Calculation method for edge thinning of medium-thickness rolled steel plate

Abstract

The invention discloses a calculation method for edge thinning of a medium-thickness rolled steel plate. The method comprises the following steps: collecting equipment parameters of a rolling mill and rolling process parameters of a rolled piece; constructing a roll system elastic deformation model, a roll thermal crown model and a roll wear model, and combining to construct a comprehensive roll gap computer mechanical model; finite element simulation, actual measurement data, a mechanism model and a machine learning model are used for accurately predicting the edge thinning of the medium-thickness plate, the outlet plate thickness distribution of each steel plate after each finish rolling pass can be predicted, the corresponding edge thinning amount is given according to different edge positions, and the accuracy of edge thinning prediction is improved. The method is used for guiding an on-site operator to adjust rolling process parameters in time, a working roll changing plan and a rolling plan are arranged more reasonably, and the rolling capacity of medium and thick plates in a steel mill is improved.

Description

Technical Field

[0001] The invention relates to thick plate rolling technology, in particular to a calculation method for edge thinning of steel plates in medium and thick plate rolling. Background Art

[0002] Currently, most patents related to edge thinning research are concentrated in the fields of cold-rolled silicon steel or strip, the production processes of which are very different from those of medium and thick plates, and are not well suited for medium and thick plate production. Secondly, most existing technologies only qualitatively study the correspondence between one or more rolling process parameters and edge thinning, and alleviate the impact of edge thinning by improving the shape of process equipment, such as the roll profile of the working roll. A small number of quantitative predictions of edge thinning under given rolling process conditions only choose to calculate through mechanism models. However, mechanism models have the disadvantage of overly idealized assumptions, resulting in a large gap between the predicted edge thinning and the actual value.

[0003] Existing technology: The patent (publication number: CN101898202B) provides a method for predicting the amount of thinning of strip rolled by an SMS-EDC rolling mill, and the patent (publication number: CN107649521A) provides a method for predicting the edge thinning of strip during the cold rolling process of a six-high rolling mill. Both are applied to cold-rolled strip. The cold rolling process of strip is significantly different from the medium and thick plate rolling process, so they are not suitable for medium and thick plate production.

[0004] Shortcomings of existing technology: 1. The existing technology only studies the edge thinning problem of cold-rolled strip steel. The cold rolling process of strip steel is different from the medium and heavy plate rolling process and is not suitable for medium and heavy plate production; 2. The existing technology only qualitatively studies the corresponding relationship between one or more rolling process parameters and edge thinning, and then reduces edge thinning by improving the shape of process equipment, such as the roll profile of the work roll, but does not quantitatively provide the edge thinning amount under specific rolling process conditions; 3. The existing technology only uses the mechanism model to calculate the edge thinning amount, ignoring the disadvantage that the mechanism model will idealize complex conditions, resulting in deviations in the prediction results and large deviations in the obtained edge thinning amount. Summary of the Invention

