Multi-objective optimization method for cold rolling process regulation

Abstract

According to the technical scheme, the cold rolling process regulation multi-objective optimization method comprises the following steps that on the basis of field production data, regression calculation compensation is carried out on key parameters in a rolling force mathematical model by combining data fusion and an intelligent algorithm; a genetic algorithm is adopted to optimize random forest hyper-parameters to predict and optimize the rolling force compensation and correction model, and the rolling force prediction model is based on the compensation and correction friction coefficient; a multi-algorithm fusion mechanism is established to optimize the rolling schedule with three targets of minimum energy consumption, stable rolling and dynamic weight balance of plate and strip surface quality and production efficiency, and different process parameters are optimized for different weights so as to adapt to different working conditions and achieve the purposes of improving product quality, reducing production energy consumption and stabilizing the rolling production process.

Description

Technical Field

[0001] The invention belongs to the technical field of rolling control, and in particular relates to a multi-objective rolling schedule optimization control method for a cold rolling process. Background Art

[0002] As the core processing technology in the field of thin plate manufacturing, the production process of cold-rolled strip steel focuses on fully tapping the potential of equipment, enhancing the quality of plate shape and reducing energy consumption. To achieve this goal, it is particularly important to deeply analyze the internal mechanism of the rolling process and accurately model the core parameters. However, it also faces many challenges, including high coupling between parameters, nonlinear characteristics, deep correlation, and non-steady-state dynamic changes. In particular, as the core variable for regulating production, the improvement of the prediction accuracy of rolling force not only helps to strengthen the quality monitoring of the product processing process, but also effectively optimizes the rolling schedule. Given that the cold-rolled strip steel process has a very high pursuit of precision and complexity, therefore, in the practice of strip steel rolling, scientific planning and adjustment of rolling procedures play a vital role in extending equipment life, reducing energy consumption and ensuring the excellence of finished product quality.

[0003] In view of the above situation, a multi-objective optimization method for cold rolling process schedule based on field data is proposed. Based on the field production data, data fusion and intelligent algorithm are combined to perform regression calculation and compensation for the key parameters in the rolling force mathematical model, and the genetic algorithm is used to optimize the random forest hyperparameters to predict and optimize the rolling force compensation correction model. Based on the rolling force prediction model of the compensated corrected friction coefficient, a multi-algorithm fusion mechanism is established to optimize the rolling schedule with the three objectives of minimum energy consumption, rolling stability, and dynamic weight balance of plate and strip surface quality and production efficiency. Different process parameters are optimized for different weights to adapt to different working conditions, so as to achieve the purpose of improving product quality, reducing production energy consumption and stabilizing the rolling production process. Summary of the invention

[0004] The purpose of the present invention is to provide a multi-objective optimization method for cold rolling process schedule, which takes minimum energy consumption, smooth rolling, and dynamic weight balance of plate and strip surface quality and production efficiency as optimization objectives, and assigns different weights to different working conditions to improve the above-mentioned technical problems.

[0005] To achieve the above objectives, the present invention provides a multi-objective optimization method for a cold rolling process schedule, comprising:

[0006] Collect production data and technical parameters from the rolling site, clean and normalize the collected data to generate the original data set;

[0007] Based on the original data set, relevant rolling process parameters are screened out by a feature selection method of recursive feature elimination and mutual information method, and relevant rolling process parameter data are fused by using data fusion technology;

[0008] Establish a mathematical mechanism model of relevant rolling process parameters, use the fused data to calculate the friction coefficient in the form of back-compensation, use the genetic algorithm to regress the coefficients to be corrected, and use the genetic algorithm to optimize the hyperparameters of the random forest algorithm to predict the corrected rolling force model;

[0009] Based on the rolling force prediction model with compensation for the modified friction coefficient, a multi-objective rolling procedure optimization model with three objectives, minimum energy consumption, smooth rolling, and dynamic weight balance between plate and strip surface quality and production efficiency, is established. The Pauta criterion is used for data preprocessing, and the exit thickness of each pass is used as the decision variable. The process constraints are used for restriction, and the intelligent algorithm is integrated to optimize and solve the quality-efficiency balance problem under different weights.

