w is determined by calculation as shown in equation (20), i This allows for the acquisition of a more appropriate value. The observation calculation unit 53 calculates the variable range Δγ of the first plate convexity ratio. hi CHG1 and the variable range of the second plate convexity ratio Δγ hi CHG2 The smaller of the two is multiplied by the correction factor k. i To calculate the first learning weighting coefficient w for each rack i .

[0197]

Formula 17

[0198]

[0199] Here, according to the different correction coefficients k of the frame i and c i Let k be any constant; typically, k i The value is the same across all racks, c i It is zero, but sometimes it is fine-tuned depending on the status of the operation.

[0200] In addition, as shown in equations (10) and (11), the first learning weighting coefficient w i Through w TOTAL Used in a standardized manner.

[0201] According to this determination method, the measured plate convexity ratio error in the final frame is allocated to the learned value of each frame based on the variable range of the plate convexity ratio of each frame. Therefore, it has the advantage of not easily causing problems such as the variable range of plate convexity ratio of a certain frame becoming without margin, the displacement position and bending force of the frame reaching the upper and lower limits, or the flatness of the frame's output side deteriorating.

[0202] 2-6. Second Learning Value Calculation Unit and Second Learning Unit

[0203] Next, the second learning method (corresponding to technical solution 2) will be described. In the second learning method, the plate thickness ratio (h) is used. i / hN ) and the second learning weighting coefficient u i The final calculated value C of the side plate convexity of the frame is... hN ACAL With the set calculated value C hN SUP The difference is allocated as the convexity error of the side panel of each frame. The amount of mechanical board convexity correction required to correct the convexity error of the side panel of each frame is expressed by the following formula (21A) using the transfer rate, which is the second learning current value (Current value: CUR) relative to the mechanical board convexity setting value.

[0204]

Formula 18

[0205]

[0206]

[0207] As shown in equation (21A), the second learning value calculation unit 56 uses the actual calculated value C of the final frame's side plate convexity. hN ACAL The calculated value C of the final frame's side panel convexity setting. hN SUP The difference between them, and the second learning weighting coefficient u for each rack. i Transfer rate ζ of each rack i Calculate the second learning current value Z for each rack. MAi CUR .

[0208] In addition, the second learning weighting coefficient u i This is the actual calculated value C of the side panel convexity used to adjust the final frame during the second learning process. hN ACAL With the set calculated value C hN SUP The learning weighting coefficient is the ratio of the difference in learning values ​​allocated to each rack. Regarding u... i The decision-making process will be described later.

[0209] The second learning department 57 stores the second learning value Z of each rack in the learning table (steel type × plate thickness × plate width). MAi OLD (OLD value) read out, using the second learned current value Z for each rack. MAi CUR The update is performed as shown in equation (22). The new second learning value Z is... MAi NEW The (NEW value) is written back to the learning table.

[0210]

Formula 19

[0211]

[0212] Here, α Mai It is a smoothing gain that can be adjusted for each rack (0≤α). MAi ≤1).

[0213] Second learning value Z Mai This includes differences in displacement and bending force between the initial calculation and the learning phase, primarily including corrections due to operator intervention. Operators are more likely to intervene manually after visually confirming the status of each rack; therefore, the second learning value Z is... MAi It is inappropriate to allocate the learning value to other racks as well; it is more appropriate to allocate it primarily to the rack in question.

[0214] Therefore, the second learning weighting coefficient u for each rack i The preferred value is the actual calculated value C of the mechanical plate convexity of each frame. Mi ACAL With the set calculated value C Mi SUP The difference is determined according to the following formula (23).

[0215]

Formula 20

[0216]

[0217] That is, the second learning weighting coefficient u of the i-th rack i The actual calculated value C of the mechanical plate convexity of the i-th frame. Mi ACAL The calculated value C of the mechanical plate convexity setting of the i-th frame Mi SUP The difference divided by the thickness h of the side plate of the i-th frame i The obtained value (corresponding to technical solution 7) is then used. Based on this, u i The value is largely allocated to the portion where the difference between the actual calculated value and the set calculated value is large.