[0005] The purpose of the present invention is to overcome the above-mentioned deficiencies of the prior art and to provide a method for calculating the edge thinning of medium and heavy plate rolling.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is: A method for calculating edge thinning of a medium and heavy plate rolled steel plate comprises the following steps: Step 1: Collect the equipment parameters of the rolling mill and the rolling process parameters of the rolled product; Collect the process parameters of the incoming materials at the current production moment of the medium and heavy plate production line, including: steel plate inlet thickness, reduction, steel plate outlet width, rolling force, bending roll force and roll shifting; collect rolling mill equipment parameters, including: work roll body length, bending roll steel spacing, backup roll body length, backup roll bearing seat spacing, work roll body nominal radius, work roll neck radius, backup roll body nominal radius, backup roll neck radius and continuously variable crown mill roll curve parameters , , , , ; Step 2: Construct the roll system elastic deformation model, roll thermal crown model and roll wear model, and combine them to construct a comprehensive roll gap calculation model; Constructing the elastic deformation model of the roller system: Both the working roll and the backup roll bear concentrated loads and distributed loads at the same time, which are expressed as a distributed force φ(x) and a concentrated force F: , Wherein, qwb(x) is the axial distribution of contact pressure between the work roll and the backup roll; P(x) is the axial distribution function of rolling pressure; Using the concept of influence function, if the deformation of unit i is g(i,j) when a unit force is applied to unit j, then the concentrated force φ is applied to unit j. j When , the deformation y0(i,j) generated in unit i is: , For all distributed forces, the deformation yi generated at element i is: , Expressed in matrix form, we have: , in, is the deformation matrix, i.e. the discretized deformation; is the load matrix, i.e. the discretized load; is the influence coefficient matrix; In the construction of the entire mechanism model, the following two equation constraints need to be satisfied: the force balance equation and the deformation coordination relationship equation; Force balance formula: , Where P is the rolling force, F w is the bending roll force, Q wb is the pressure between rollers, I is the unit column vector; Deformation coordination formula: , , Among them, H is the thickness of the rolled piece, H0 is half the height of the center point of the rolled piece, and Y ws Y is the flattening amount of the working roll due to rolling pressure, wb is the flattening amount between the working roll and the support roll, C b is the crown of the support roller, C w is the crown of the working roll, Y w is the working roll deflection, Y b is the support roller deflection; Constructing the roll thermal crown model: The area corresponding to the temperature field is discretized and divided into grids. Based on the heat conduction equation and the actual working conditions of contact heat transfer of each medium, differential equations are constructed for the surface grid points and internal grid points of the work roll respectively. The temperature of each grid point at any time during the operation of the work roll is obtained, and the diameter thermal expansion u(x,t) of the roll axial distance from the center of the steel plate at any time is calculated. The calculation formula is: , Wherein, T(x,r,t) is the numerical calculation temperature of the discrete unit at the axial distance x from the center of the steel plate and the radial distance r from the center of the steel plate, and T0 is the initial temperature; Constructing a roll wear model: After the rolling of a single piece of rolled product is completed, a coordinate system is established with the center of the rolled product as the origin, and the wear of the rolls corresponding to each position is calculated. The calculation formula for the wear is: , in, is the rolling pressure, is the width of the rolled piece, is the contact arc length, is the working roll diameter, is the horizontal coordinate, , is the rolling kilometres, is the lateral distribution of rolling pressure, and These are two undetermined parameters, which are determined by regression analysis based on actual production data; The wear amount of each roller position after rolling a single piece of rolled product is , the formula is: , Among them, S is the actual roller shifting amount; After completing a rolling unit, the wear of the working roll is the sum of the wear of the individual rolled pieces, and the formula is: , in, is the total number of rolled blocks; The roll thermal crown model and roll wear model are calculated cumulatively from the time the rolls are put on the mill, and then the single-pass roll elastic deformation model is superimposed to form a comprehensive roll gap calculation model, which is then used to determine the expression for the workpiece thickness distribution H. Step 3: Construct a finite element simulation model based on the rolling process parameters of the rolled piece and the equipment parameters of the rolling mill in step 1, collect thickness distribution data of the deformation zone of the rolled piece, obtain the distribution of the rolling pressure along the plate width direction, and fit it into a high-order polynomial; based on the constructed finite element simulation model, obtain the distribution of the rolling force along the plate width direction during the rolling process, and fit it into a rolling force distribution function P(x); Step 4: Substitute the rolling force distribution function P(x) obtained in step 3 into the comprehensive roll gap calculation model constructed in step 2 to calculate the outlet thickness distribution of the steel plate; substitute P(x) back into step 2 and iterate the roll gap pressure Q wb The roll elastic deformation model and roll wear model are obtained, and the roll thermal crown model is superimposed to obtain the specific steel plate thickness distribution H; Step 5: Using the difference between the steel plate outlet thickness distribution and the actual steel plate crown as the dependent variable, and the reduction rate of the last finishing pass of the steel plate without empty passes, outlet thickness, outlet width, rolling force, bending roll force, roll shifting amount, and rolling time as characteristic parameters, a regression tree model is constructed to predict the error of the mechanism model. The mechanism model in step 4 is corrected by the machine learning model to obtain a corrected steel plate crown calculation model; Step 6: Based on the modified steel plate convexity calculation model, calculate the edge thinning amount , it is used as the output variable, and the outlet thickness, reduction rate, outlet width, rolling force, bending roll force and roll shifting amount of each pass are used as input variables. A neural network prediction model is constructed, and the mean square error (MSE) is used as the loss function. Starting from a single hidden layer, the network parameters are corrected by the stepwise growth method. The neural network combination with the lowest MSE value is selected to obtain a prediction model for the edge thinning of medium and thick plates. The edge thinning distribution map after each pass is calculated in real time, and the rolling parameters of the next pass are adjusted.