[0010] Optionally, the production data and technical parameters include: emulsion flow, temperature, injection angle, bending roll force of working rolls and intermediate rolls, rolling force, reduction, rolling speed, plate and strip thickness, and flatness data.

[0011] Optionally, the data fusion technology is:

[0012]

[0013] Where n is the data type, w is the weight coefficient of each process parameter, i is each related rolling process, n is the total number of rolling process parameters, and X is the new data set after fusion.

[0014] Optionally, establishing a mathematical mechanism model of relevant rolling process parameters includes:

[0015] According to the plate shape control and roller elastic-plastic deformation theory, the rolling force model is established:

[0016]

[0017] Where F is the rolling force, B is the width of the rolled piece, l' is the contact arc length between the roller and the deformation zone of the rolled piece after the roller is flattened during rolling, R is the roller radius, R' is the radius of the roller after flattening, Q P is the external friction stress state coefficient of the roller after flattening, K T is the tension influence coefficient, K is the deformation resistance of the principal stress in the width direction, μ is the friction coefficient, Δh is the absolute reduction, τ f , τ b are the front and rear tensions respectively, ε is the degree of deformation, ε=(h 0 +h 1 ) / h0 , where h 0 is the entrance thickness of each pass, h 1 is the exit thickness of each pass;

[0018] Based on the theory of elastic-plastic mechanics and contact mechanics, the interaction model between the working roll and the intermediate roll is established:

[0019]

[0020] Where p(x) is the contact pressure, P is the rolling force, a is the width of the contact area, x is the position in the contact area, E is the equivalent Young's modulus, R is the roller radius, δ is the deformation, L is the contact length, and h is 0 is the inlet thickness of the strip, h f is the exit thickness of the plate;

[0021] Based on elastic mechanics and beam bending theory, the bending roll force model of the working roll and the intermediate roll is established:

[0022]

[0023] Where, δ(x) is the deflection of the working roll and the intermediate roll, F b is the bending roll force, x is the distance from one end of the working roll and the intermediate roll to any point, L is the length of the working roll and the intermediate roll, E is the elastic modulus of the material, p(x) is the contact pressure, p 0 is the initial contact pressure when no bending roll force is applied, A is the contact area, f(x) is the distribution function of the contact pressure along the length of the working roll, δ max is the maximum deflection of the working roll or the intermediate roll, K is the bending stiffness, and I is the moment of inertia of the working roll and the intermediate roll;

[0024] Based on fluid mechanics and tribology, the emulsion and plate quality model is established:

[0025]

[0026] Where Q is the flow rate of the emulsion, μ 0 is the emulsion at the reference temperature T 0 The viscosity under the condition of temperature, k is the coefficient of viscosity change with temperature, T is the actual temperature of the emulsion, v r is the speed of the roller surface, A is the contact area, μ f is the friction coefficient, h f is the actual strip thickness, h 0 is the target thickness of the plate, k 1 is the coefficient between friction and plate quality, F r is the rolling force.

[0027] Optionally, the friction coefficient is calculated using the back-compensation form:

[0028]

[0029] In the formula, α is the lubricant influence coefficient, v is the roller linear speed, a, b, c, d are the coefficients to be corrected, and h out Rolled product outlet thickness, h in Thickness of rolled product at entry.

[0030] Optionally, the steps of using a genetic algorithm to optimize the hyperparameters of a random forest algorithm to predict the modified rolling force model are:

[0031] Select a certain amount of experimental data and divide it into training set and test set in a ratio of 7:3;

[0032] The genetic algorithm uses the number of decision trees and the model depth of each tree as decision variables, extracts the corresponding hyperparameters within the specified range, and inputs the extracted hyperparameters into the random forest algorithm for training;

[0033] Select the hyperparameters that give the best prediction results;

[0034] Use the test set data to verify the prediction results of the model under the corresponding hyperparameters.