[0218] 2-7. Setting Calculation Department (Setting calculation for the next rolled material)

[0219] Next, the correction in the setting calculation after the next rolled material will be explained. As shown in equation (24), the setting calculation unit 51 performs the following setting calculation value correction 51a in the setting calculation after the next rolled material: calculates the mechanical plate crown setting calculation value C for each stand. Mi SUP Add the first learning value Z of each rack read from the learning table MOi NEW and the second learning value Z for each rack MAiNEW The calculated values ​​for the corrected mechanical plate convexity settings of each frame are obtained.

[0220]

Formula 21

[0221] C Mi =f(H i h i P i F i C wi C bi )+Z MOi +Z MAi (twenty four)

[0222] Therefore, by correcting the calculated value of the mechanical plate convexity setting of each frame based on the learning results, and substituting it into equations (2) and (3) to solve in reverse, the setting values ​​of bending force and displacement position are corrected, so that the convexity of the precision rolled side plate can be close to the target value.

[0223] 3. Effects

[0224] As explained above, the plate crown control device 10 according to this embodiment can infer the observed mechanical plate crown value based on the measured value measured by the plate crown gauge 4 disposed on the final stand exit side, according to the crown variable range, the quadratic planning method, etc., and learn the difference between the observed mechanical plate crown value and the actual calculated mechanical plate crown value. Using the learning result (first learning value), the predicted mechanical plate crown value (set calculation value) can be corrected in the setting calculation after the next rolled material.

[0225] Therefore, it is possible to suppress the problem of flatness deterioration on the exit side of each frame, and the problem of bending force and displacement position of each frame reaching mechanical and operational limits, thus failing to achieve the target plate convexity. As a result, the setting accuracy of work roll bending, curve roll displacement, etc., can be improved, so that the final exit side plate convexity of the frame is close to the target value, achieving stable plate passing and improving the yield.

[0226] Furthermore, according to the plate convexity control device 10 of this embodiment, as shown in equation (20), the plate convexity model of each rack can be learned in a way that ensures a margin by taking into account both the variable range of mechanical plate convexity of each rack and the shape insensitivity area. Therefore, it is possible to obtain the advantages of making it difficult to reach the upper or lower limit of mechanical convexity, and increasing the possibility of making the flatness between racks and the final rack exit convexity consistent with the target value.

[0227] Furthermore, the plate convexity control device 10 according to this embodiment can learn (a second learning value) the difference between the actual calculated value of the convexity of the final frame's exit side plate and the set calculated value of the convexity of the exit side plate. Accordingly, it can also learn the correction results of the operator on the bending force and displacement position.

[0228] 4. Variations

[0229] However, in the above implementation, with minimal manual intervention from the operator, the second learning value Z, which is the difference between the actual calculated value and the set calculated value, can be omitted. MAi And its calculation (corresponding to technical solution 1). Alternatively, it can also omit the first learning value Z. MOi Its calculation structure (corresponding to technical solution 8).

[0230] Furthermore, when the plate convexity gauge is installed on the exit side of the frame other than the final frame, the present invention can be applied by replacing the final frame with the plate thickness gauge installation frame and replacing each frame with a frame upstream of the plate thickness gauge.

[0231] Furthermore, when using a plate crown control device 10 other than a pairing cross device, a VC roll device, an intermediate roll displacement device for a 6-stage rolling mill, or a work roll displacement mechanism 7, the present invention can be applied by replacing the displacement position with the operating amount of the actuator of each device.

[0232] Furthermore, in the case of a reversing mill that performs multiple (passes) of rolling while changing the rolling direction through a single stand, the present invention can be applied by replacing the stand with passes.

[0233] 5. Hardware Configuration Example

[0234] Figure 9 This is a conceptual diagram illustrating an example of the hardware configuration of the processing circuit of the board convexity control device 10 in the above embodiment. The function of the board convexity control device 10 is implemented by the processing circuit. In one embodiment, the processing circuit includes at least one processor 91 and at least one memory 92. In another embodiment, the processing circuit includes at least one dedicated hardware 93.