[0007] Furthermore, the specific steps of step 5 are: Step 5.1: Based on the mechanism model of steel plate thickness distribution H constructed in step 4, calculate the outlet plate thickness distribution of each slab after the last non-empty finishing pass, and set mm is the edge thickness calculation point, the center thickness minus the distance from the edge on both sides The average thickness at mm is used to obtain the theoretical steel plate convexity; Step 5.2: Subtract the theoretical steel plate crown from the actual steel plate crown to obtain the error between the elastic deformation model and the actual steel plate crown. Using this error as the dependent variable and the reduction ratio of the last finishing pass of the steel plate without empty passes, exit thickness, exit width, rolling force, bending force, roll shifting, and rolling time as characteristic parameters, a regression tree model of the steel plate crown error model is constructed. Step 5.3: Integrate the steel plate Baidu distribution H mechanism model and the steel plate convexity error model to obtain the corrected steel plate convexity calculation model.

[0008] Furthermore, the specific steps of step 6 are: Step 6.1: Based on the modified steel plate crown calculation model, calculate the outlet plate thickness distribution of each slab in each finishing pass, use the sixth-order function to fit the thickness distribution, and obtain the thickness distribution polynomial coefficient, which is recorded as , and the edge thinning amount is calculated and recorded as ; Step 6.2: Thinning the edges As dependent variables, the exit thickness, reduction rate, exit width, rolling force, roll bending force, roll shifting amount and rolling time of each pass were used as independent variables to construct a neural network prediction model. The mean square error (MSE) was used to evaluate the prediction effect of the neural network model. Starting from a single hidden layer, the network parameters were modified by the stepwise growth method, and the neural network combination with the lowest MSE value was selected. Step 6.3: Based on the medium and thick plate slab rolling data, input the exit thickness, reduction rate, exit width, rolling force, bending roll force, roll shifting amount and rolling time, calculate the edge thinning distribution map after each pass in real time, and adjust the rolling parameters of the next pass.

[0009] The beneficial effects of the present invention are: the present invention uses simulation calculation results and actual data to jointly construct a roll gap mechanism model, and combines it with a machine learning model for precision correction, and predicts the outlet plate thickness distribution of each steel plate after each finishing pass. The edge thinning amount can be obtained according to different edge positions, which is used to guide on-site operators to adjust the rolling process parameters in time, arrange the working roll change plan and rolling plan more reasonably, and improve the rolling capacity of the steel plant.

[0010] Through the implementation of this invention, the cumulative change in edge thinning can be calculated based on the rolling table, allowing for corresponding optimization and adjustment of the rolling schedule. As simulation data gradually accumulates, a neural network model can be constructed that uses process parameters such as entry thickness, reduction ratio, exit width, rolling force, roll bending force, roll shifting, and rolling temperature as input and outputs a rolling force distribution coefficient. This reduces the amount of subsequent simulation calculations and allows for rapid implementation of the overall invention.