[0035] Optionally, the objective function of the minimum energy consumption is:

[0036] minf=∑P i

[0037] In the formula, f is the total rolling power, P is the rolling power of each pass, and i is the pass;

[0038] The objective function of rolling stability is:

[0039]

[0040] Where f is the equal power margin objective function, P r is the rated power of the rolling mill motor, P is the rolling mill motor power, i, j are the passes;

[0041] The objective function of the dynamic weight balance between the strip surface quality and production efficiency is:

[0042]

[0043] In the formula, f 1 is the quality-efficiency balance function, w 1 is the surface quality weight, w 2 is the production efficiency weight, and w 1 +w 2 =1,h f h is the strip outlet thickness, 0is the target thickness of the plate, f s k is the strip straightness deviation, 1 , k 2 is the correlation coefficient weight of process parameters, f q is the effect of emulsion on the quality of strip, v r is the rolling speed.

[0044] Optionally, the process constraints are:

[0045]

[0046] In the formula, F r is the rolling force, F max is the maximum rolling force allowed by the rolling mill, F b is the bending roll force, is the maximum capacity of the roll bending system, v r is the rolling speed, v max is the maximum allowable rolling speed, Q and T are the flow rate and speed of the emulsion respectively.

[0047] Optionally, the steps of optimizing and solving the quality-efficiency balance problem under different weights by the fusion intelligent algorithm are as follows:

[0048] Optimize the HOA using NSGA-II:

[0049] Set the initial parameters of each algorithm, including particle swarm size, population size, number of iterations, etc.

[0050] Define the objective function, define the multi-objective optimization function based on the data in the production process, and consider the constraints of the actual process;

[0051] Perform multi-objective optimization, use NSGA-II for diversity enhancement and optimization of HOA algorithm, generate updated Pareto solution set, perform elite selection and non-dominated sorting, retain the optimal solution, ensure uniform distribution of solutions in the solution set in the target space based on congestion calculation, and select the optimal solution for implementation and application of process parameters.

[0052] Optionally, the Pauta criterion calculation formula is as follows:

[0053]

[0054] In the formula, is the mean of the data set, S x is the standard deviation of the data set, and N is the amount of data in the data set.

[0055] The technical effects of the present invention are:

[0056] The present invention provides a multi-objective optimization method for cold rolling process schedule, which optimizes the rolling schedule with three objectives of minimum energy consumption, smooth rolling, and dynamic weight balance between plate and strip surface quality and production efficiency. Different process parameters are optimized according to different weights to adapt to different working conditions, thereby achieving the purpose of improving product quality, reducing production energy consumption, and stabilizing the rolling production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 It is a schematic diagram of the principle of the drive and sensor of a single-stand reversible cold rolling mill;

[0059] Figure 2 It is a flow chart for optimizing the comprehensive procedure for quality-efficiency balance in the present invention;

[0060] Among them, 1. left coiler; 2. right coiler; 3. left roller tension meter; 4. right roller tension meter; 5. left flatness roller; 6. right flatness roller; 7. left laser speed meter; 8. right laser speed meter; 9. left thickness gauge; 10. right thickness gauge; 11. rolling mill; 12. hydraulic cylinder position pressure sensor. DETAILED DESCRIPTION

[0061] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0062] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0063] Embodiment 1

[0064] like Figure 1-Figure 2 As shown in the figure, the principle of the tension control system for single-stand reversible cold rolling coiling Figure 1In the figure, the cold rolling mill system structure includes a left coiler 1, a right coiler 2, a left roller tension meter 3, a right roller tension meter 4, a left flat roller 5, a right flat roller 6, a left laser speed meter 7, a right laser speed meter 8, a left thickness gauge 9, a right thickness gauge 10, a rolling mill 11, and a hydraulic cylinder position pressure sensor 12. The roller tension meter, the laser speed meter and the thickness gauge collect technical parameters and production data during the cold rolling process, and are electrically connected to the left coiler 1 and the right coiler 2 via the PLC controller. The working process of the system is: the strip steel is rolled from the left coiler 1 through the left flat roller 4 to the rolling mill 10, and after rolling is completed, it is coiled at the right coiler 2 via the right flat roller 5. There is an inlet left thickness gauge 8 and a right thickness gauge 9 at the inlet and outlet sides of the rolling mill, respectively, to measure the outlet and inlet thickness, and output analog or digital signals. A hydraulic cylinder position pressure sensor 12 is installed on the rolling mill 10, and the hydraulic pressure adjustment of the working roll is performed according to the rolling pressure or the pressure position setting value calculated by the mathematical model. Encoders are installed on the left coiler 1 and the right coiler 2 to measure the coiling speed, which is used to control the speed of the coiler and calculate the coil diameter. All signals are finally summarized in the PLC controller for relevant calculation and output to realize the rolling process of the plate and strip.

