Cold rolling temper mill plate shape closed-loop control method based on data driving

Abstract

The invention discloses a cold rolling temper mill strip shape closed-loop control method based on data driving, and relates to the technical field of strip steel rolling and temper production, and the method comprises the following steps: carrying out strip shape mode identification on outlet strip steel of a cold rolling unit, converting a complex strip shape into a strip shape mode coefficient, and based on historical production data, carrying out cleaning, screening and analysis to obtain a strip shape model coefficient; the method comprises the following steps: establishing a temper mill data model between a strip shape regulation and control means and a strip shape mode coefficient, utilizing model prediction control to construct a strip shape closed-loop controller of the temper mill, and utilizing the strip shape closed-loop controller to calculate an optimal control parameter of the strip shape regulation and control means based on an actually measured strip steel strip shape mode coefficient and a target strip shape mode coefficient. And the information is issued to the plate shape regulation and control mechanism for closed-loop regulation. According to the method, data analysis, model prediction control and real-time closed-loop control are combined, the problems that the cold rolling unit temper mill plate shape control precision is poor, and stable and continuous production is difficult are solved, and the plate shape quality and yield of plate and strip products are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of strip rolling and tempering production, and particularly relates to a data-driven closed-loop shape control method for a cold rolling temper mill. Background Art

[0002] Cold-rolled strip steel is an important product in the iron and steel industry. Due to its excellent performance and wide applicability, it has been widely used in many fields such as automobile manufacturing and household appliances. The shape flatness is one of the important criteria for measuring the quality of strip steel. However, it is affected by various factors, and the control process involves multi-variable coupling, which has always been a research hotspot and difficulty in cold rolling. In the closed-loop shape control, most current units still adopt traditional control strategies such as manual control, sequential control, or proportional control. However, due to the coupling effect between various regulating mechanisms, it is difficult for such traditional control strategies to efficiently and stably complete the shape control. In view of this, some scholars have proposed various optimized control methods to improve the automation and intelligence level of shape control.

[0003] Li Jing et al. [1] Aiming at the shape control problem of a four-high temper mill, a shape feedforward model based on the rolling force change was developed. By establishing a mathematical model based on the rolling force change and embedding it into the PLC program, the automatic adjustment of the bending roll force was realized. The actual application results show that the feedforward control system can effectively avoid shape defects caused by drastic changes in rolling force and significantly improve the shape quality. However, this method only considers the influence of rolling force on the shape and does not study complex temper mills such as six-high mills, so the application range is relatively narrow.

[0004] Chen Caijun et al. [2] Based on the production data of a 1550 cold rolling production line in a certain factory, using a machine learning algorithm based on python, with parameters such as temper rolling force, elongation, and tension as inputs and the control quantity of the temper mill as the output value, a prediction regression model was trained. The model operation results show that its control effect is relatively excellent, it runs stably, and it is seamlessly integrated with the existing L1, L2, and L3 systems, improving the overall automation level of the system. However, this machine learning model is too simple, and it is difficult to obtain an ideal control effect when facing complex tempering situations.

[0005] It is worth noting that Li Bo et al. [3] Adopting the model predictive control theory, a shape prediction controller was designed and applied to the shape control system of a four-high CVC cold strip mill, realizing strong robustness and anti-interference ability. The closed-loop shape control system of the temper mill can be regarded as a multi-input multi-output system that changes in real time [4]The calculation of the regulation amounts of its multiple flatness control means can be regarded as the solution of a multi-output optimal control system. MPC is exactly applicable to this situation. The simulation results of Li Bo et al. show that, compared with the traditional PID control, this MPC model has a fast response speed, small overshoot, and significantly improved dynamic performance. However, this method has a large dependence on the model. The modeling process of Li Bo et al. is divorced from actual production data and only through MATLAB simulation. The simulation results cannot fully reflect the complex situations in actual production.

[0006] In summary, there is an urgent need for a data-driven flatness closed-loop control method for cold rolling temper mills in China at present.

[0007] The references are as follows:

[0008] [1] Li Jing, Ma Jinfeng, Wang Fei, et al. Research and application of flatness feedforward model for four-high temper mill [J / OL]. Steel Rolling, 2019, 36(1): 58 - 59. DOI: 10.13228 / j.boyuan.issn1003 - 9996.20180003.

[0009] [2] Chen Caijun, Yang Zhiying, Zheng Haicheng. Research and application of automatic flatness control model for cold rolling galvanizing production line [J]. Metallurgical Industry Automation, 2023, 47(S1): 387 - 390.

[0010] [3] Li Bo, Zhang Qingdong. Research on flatness predictive control of four-high CVC cold strip mill [J]. Metallurgical Equipment, 2008(4): 9 - 12.

[0011] [4] Zhang Qingdong, Chen Xianlin, He Anrui, et al. Flatness automatic control system model for wide strip cold rolling mill [J]. Steel Rolling, 1998(5): 11 - 15. Summary of the Invention

[0012] The purpose of the present invention is to provide a data-driven flatness closed-loop control method for cold rolling temper mills to overcome the above defects in the prior art.

[0013] A data-driven flatness closed-loop control method for cold rolling temper mills includes the following steps:

[0014] S1. Identify the mode coefficients of the strip flatness at the outlet of the cold rolling mill;

[0015] S2. Through cleaning, screening, and analyzing historical production data, use the method of system identification to fit the temper mill data model between the flatness control means and the flatness mode coefficients.