[0011] The present invention is flexible and convenient to use, simple to operate, can be applied to similar thick plate factories, and has broad prospects for technical promotion. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is the calculation logic diagram of the present invention; Figure 2 This is a diagram showing the predicted amount of thinning on the left and right sides of a slab according to an embodiment of the present invention; Figure 3 This is a graph showing how the thinning amount on the left and right sides of the slab changes with the bending roll force in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The present invention will be further described in detail below in conjunction with the embodiments. Figure 1 As shown in the figure, the steps of the calculation method for the edge thinning of medium and heavy plate rolling are as follows: Step 1: Collect the equipment parameters of the rolling mill and the rolling process parameters of the rolled product; Collect the process parameters of the incoming materials at the current production moment of the medium and heavy plate production line, including: steel plate inlet thickness, reduction, steel plate outlet width, rolling force, bending roll force and roll shifting; collect rolling mill equipment parameters, including: work roll body length, bending roll steel spacing, backup roll body length, backup roll bearing seat spacing, work roll body nominal radius, work roll neck radius, backup roll body nominal radius, backup roll neck radius and continuously variable crown mill roll curve parameters , , , , ; Step 2: Construct the roll system elastic deformation model, roll thermal crown model and roll wear model, and combine them to construct a comprehensive roll gap calculation model; Constructing the elastic deformation model of the roller system: Both the working roll and the backup roll bear concentrated loads and distributed loads at the same time, which are expressed as a distributed force φ(x) and a concentrated force F: , Wherein, qwb(x) is the axial distribution of contact pressure between the work roll and the backup roll; P(x) is the axial distribution function of rolling pressure; Using the concept of influence function, if the deformation of unit i is g(i,j) when a unit force is applied to unit j, then the concentrated force φ is applied to unit j. j When , the deformation y0(i,j) generated in unit i is: , For all distributed forces, the deformation yi generated at element i is: , Expressed in matrix form, we have: , in, is the deformation matrix, i.e. the discretized deformation; is the load matrix, i.e. the discretized load; is the influence coefficient matrix; In the construction of the entire mechanism model, the following two equation constraints need to be satisfied: the force balance equation and the deformation coordination relationship equation; Force balance formula: , Where P is the rolling force, F w is the bending roll force, Q wb is the pressure between rollers, I is the unit column vector; Deformation coordination formula: , , Among them, H is the thickness of the rolled piece, H0 is half the height of the center point of the rolled piece, and Y ws Y is the flattening amount of the working roll due to rolling pressure, wb is the flattening amount between the working roll and the support roll, C b is the crown of the support roller, C w is the crown of the working roll, Y w is the working roll deflection, Y b is the support roller deflection; Constructing the roll thermal crown model: The area corresponding to the temperature field is discretized and meshed. Based on the heat conduction equation and the actual working conditions of contact heat transfer of each medium, differential equations are constructed for the surface grid points and internal grid points of the work roll. The heat conduction equation is as follows: , The temperature of each grid point at any time during the working roll is obtained, and the diameter thermal expansion u(x,t) of the roll axial distance from the center of the steel plate at any time is calculated. The calculation formula is: , Wherein, T(x,r,t) is the numerical calculation temperature of the discrete unit at the axial distance x from the center of the steel plate and the radial distance r from the center of the steel plate, and T0 is the initial temperature; Constructing a roll wear model: According to the general law of tribology, the amount of wear on the contact surface is proportional to the load it bears and the distance of relative sliding and rolling. Therefore, the specific form of wear is as follows: , In the formula is the wear amount, is the wear coefficient, is the normal pressure on the contact surface, is the sliding or rolling distance of the contact surface.