[0065] Embodiment 2

[0066] like Figure 2 As shown, this embodiment provides a multi-objective optimization method for cold rolling process schedule, including:

[0067] Collect production data and technical parameters from the rolling site, clean and normalize the collected data to generate the original data set;

[0068] Relevant process parameters are screened out through recursive feature elimination and mutual information method feature selection methods, and data fusion technology is used to further improve the data utilization and accuracy of the system;

[0069] Establish a mathematical mechanism model of relevant rolling process parameters, use the fused data to calculate the friction coefficient in the form of back-compensation, use the genetic algorithm to regress the coefficients to be corrected, and use the genetic algorithm to optimize the hyperparameters of the random forest algorithm to predict the corrected rolling force model;

[0070] Based on the rolling force prediction model with compensated modified friction coefficient, a multi-objective rolling schedule optimization model with three objectives, minimum energy consumption, rolling stability, and dynamic weight balance between plate and strip surface quality and production efficiency, is established. Taking the actual process constraints into consideration, the exit thickness of each pass is used as the decision variable, and the intelligent algorithm is integrated to optimize and solve the quality-efficiency balance problem under different weights.

[0071] The production data and technical parameters include: emulsion flow, temperature, injection angle, bending force of working rolls and intermediate rolls, rolling force, reduction, rolling speed, plate and strip thickness, straightness and other data.

[0072] The mutual information method for nonlinear correlation analysis is as follows:

[0073]

[0074] Where X and Y are the input and output variables respectively, p(x,y) is the joint distribution of variables X and Y, and p(x) and p(y) are the marginal distributions of variables X and Y.

[0075] The mathematical mechanism model of the relevant rolling process parameters:

[0076] Based on the plate shape control and roller elastic-plastic theory, the rolling force model is established:

[0077]

[0078] Where F is the rolling force, B is the width of the rolled piece, l' is the contact arc length between the roller and the deformation zone of the rolled piece after the roller is flattened during rolling, R is the roller radius, R' is the radius of the roller after flattening, Q P is the external friction stress state coefficient of the roller after flattening, K T is the tension influence coefficient, K is the deformation resistance of the principal stress in the width direction, μ is the friction coefficient, Δh is the absolute reduction, τ f , τ b are the front and rear tensions respectively, ε is the degree of deformation, ε=(h 0 +h 1 ) / h 0 , where h 0 is the entrance thickness of each pass, h 1 is the exit thickness of each pass.

[0079] Based on the theories of elastic-plastic mechanics and contact mechanics, the interaction model between the working roll and the intermediate roll is established as follows:

[0080]

[0081] Where p(x) is the contact pressure, P is the rolling force, a is the width of the contact area, x is the position in the contact area, E is the equivalent Young's modulus, R is the roller radius, δ is the deformation, L is the contact length, and h is 0 is the inlet thickness of the strip, h f is the exit thickness of the strip.

[0082] According to the theory of elastic-plastic mechanics, the working roll and the intermediate roll will undergo elastic deformation during the rolling process, and the deformation amount is:

[0083]

[0084] Where δ is the deformation, P is the rolling force, a is the width of the contact area, R is the roller radius, and E is the equivalent Young's modulus, which represents the elastic stiffness between the two rollers and is defined as:

[0085]

[0086] In the formula, E 1 、E 2 are the Young's modulus of the working roll and the intermediate roll, respectively, v 1 、v 2 is their Poisson's ratio.