[0235] When the processing circuit includes a processor 91 and a memory 92, each function is implemented through software, firmware, or a combination of software and firmware. At least one of the software and firmware is described as a program. At least one of the software and firmware is stored in the memory 92. The processor 91 executes the program and various data stored in the memory 92, thereby implementing the functions. The various data stored in the memory 92 include measured values ​​obtained by the aforementioned plate crown gauge 4 and flatness gauge 5. Additionally, it includes actual values ​​of rolling load, bending force, displacement position, etc., for each stand. Furthermore, it includes a learning table storing first and second learning values ​​for each stand.

[0236] When the processing circuit has dedicated hardware 93, the processing circuit may be, for example, a single circuit, a composite circuit, a programmed processor, or a circuit combining these. Each function is implemented through the processing circuit.

[0237] The embodiments of the present invention have been described above, but the present invention is not limited to the above embodiments, and various modifications can be made without departing from the spirit of the present invention. When the number, quantity, quantity, range, etc., of each element are mentioned in the above embodiments, the present invention is not limited to the mentioned values, except where specifically stated or clearly determined in principle. Furthermore, the structures described in the above embodiments are not essential to the present invention, except where specifically stated or clearly determined in principle.

[0238] Explanation of symbols

[0239] 1. Rolling equipment

[0240] 2. Rolled material

[0241] 3 Plate thickness gauge

[0242] 4-plate convexity gauge

[0243] 5. Flatness gauge

[0244] 6 working rolls

[0245] 7. Working Roller Displacement Mechanism

[0246] 8. Working Roller Bending Mechanism

[0247] 10 Plate Convexity Control Device

[0248] 51 Setting Calculation Department

[0249] 51a Setting Calculated Value Correction

[0250] 52 Actual Calculation Value Calculation Department

[0251] 53 Observation Calculation Department

[0252] 54 First Learning Value Calculation Department

[0253] 55 First Study Department

[0254] 56 Second Learning Value Calculation Department

[0255] 57 Second Study Department

[0256] 91 processor

[0257] 92 Memory

[0258] 93 Hardware

Claims

1. A plate crown control device, comprising, based on a set calculation of the exit plate crown setting value of each stand, a set calculation value of the mechanical plate crown setting value of each stand, and set values ​​of the bending force and displacement position of each stand, controlling a series rolling mill that continuously rolls a workpiece through multiple stands, characterized in that, It has at least one processor and memory. The memory stores: For the measurement point determined in the length direction of the rolled material, the measured value of the exit side plate convexity is determined by the plate convexity meter installed on the exit side of the final stand of the plurality of stands; The actual values ​​of the rolling load, bending force, and displacement position of the measuring points as they pass through each frame; and A learning table that stores the first learning values ​​for each rack. The processor is configured to execute: The actual calculated value is based on the plate thickness of the rolled material on the inlet side of each stand, the plate thickness of the rolled material on the outlet side of each stand, and the actual value to calculate the actual calculated value of the mechanical plate convexity of each stand. The observation values ​​are calculated using the first learning weighting coefficients of each rack, in a manner consistent with the measured values ​​of the convexity of the output side plate of the final rack. The first learning value is calculated based on the difference between the observed value of the mechanical plate convexity of each rack and the actual calculated value of the mechanical plate convexity of each rack, and the first learning current value of each rack is calculated. The first learning process involves updating the first learning values ​​of each rack stored in the learning table by proportionally allocating the current first learning values ​​of each rack and the smoothing gain. as well as The calculation value is corrected by adding the first learning value of each stand read from the learning table to the calculated mechanical plate crown setting value of each stand in the subsequent calculations for the next rolled material. Then, the bending force and displacement position setting values ​​of each stand are calculated. The learning table stores the second learning value for each rack. The actual calculation of the calculated value includes calculating the actual calculated value of the convexity of the exit side plate of each frame based on the actual calculated value of the convexity of the mechanical plate of each frame. The processor is configured to also perform: The second learning value is calculated by using the difference between the actual calculated value of the exit side panel convexity of the final rack and the set calculated value of the exit side panel convexity of the final rack, the second learning weighting coefficient calculated for each rack without using the measured value of the exit side panel convexity, and the transfer rate of each rack to calculate the second learning current value of each rack. as well as The second learning process involves updating the second learning values ​​for each rack stored in the learning table by proportionally allocating the current second learning values ​​for each rack and the smoothing gain. The setting calculation value correction is performed by, in the setting calculation after the next rolled material, using the set calculation value of the mechanical plate crown of each stand, plus the first learning value and the second learning value of each stand read from the learning table, to obtain the corrected mechanical plate crown calculation value of each stand, and then calculating the setting values ​​of the bending force and displacement position of each stand. The first learning weighting coefficient for each rack is a value obtained by multiplying the smaller of the variable range of the first plate convexity ratio and the variable range of the second plate convexity ratio of each rack by a correction coefficient. The variable range of the first plate convexity ratio of each rack is obtained by multiplying the variable range of the mechanical plate convexity of each rack (calculated based on the upper and lower limits of the bending force and displacement position of each rack) by the transfer rate and then dividing by the thickness of the rack's side plate. The variable range of the second plate convexity ratio of each rack is calculated based on the flatness limit of each rack. The second learning weighting coefficient for each rack is the difference between the actual calculated value of the mechanical plate convexity of each rack and the set calculated value of the mechanical plate convexity of each rack, divided by the thickness of the side plate of each rack.