[0014] The "slice method" is used to establish a wear roll profile model, using previously measured wear roll profile data as a reference. Furthermore, the lateral distribution of rolling pressure must be factored in to more accurately characterize the wear roll profile. Specifically, after rolling a single piece of workpiece, a coordinate system is established with the workpiece center as the origin, and the roll wear at each position is calculated. The wear calculation formula is: , in, is the rolling pressure, is the width of the rolled piece, is the contact arc length, is the working roll diameter, is the horizontal coordinate, , is the rolling kilometres, is the lateral distribution of rolling pressure, and These are two undetermined parameters, which are determined by regression analysis based on actual production data; The wear amount of each roller position after rolling a single piece of rolled product is , the formula is: , Among them, S is the actual roller shifting amount; After completing a rolling unit, the wear of the working roll is the sum of the wear of the individual rolled pieces, and the formula is: , in, is the total number of rolled blocks; The roll thermal crown model and roll wear model are calculated cumulatively from the time the rolls are put on the mill, and then the single-pass roll elastic deformation model is superimposed to form a comprehensive roll gap calculation model, which is then used to determine the expression for the workpiece thickness distribution H. Step 3: Construct a finite element simulation model based on the rolling process parameters of the rolled piece and the equipment parameters of the rolling mill in step 1, collect thickness distribution data of the deformation zone of the rolled piece, obtain the distribution of the rolling pressure along the plate width direction, and fit it into a high-order polynomial; based on the constructed finite element simulation model, obtain the distribution of the rolling force along the plate width direction during the rolling process, and fit it into a rolling force distribution function P(x); Step 4: Substitute the rolling force distribution function P(x) obtained in step 3 into the comprehensive roll gap calculation model constructed in step 2 to calculate the outlet thickness distribution of the steel plate; substitute P(x) back into step 2 and iterate the roll gap pressure Qwb The roll elastic deformation model and roll wear model are obtained, and the roll thermal crown model is superimposed to obtain the specific steel plate thickness distribution H; Step 5: Using the difference between the steel plate outlet thickness distribution and the actual steel plate crown as the dependent variable, and the reduction rate of the last finishing pass of the steel plate without empty passes, outlet thickness, outlet width, rolling force, bending roll force, roll shifting amount, and rolling time as characteristic parameters, a regression tree model is constructed to predict the error of the mechanism model. The mechanism model in step 4 is corrected by the machine learning model to obtain a corrected steel plate crown calculation model; Step 5.1: Based on the mechanism model of steel plate thickness distribution H constructed in step 4, calculate the outlet plate thickness distribution of each slab after the last non-empty finishing pass, and set mm is the edge thickness calculation point, the center thickness minus the distance from the edge on both sides The average thickness at mm is used to obtain the theoretical steel plate convexity; Step 5.2: Subtract the theoretical steel plate crown from the actual steel plate crown to obtain the error between the elastic deformation model and the actual steel plate crown. Using this error as the dependent variable and the reduction ratio of the last finishing pass of the steel plate without empty passes, exit thickness, exit width, rolling force, bending force, roll shifting, and rolling time as characteristic parameters, a regression tree model of the steel plate crown error model is constructed. Step 5.3: Integrate the steel plate Baidu distribution H mechanism model and the steel plate convexity error model to obtain the corrected steel plate convexity calculation model.

[0015] Step 6: Based on the modified steel plate convexity calculation model, calculate the edge thinning amount , it is used as the output variable, and the outlet thickness, reduction rate, outlet width, rolling force, bending roll force and roll shifting amount of each pass are used as input variables. A neural network prediction model is constructed, and the mean square error (MSE) is used as the loss function. Starting from a single hidden layer, the network parameters are corrected by the stepwise growth method. The neural network combination with the lowest MSE value is selected to obtain a prediction model for the edge thinning of medium and thick plates. The edge thinning distribution map after each pass is calculated in real time, and the rolling parameters of the next pass are adjusted.

[0016] Step 6.1: Based on the modified steel plate crown calculation model, calculate the outlet plate thickness distribution of each slab in each finishing pass, use the sixth-order function to fit the thickness distribution, and obtain the thickness distribution polynomial coefficient, which is recorded as , and the edge thinning amount is calculated and recorded as ; Step 6.2: Thinning the edges As dependent variables, the exit thickness, reduction rate, exit width, rolling force, roll bending force, roll shifting amount and rolling time of each pass were used as independent variables to construct a neural network prediction model. The mean square error (MSE) was used to evaluate the prediction effect of the neural network model. Starting from a single hidden layer, the network parameters were modified by the stepwise growth method, and the neural network combination with the lowest MSE value was selected. Step 6.3: Based on the medium and thick plate slab rolling data, input the exit thickness, reduction rate, exit width, rolling force, bending roll force, roll shifting amount and rolling time, calculate the edge thinning distribution map after each pass in real time, and adjust the rolling parameters of the next pass.

[0017] Specific implementation examples Figure 2 、 Figure 3 shown. Example 1

[0018] Set edge thickness calculation point mm, taking the actual rolling process data of a steel plate on the rolling line of a thick plate mill as an example, the thinning amount of the steel plate on both sides is obtained by using the invented technology. Figure 2 As shown in the figure, if the edge thinning value is specified as the difference between the thickness of the steel plate at 130mm and the thickness of the steel plate at 15mm, then the thinning amount of the left edge of the steel plate can be seen from the figure. 23.22µm, right side thinning 23.16µm. Example 2

[0019] Set edge thickness calculation point mm, taking the actual rolling process data of a slab on the rolling line of a thick plate mill as an example, using the invented technology, such as Figure 2 As shown in the figure, it can be concluded that when rolling with the same width, if other rolling variables are kept unchanged and the bending roll force is increased, the edge thinning will be reduced. This provides theoretical support for further developing the edge thinning calculation model for rolling with the same width, guiding on-site operators to adjust the finishing rolling parameters in time, and optimizing the rolling plan arrangement.