[0087] When the contact load is large, plastic deformation may occur between the rollers. In this case, the relationship between contact pressure and deformation is nonlinear. Usually, a modified model in elastic-plastic mechanics is used. Assuming that plastic yield occurs in some areas, the stress of the plastic deformation part can be expressed as:

[0088] σ P =σ y +kε n

[0089] In the formula, σ P is the stress of plastic deformation, σ y is the yield strength of the material, k and n are the plastic constants of the material, and ε is the strain of the material.

[0090] During the rolling process, the contact length between the working roll and the strip and the intermediate roll is an important parameter for calculating stress distribution and deformation. The contact length can be expressed by the formula:

[0091]

[0092] Where L is the contact length, R is the roller radius, and h 0 is the inlet thickness of the strip, h f is the exit thickness of the strip.

[0093] Based on elastic mechanics and beam bending theory, the bending roll force model of the working roll and the intermediate roll is established as follows:

[0094]

[0095] Where, δ(x) is the deflection of the working roll and the intermediate roll, F b is the bending roll force, x is the distance from one end of the working roll and the intermediate roll to any point, L is the length of the working roll and the intermediate roll, E is the elastic modulus of the material, p(x) is the contact pressure, p 0is the initial contact pressure when no bending roll force is applied, A is the contact area, f(x) is the distribution function of the contact pressure along the length of the working roll, δ max is the maximum deflection of the working roll or the intermediate roll, K is the bending stiffness, and I is the moment of inertia of the working roll and the intermediate roll. The moment of inertia for the cylindrical roll can be expressed as:

[0096]

[0097] Where I is the moment of inertia and D is the diameter of the working roll or intermediate roll.

[0098] Among them, the bending roll force changes the deflection of the working roll and the intermediate roll, thereby affecting the contact pressure distribution between them. Assuming that the contact pressure between the two rolls is evenly distributed, when the bending roll force acts, the contact pressure will change along the contact line. The contact pressure can be expressed as:

[0099]

[0100] Where p(x) is the contact pressure, F b is the bending roll force, p 0 is the initial contact pressure when no bending roll force is applied, A is the contact area, and f(x) is the distribution function of the contact pressure along the length of the working roll.

[0101] Among them, the bending stiffness of the working roll and the intermediate roll is one of the key parameters for controlling the deflection. The bending stiffness can be expressed as:

[0102]

[0103] Where K is the bending stiffness, I is the moment of inertia of the working roll and the intermediate roll, and E is the elastic modulus of the material.

[0104] Among them, when the bending roll force acts on the roller, the deformation (deflection) of the roller is inversely proportional to the bending stiffness and the bending roll force, specifically:

[0105]

[0106] In the formula, δ max is the maximum deflection of the working roll or intermediate roll, F b is the bending roll force, L is the length of the working roll and the intermediate roll, E is the elastic modulus of the material, and I is the moment of inertia of the working roll and the intermediate roll.

[0107] Based on fluid mechanics and tribology, the emulsion and plate quality model is established as follows:

[0108]

[0109] Where Q is the flow rate of the emulsion, μ 0is the emulsion at the reference temperature T 0 The viscosity under the condition of temperature, k is the coefficient of viscosity change with temperature, T is the actual temperature of the emulsion, v r is the speed of the roller surface, A is the contact area, μ f is the friction coefficient, h f is the actual strip thickness, h 0 is the target thickness of the plate, k 1 is the coefficient between friction and plate quality, F r is the rolling force.

[0110] Among them, the emulsion forms a lubricating film between the roller and the strip, and its thickness depends on the emulsion flow rate, roller speed and emulsion viscosity. Therefore, the emulsion film thickness can be expressed as:

[0111]

[0112] In the formula, h l is the thickness of the emulsion film, Q is the flow rate of the emulsion, μ is the dynamic viscosity of the emulsion, v r is the speed of the roller surface and A is the contact area.

[0113] Among them, the viscosity of the emulsion changes with temperature, which can generally be expressed by the empirical formula:

[0114]

[0115] In the formula, μ is the viscosity of the emulsion, μ 0 is the emulsion at the reference temperature T 0 where k is the coefficient of viscosity change with temperature, and T is the actual temperature of the emulsion.