2. The plate convexity control device as described in claim 1, characterized in that, The actual calculation of the calculated value includes calculating the actual calculated value of the convexity of the exit side plate of each frame based on the actual calculated value of the convexity of the mechanical plate of each frame. The observed value is calculated as follows: The deviation between the measured value and the calculated value of the side panel convexity of the final frame is multiplied by the first learning weighting coefficient and the plate thickness ratio of each frame, and then added to the calculated value of the side panel convexity of each frame to calculate the observed value of the side panel convexity of each frame. The mechanical plate convexity observation value of each rack is calculated using the inlet side plate convexity observation value, the outlet side plate convexity observation value, the genetic coefficient, the transfer rate, and the ratio of the inlet side plate thickness to the outlet side plate thickness of each rack.

3. The plate convexity control device as described in claim 1, characterized in that, In the calculation of the observed values, the processor is further configured to, Based on the upper and lower limits of the bending force and displacement position of each frame, the variable range of the mechanical plate convexity of each frame is calculated. The variable range of the first plate convexity ratio for each frame is calculated by multiplying the variable range of the mechanical plate convexity of each frame by the transfer rate and dividing by the thickness of the frame's side plate. The variable range of the second plate convexity ratio for each rack is calculated based on the flatness limit of each rack. The first learning weighting coefficient of each rack is determined by the method that the larger the variable range of the output side plate convexity ratio of each rack obtained by the variable range of the first plate convexity ratio and the variable range of the second plate convexity ratio, the larger the learning weighting coefficient.

4. A plate crown control device, based on the calculated values ​​of the exit plate crown of each stand, the calculated values ​​of the mechanical plate crown of each stand, and the set values ​​of the bending force and displacement position of each stand, controls a series rolling mill that continuously rolls a workpiece through multiple stands, characterized in that... It has at least one processor and memory. The memory stores: For the measurement point determined in the length direction of the rolled material, the measured value of the exit side plate convexity is determined by the plate convexity meter installed on the exit side of the final stand of the plurality of stands; The actual values ​​of the rolling load, bending force, and displacement position of the measuring points as they pass through each frame; and A learning table that stores the first learning values ​​for each rack. The processor is configured to execute: The actual calculated value is based on the plate thickness of the rolled material on the inlet side of each stand, the plate thickness of the rolled material on the outlet side of each stand, and the actual value to calculate the actual calculated value of the mechanical plate convexity of each stand. The observation values ​​are calculated using the first learning weighting coefficients of each rack, in a manner consistent with the measured values ​​of the convexity of the output side plate of the final rack. The first learning value is calculated based on the difference between the observed value of the mechanical plate convexity of each rack and the actual calculated value of the mechanical plate convexity of each rack, and the first learning current value of each rack is calculated. The first learning process involves updating the first learning values ​​of each rack stored in the learning table by proportionally allocating the current first learning values ​​of each rack and the smoothing gain. as well as The calculation value is corrected by adding the first learning value of each stand read from the learning table to the calculated mechanical plate crown setting value of each stand in the subsequent calculations for the next rolled material. Then, the bending force and displacement position setting values ​​of each stand are calculated. The observed value is calculated as follows: The plate convexity prediction formula for each frame is used as a constraint condition. The observed convexity values ​​of the mechanical plate of each frame, as well as the observed convexity values ​​of the exit side plate of each frame other than the final frame, are used as design variables. The sum or sum of squares of the absolute values ​​of the deviations between the observed and actual calculated values ​​of the mechanical plate convexity of each rack, multiplied by the first learning weighting coefficient of each rack, is used as the objective function. Each of the design variables is calculated in a manner that minimizes the objective function. The first learning weighting coefficient for each rack is a value obtained by multiplying the smaller of the first plate convexity ratio variable range and the second plate convexity ratio variable range of each rack by a correction coefficient. The first plate convexity ratio variable range of each rack is obtained by multiplying the mechanical plate convexity variable range of each rack (calculated based on the upper and lower limits of the bending force and displacement position of each rack) by the transfer rate of each rack and then dividing by the thickness of the rack's side plate. The second plate convexity ratio variable range of each rack is calculated based on the flatness limit of each rack.