[0020] The above content is only used to illustrate the technical solution of the present invention. Simple modifications or equivalent substitutions of the technical solution of the present invention made by ordinary technicians in this field do not deviate from the essence and scope of the technical solution of the present invention.

Claims

1. A method for calculating edge thinning of medium and heavy plate rolling, characterized by: The steps include: Step 1: Collect the equipment parameters of the rolling mill and the rolling process parameters of the rolled product; Collect the process parameters of the incoming materials at the current production moment of the medium and heavy plate production line, including: steel plate inlet thickness, reduction, steel plate outlet width, rolling force, bending roll force and roll shifting; collect rolling mill equipment parameters, including: work roll body length, bending roll steel spacing, backup roll body length, backup roll bearing seat spacing, work roll body nominal radius, work roll neck radius, backup roll body nominal radius, backup roll neck radius and continuously variable crown mill roll curve parameters , , , , ; Step 2: Construct the roll system elastic deformation model, roll thermal crown model and roll wear model, and combine them to construct a comprehensive roll gap calculation model; Constructing the elastic deformation model of the roller system: Both the working roll and the backup roll bear concentrated loads and distributed loads at the same time, which are expressed as a distributed force φ(x) and a concentrated force F: Wherein, qwb(x) is the axial distribution of contact pressure between the work roll and the backup roll; P(x) is the axial distribution function of rolling pressure; Using the concept of influence function, if the deformation of unit i is g(i,j) when a unit force is applied to unit j, then the concentrated force φ is applied to unit j. j When , the deformation y0(i,j) generated in unit i is: For all distributed forces, the deformation yi generated at element i is: Expressed in matrix form, we have: in, is the deformation matrix, i.e. the discretized deformation; is the load matrix, i.e. the discretized load; is the influence coefficient matrix; In the construction of the entire mechanism model, the following two equation constraints need to be satisfied: the force balance equation and the deformation coordination relationship equation; Force balance formula: ， Where P is the rolling force, F w is the bending roll force, Q wb is the pressure between rollers, I is the unit column vector; Deformation coordination formula: Among them, H is the thickness of the rolled piece, H0 is half the height of the center point of the rolled piece, and Y ws Y is the flattening amount of the working roll due to rolling pressure, wb is the flattening amount between the working roll and the support roll, C b is the crown of the support roller, C w is the crown of the working roll, Y w is the working roll deflection, Y b is the support roller deflection; Constructing the roll thermal crown model: The area corresponding to the temperature field is discretized and divided into grids. Based on the heat conduction equation and the actual working conditions of contact heat transfer of each medium, differential equations are constructed for the surface grid points and internal grid points of the work roll respectively. The temperature of each grid point at any time during the operation of the work roll is obtained, and the diameter thermal expansion u(x,t) of the roll axial distance from the center of the steel plate at any time is calculated. The calculation formula is: Wherein, T(x,r,t) is the numerical calculation temperature of the discrete unit at the axial distance x from the center of the steel plate and the radial distance r from the center of the steel plate, and T0 is the initial temperature; Constructing a roll wear model: After the rolling of a single piece of rolled product is completed, a coordinate system is established with the center of the rolled product as the origin, and the wear of the rolls corresponding to each position is calculated. The calculation formula for the wear is: in, is the rolling pressure, is the width of the rolled piece, is the contact arc length, is the working roll diameter, is the horizontal coordinate, , is the rolling kilometres, is the lateral distribution of rolling pressure, and These are two undetermined parameters, which are determined by regression analysis based on actual production data; The wear amount of each roller position after rolling a single piece of rolled product is , the formula is: Among them, S is the actual roller shifting amount; After completing a rolling unit, the wear of the working roll is the sum of the wear of the individual rolled pieces, and the formula is: in, is the total number of rolled blocks; The roll thermal crown model and roll wear model are calculated cumulatively from the time the rolls are put on the mill, and then the single-pass roll elastic deformation model is superimposed to form a comprehensive roll gap calculation model, which is then used to determine the expression for the workpiece thickness distribution H. Step 3: Construct a finite element simulation model based on the rolling process parameters of the rolled piece and the equipment parameters of the rolling mill in step 1, collect thickness distribution data of the deformation zone of the rolled piece, obtain the distribution of the rolling pressure along the plate width direction, and fit it into a high-order polynomial; based on the constructed finite element simulation model, obtain the distribution of the rolling force along the plate width direction during the rolling process, and fit it into a rolling force distribution function P(x); Step 4: Substitute the rolling force distribution function P(x) obtained in step 3 into the comprehensive roll gap calculation model constructed in step 2 to calculate the outlet thickness distribution of the steel plate; substitute P(x) back into step 2 and iterate the roll gap pressure Q wb The roll elastic deformation model and roll wear model are obtained, and the roll thermal crown model is superimposed to obtain the specific steel plate thickness distribution H; Step 5: Using the difference between the steel plate outlet thickness distribution and the actual steel plate crown as the dependent variable, and the reduction rate of the last finishing pass of the steel plate without empty passes, outlet thickness, outlet width, rolling force, bending roll force, roll shifting amount, and rolling time as characteristic parameters, a regression tree model is constructed to predict the error of the mechanism model. The mechanism model in step 4 is corrected by the machine learning model to obtain a corrected steel plate crown calculation model; Step 6: Based on the modified steel plate convexity calculation model, calculate the edge thinning amount , it is used as the output variable, and the outlet thickness, reduction rate, outlet width, rolling force, bending roll force and roll shifting amount of each pass are used as input variables. A neural network prediction model is constructed, and the mean square error (MSE) is used as the loss function. Starting from a single hidden layer, the network parameters are corrected by the stepwise growth method. The neural network combination with the lowest MSE value is selected to obtain a prediction model for the edge thinning of medium and thick plates. The edge thinning distribution map after each pass is calculated in real time, and the rolling parameters of the next pass are adjusted.