[0116] Among them, the friction between the strip and the roller is mainly affected by the lubrication effect of the emulsion and the rolling force. According to Coulomb's friction law, the relationship between the friction force, the rolling force and the friction coefficient is:

[0117] F f =μ f F r

[0118] In the formula, F f is the friction force, μ f is the friction coefficient, F r is the rolling force.

[0119] Among them, the friction coefficient changes with the viscosity of the emulsion and can be expressed as:

[0120]

[0121] In the formula, μ f is the friction coefficient, μ0 is the initial value of the friction coefficient, and μ(T) is the viscosity of the emulsion determined by the aforementioned temperature relationship.

[0122] Among them, the influence of emulsion on the strip shape quality affects the flatness and thickness distribution of the strip through friction. Combining fluid mechanics and tribology, the strip shape quality can be expressed as:

[0123]

[0124] In the formula, h l is the thickness of the emulsion film, μ f is the friction coefficient, h f is the actual strip thickness, h 0 is the target thickness of the plate, k 1 is the coefficient between friction and plate quality, F f For friction.

[0125] The friction coefficient calculation model using compensation is:

[0126]

[0127] In the formula, α is the lubricant influence coefficient, v is the roller linear speed, a, b, c, d are the coefficients to be corrected, and h out Rolled product outlet thickness, h in Thickness of rolled product at entry.

[0128] The steps of using the genetic algorithm to optimize the hyperparameters of the random forest algorithm to predict the rolling force are as follows:

[0129] Select a certain amount of experimental data and divide it into training set and test set in a ratio of 7:3;

[0130] The genetic algorithm uses the number of decision trees and the model depth of each tree as decision variables, extracts the corresponding hyperparameters within the specified range, and inputs the extracted hyperparameters into the random forest algorithm for training;

[0131] Select the hyperparameters that give the best prediction results;

[0132] Use the test set data to verify the prediction results of the model under the corresponding hyperparameters.

[0133] The minimum energy consumption objective function is:

[0134] minf=∑P i

[0135] Where f is the total rolling power, P i is the rolling power of each pass.

[0136] The rolling stability equal power margin objective function is:

[0137]

[0138] Where f is the equal power margin objective function, P r is the rated power of the motor of the rolling mill, P i is the rolling mill motor power of the i-th pass.

[0139] The dynamic weight balance function between the strip surface quality and production efficiency is:

[0140]

[0141] In the formula, w 1 is the surface quality weight, w 2 is the production efficiency weight, and w 1 +w 2 =1,h f h is the strip outlet thickness, 0 is the target thickness of the plate, f s k is the strip straightness deviation, 1 , k 2 is the correlation coefficient weight of process parameters, f q is the effect of emulsion on the quality of strip, v r is the rolling speed.

[0142] Among them, the flatness deviation of the strip is f s Mainly affected by the bending roll force F b and rolling force F r The bending roll force is used to offset the elastic deformation of the roll, thereby ensuring the flatness of the strip surface, which can be expressed as:

[0143] f s =k 1 ·F r +k 2 ·F b

[0144] In the formula, k 1 and k 2 It is the empirical coefficient between straightness and rolling force and bending roll force.

[0145] Among them, the rolling speed v r Directly affects the time it takes for the strip to pass through the rolling mill, which in turn affects production efficiency, rolling speed and rolling force F r There is an inverse relationship between the friction coefficient μ and the emulsion parameters (flow rate Q and temperature T) have an impact on the friction coefficient, thus indirectly affecting the rolling speed:

[0146]

[0147] In the formula, vmax is the maximum allowable rolling speed, k is a process-related constant, and μ is the friction coefficient under the influence of the emulsion.

[0148] Among them, the energy consumption E can be obtained by integrating the rolling power P:

[0149]

[0150] Where, t is the rolling time, M is the main motor torque, v is the rolling speed, η is the motor efficiency, R' is the radius of the roller after flattening, M p is the rolling torque, M f is the additional friction torque, M x is the idling torque, M d For the dynamic torque.

[0151] The process constraints are:

[0152]

[0153] In the formula, F r is the rolling force, F max is the maximum rolling force allowed by the rolling mill, F b is the bending roll force, is the maximum capacity of the roll bending system, v r is the rolling speed, v max is the maximum allowable rolling speed, Q and T are the flow rate and speed of the emulsion respectively.