5. The plate convexity control device as described in claim 4, characterized in that, The design variables include the heritability coefficient of the plate convexity ratio or the transfer rate of each rack. The constraints include inequalities such as the rear rack having a larger plate convexity ratio genetic coefficient compared to the front rack, or the rear rack having a smaller transfer rate compared to the front rack.

6. A plate crown control device, based on the calculated values ​​of the exit plate crown of each stand, the calculated values ​​of the mechanical plate crown of each stand, and the set values ​​of the bending force and displacement position of each stand, controls a series rolling mill that continuously rolls a workpiece through multiple stands, characterized in that, It has at least one processor and memory. The memory stores: For the measurement point determined in the length direction of the rolled material, the measured value of the exit side plate convexity is determined by the plate convexity meter installed on the exit side of the final stand of the plurality of stands; The actual values ​​of the rolling load, bending force, and displacement position of the measuring points as they pass through each frame; and A learning table that stores the second learning values ​​for each rack. The processor is configured to execute: The actual calculation values ​​are calculated based on the plate thickness of the rolled material on the inlet side of each stand, the plate thickness of the rolled material on the outlet side of each stand, and the actual values, to calculate the actual calculated values ​​of the mechanical plate convexity of each stand and the actual calculated values ​​of the outlet plate convexity. The observation values ​​are calculated using the first learning weighting coefficients of each rack, in a manner consistent with the measured values ​​of the convexity of the output side plate of the final rack. The second learning value is calculated by using the difference between the actual calculated value of the exit side panel convexity of the final rack and the set calculated value of the exit side panel convexity of the final rack, the second learning weighting coefficient calculated for each rack without using the measured value of the exit side panel convexity, and the transfer rate of each rack to calculate the second learning current value of each rack. The second learning process involves updating the second learning values ​​of each rack stored in the learning table by proportionally allocating the current second learning values ​​of each rack and the smoothing gain. as well as The calculation value is corrected by adding the second learning value of each stand, which is read from the learning table, to the calculated mechanical plate crown setting value of each stand for the next rolled material. The corrected mechanical plate crown setting value of each stand is then calculated, and the setting values ​​of bending force and displacement position of each stand are calculated. The first learning weighting coefficient for each rack is a value obtained by multiplying the smaller of the variable range of the first plate convexity ratio and the variable range of the second plate convexity ratio of each rack by a correction coefficient. The variable range of the first plate convexity ratio of each rack is obtained by multiplying the variable range of the mechanical plate convexity of each rack (calculated based on the upper and lower limits of the bending force and displacement position of each rack) by the transfer rate and then dividing by the thickness of the rack's side plate. The variable range of the second plate convexity ratio of each rack is calculated based on the flatness limit of each rack. The second learning weighting coefficient for each rack is the difference between the actual calculated value of the mechanical plate convexity of each rack and the set calculated value of the mechanical plate convexity of each rack, divided by the thickness of the side plate of each rack.

Images