2. The method for calculating edge thinning of plate rolled steel plates according to claim 1, characterized in that: The specific steps of step 5 are: Step 5.1: Based on the mechanism model of steel plate thickness distribution H constructed in step 4, calculate the outlet plate thickness distribution of each slab after the last non-empty finishing pass, and set mm is the edge thickness calculation point, the center thickness minus the distance from the edge on both sides The average thickness at mm is used to obtain the theoretical steel plate convexity; Step 5.2: Subtract the theoretical steel plate crown from the actual steel plate crown to obtain the error between the elastic deformation model and the actual steel plate crown. Using this error as the dependent variable and the reduction ratio of the last finishing pass of the steel plate without empty passes, exit thickness, exit width, rolling force, bending force, roll shifting, and rolling time as characteristic parameters, a regression tree model of the steel plate crown error model is constructed. Step 5.3: Integrate the steel plate Baidu distribution H mechanism model and the steel plate convexity error model to obtain the corrected steel plate convexity calculation model.

3. The method for calculating edge thinning of plate rolled steel plates according to claim 1, characterized in that: The specific steps of step 6 are: Step 6.1: Based on the modified steel plate crown calculation model, calculate the outlet plate thickness distribution of each slab in each finishing pass, use the sixth-order function to fit the thickness distribution, and obtain the thickness distribution polynomial coefficient, which is recorded as , and the edge thinning amount is calculated and recorded as ; Step 6.2: Thinning the edges As dependent variables, the exit thickness, reduction rate, exit width, rolling force, roll bending force, roll shifting amount and rolling time of each pass were used as independent variables to construct a neural network prediction model. The mean square error (MSE) was used to evaluate the prediction effect of the neural network model. Starting from a single hidden layer, the network parameters were modified by the stepwise growth method, and the neural network combination with the lowest MSE value was selected. Step 6.3: Based on the medium and thick plate slab rolling data, input the exit thickness, reduction rate, exit width, rolling force, bending roll force, roll shifting amount and rolling time, calculate the edge thinning distribution map after each pass in real time, and adjust the rolling parameters of the next pass.

Images