[0154] The steps of optimizing and solving the optimization problem by the fusion intelligent algorithm are as follows:

[0155] Optimize the HOA using NSGA-II:

[0156] Set the initial parameters of each algorithm, including particle swarm size, population size, number of iterations, etc.

[0157] Define the objective function, define the multi-objective optimization function based on the data in the production process, and consider the constraints of the actual process;

[0158] Perform multi-objective optimization, use NSGA-II for diversity enhancement and optimization of HOA algorithm, generate updated Pareto solution set, perform elite selection and non-dominated sorting, retain the optimal solution, ensure uniform distribution of solutions in the solution set in the target space based on congestion calculation, and select the optimal solution for implementation and application of process parameters.

[0159] The Pauta criterion calculation formula is as follows:

[0160]

[0161] In the formula, is the mean of the data set, S x is the standard deviation of the data set, and N is the amount of data in the data set.

[0162] The above is only a preferred specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A multi-objective optimization method for cold rolling process schedule, characterized in that: include: Collect production data and technical parameters from the rolling site, clean and normalize the collected data to generate the original data set; Based on the original data set, relevant rolling process parameters are screened out by a feature selection method of recursive feature elimination and mutual information method, and relevant rolling process parameter data are fused by using data fusion technology; Establish a mathematical mechanism model of relevant rolling process parameters, use the fused data to calculate the friction coefficient in the form of back-compensation, use the genetic algorithm to regress the coefficients to be corrected, and use the genetic algorithm to optimize the hyperparameters of the random forest algorithm to predict the corrected rolling force model; Based on the rolling force prediction model with compensation for the modified friction coefficient, a multi-objective rolling procedure optimization model with three objectives, minimum energy consumption, smooth rolling, and dynamic weight balance between plate and strip surface quality and production efficiency, is established. The Pauta criterion is used for data preprocessing, and the exit thickness of each pass is used as the decision variable. The process constraints are used for restriction, and the intelligent algorithm is integrated to optimize and solve the quality-efficiency balance problem under different weights.

2. A multi-objective optimization method for cold rolling process schedule according to claim 1, characterized in that: The production data and technical parameters include: emulsion flow, temperature, injection angle, bending roll force of working roll and intermediate roll, rolling force, reduction, rolling speed, plate and strip thickness, and flatness data.

3. A multi-objective optimization method for cold rolling process schedule according to claim 1, characterized in that: The data fusion technology is: Where n is the data type, w is the weight coefficient of each process parameter, i is each related rolling process, n is the total number of rolling process parameters, and X is the new data set after fusion.

4. A multi-objective optimization method for cold rolling process schedule according to claim 1, characterized in that: The mathematical mechanism model of the relevant rolling process parameters is established including: According to the plate shape control and roller elastic-plastic deformation theory, the rolling force model is established: Where F is the rolling force, B is the width of the rolled piece, l' is the contact arc length between the roller and the deformation zone of the rolled piece after the roller is flattened during rolling, R is the roller radius, R' is the radius of the roller after flattening, Q P is the external friction stress state coefficient of the roller after flattening, K T is the tension influence coefficient, K is the deformation resistance of the principal stress in the width direction, μ is the friction coefficient, Δh is the absolute reduction, τ f , τ b are the front and rear tensions, ε is the degree of deformation, ε=(h0+h1) / h0, where h0 is the entrance thickness of each pass and h1 is the exit thickness of each pass; Based on the theory of elastic-plastic mechanics and contact mechanics, the interaction model between the working roll and the intermediate roll is established: Where p(x) is the contact pressure, P is the rolling force, a is the width of the contact area, x is the position in the contact area, E is the equivalent Young's modulus, R is the roller radius, δ is the deformation, L is the contact length, h0 is the inlet thickness of the strip, and h f is the exit thickness of the plate; Based on elastic mechanics and beam bending theory, the bending roll force model of the working roll and the intermediate roll is established: Where, δ(x) is the deflection of the working roll and the intermediate roll, F b is the bending roll force, x is the distance from one end of the working roll and the intermediate roll to any point, L is the length of the working roll and the intermediate roll, E is the elastic modulus of the material, p(x) is the contact pressure, p0 is the initial contact pressure when no bending roll force is applied, A is the contact area, f(x) is the distribution function of the contact pressure along the length of the working roll, δ max is the maximum deflection of the working roll or the intermediate roll, K is the bending stiffness, and I is the moment of inertia of the working roll and the intermediate roll; Based on fluid mechanics and tribology, the emulsion and plate quality model is established: In the formula, Q is the flow rate of the emulsion, μ0 is the viscosity of the emulsion at the reference temperature T0, k is the coefficient of viscosity change with temperature, T is the actual temperature of the emulsion, and v r is the speed of the roller surface, A is the contact area, μ f is the friction coefficient, h f is the actual strip thickness, h0 is the target strip thickness, k1 is the coefficient between friction and strip quality, F r is the rolling force.

5. A multi-objective optimization method for cold rolling process schedule according to claim 1, characterized in that: The friction coefficient is calculated using the back-compensation method: In the formula, α is the lubricant influence coefficient, v is the roller linear speed, a, b, c, d are the coefficients to be corrected, and h out Rolled product outlet thickness, h in Thickness of rolled product at entry.

6. A multi-objective optimization method for cold rolling process schedule according to claim 1, characterized in that: The steps of using genetic algorithm to optimize the hyperparameters of random forest algorithm to predict the modified rolling force model are: Select a certain amount of experimental data and divide it into training set and test set in a ratio of 7:3; The genetic algorithm uses the number of decision trees and the model depth of each tree as decision variables, extracts the corresponding hyperparameters within the specified range, and inputs the extracted hyperparameters into the random forest algorithm for training; Select the hyperparameters that give the best prediction results; Use the test set data to verify the prediction results of the model under the corresponding hyperparameters.

7. A multi-objective optimization method for cold rolling process schedule according to claim 1, characterized in that: The objective function of the minimum energy consumption is: minf=∑P i In the formula, f is the total rolling power, P is the rolling power of each pass, and i is the pass; The objective function of rolling stability is: Where f is the equal power margin objective function, P r is the rated power of the rolling mill motor, P is the rolling mill motor power, i, j are the passes; The objective function of the dynamic weight balance between the strip surface quality and production efficiency is: Where f1 is the quality-efficiency balance function, w1 is the surface quality weight, w2 is the production efficiency weight, and w1+w2=1, h f is the strip exit thickness, h0 is the strip target thickness, f s is the flatness deviation of the plate strip, k1 and k2 are the correlation coefficient weights of the process parameters, and f q is the effect of emulsion on the quality of strip, v r is the rolling speed.

8. A multi-objective optimization method for cold rolling process schedule according to claim 1, characterized in that: The process constraints are: In the formula, F r is the rolling force, F max is the maximum rolling force allowed by the rolling mill, F b is the bending roll force, is the maximum capacity of the roll bending system, v r is the rolling speed, v max is the maximum allowable rolling speed, Q and T are the flow rate and speed of the emulsion respectively.

9. A multi-objective optimization method for cold rolling process schedule according to claim 1, characterized in that: The steps of optimizing and solving the quality-efficiency balance problem under different weights by the fusion intelligent algorithm are as follows: Optimize the HOA using NSGA-II: Set the initial parameters of each algorithm, including particle swarm size, population size, and number of iterations; Define the objective function, define the multi-objective optimization function based on the data in the production process, and consider the constraints of the actual process; Perform multi-objective optimization, use NSGA-II for diversity enhancement and optimization of HOA algorithm, generate updated Pareto solution set, perform elite selection and non-dominated sorting, retain the optimal solution, ensure uniform distribution of solutions in the solution set in the target space based on congestion calculation, and select the optimal solution for implementation and application of process parameters.

10. A multi-objective optimization method for cold rolling process schedule according to claim 1, characterized in that: The Pauta criterion calculation formula is as follows: In the formula, is the mean of the data set, S x is the standard deviation of the data set, and N is the amount of data in the data set.

Images