[0016] S3. Through the leveler data model, a flatness closed-loop controller is established using the model predictive control algorithm, where the model predictive control algorithm is based on a discrete state-space model and consists of model prediction, rolling optimization, and feedback correction processes.

[0017] S4. Through the flatness mode coefficient of the strip steel at the outlet of the cold rolling mill, by taking the difference from the target flatness mode coefficient, the flatness deviation value is obtained and input into the flatness closed-loop controller to obtain the adjustment amount of the flatness adjustment mechanism, which is then sent to the actuator to complete the closed-loop adjustment of the strip steel flatness.

[0018] Preferably, the identification of the flatness mode coefficient of the strip steel at the outlet of the cold rolling mill in step S1 includes:

[0019] Scaling the bandwidths of different strip steels to the range of [-1, 1].

[0020] Based on the flatness distribution of the strip steel at the outlet of the cold rolling mill obtained by the flatness meter, using Legendre orthogonal polynomials, the flatness mode coefficient of the strip steel at the outlet of the cold rolling mill is identified. The relationship among the measured flatness, Legendre orthogonal polynomials, and flatness mode coefficient is shown in the following formula:

[0021] Y(y) = c1p1(y) + c2p2(y) + c4p4(y)

[0022] where Y(y) is the measured flatness, c1, c2, and c4 are the flatness mode coefficients, and p1(y), p2(y), and p4(y) are the first, second, and fourth Legendre polynomials respectively, as shown in the following formula:

[0023] p1(y) = y

[0024]

[0025] Preferably, the cleaning, screening, and analysis of the historical production data in step S2 include:

[0026] The historical production data includes strip steel width, strip steel inlet thickness, yield strength, inlet speed, outlet speed, target elongation, actual elongation, rolling force, inlet tension, outlet tension, roll tilt, intermediate roll bending force, and work roll bending force. The historical production data is divided by the interquartile range, and some moderate outliers and extreme outliers are deleted.

[0027] Data analysis is based on the cleaned and screened dataset, and the Pearson correlation coefficient is used to analyze the correlation degree between variables.

[0028] Preferably, the calculation formula for the Pearson correlation coefficient in step S2 is:

[0029]

[0030] where Xi is a sample point, is the sample average value, and n is the number of samples.

[0031] Preferably, the system identification of the planishing mill model in step S2 includes:

[0032] System identification uses the least squares method as the parameter fitting method and the discrete state space as the model. The input parameters of the model are strip inlet thickness, strip inlet speed, rolling force, outlet tension, roll tilt, intermediate roll bending force, and work roll bending force. The output parameters of the model are the shape mode coefficients c1, c2, and c4. The expression of the discrete state space model is as follows:

[0033]

[0034] In the formula: x(k) represents the state quantity at time k; y(k) represents the output quantity at time k; A represents the system state matrix; B represents the system input matrix; C represents the system output matrix.

[0035] Preferably, the model predictive control in step S3 is based on the prediction of the future behavior of the model, and determines the optimal input by solving the optimal solution of the loss function within a finite number of steps.

[0036] Preferably, at time k in step S3, the model predictive control is divided into the following three steps:

[0037] S3.1. Estimate or measure the current system state x k ;

[0038] S3.2. Based on u k 、u k+1 、u k+2 、…、u k+N find the optimal solution of the loss function for optimal control, where u is the input at each time and N is the prediction horizon;

[0039] The loss function is:

[0040]

[0041] In the formula, e is the error between the output value and the reference value; Q is the error weight coefficient; u is the input; R is the input weight coefficient; F is the weight coefficient of the terminal state quantity; (k + i|k) represents the quantity when predicting i steps forward from time k;

[0042] S3.3. Only execute the current u at time k k ;

[0043] Preferably, the adjustment of the shape adjustment mechanism in step S4 includes roll tilt, intermediate roll bending, and work roll bending.

[0044] The beneficial effects achieved by the present invention are as follows:

[0045] In this application, shape pattern recognition is performed on the strip steel at the outlet of the cold rolling mill. The complex shape can be converted into shape pattern coefficients. Based on historical production data, after cleaning, screening, and analysis, a leveler data model between shape control means and shape pattern coefficients is established. Using model predictive control, a shape closed-loop controller for the leveler is constructed. Based on the shape pattern coefficients of the strip steel and the target shape pattern coefficients, the optimal control parameters of the shape control means are calculated using the established shape closed-loop controller, and then sent to the shape control mechanism for closed-loop adjustment. The present invention effectively combines data analysis, model predictive control, and real-time closed-loop control technologies, can solve the problems of poor shape control accuracy and difficulty in stable continuous production of the existing cold rolling mill leveler, and thus significantly improves the shape quality and output of the strip product. Description of the Drawings

[0046] Figure 1 It is a schematic diagram of the distribution of a section of historical processing data of a cold rolling leveler collected in an embodiment of the present invention. Among them, (a) is a violin distribution diagram of strip steel specification parameters, (b) is a violin distribution diagram of rolling parameters, and (c) is a violin distribution diagram of leveling parameters;

[0047] Figure 2 They are the images of Legendre first, second, and fourth polynomials;

[0048] Figure 3 It is a fitting curve diagram of the original shape curve and Legendre orthogonal polynomial of a certain cross-section in an embodiment of the present invention;

[0049] Figure 4 It is a Pearson correlation heat map of each variable based on the historical processing data of the leveler in an embodiment of the present invention;

[0050] Figure 5 It is a fitting curve of the original shape pattern coefficients c1, c2, and c4 and the discrete state space model of a certain section in an embodiment of the present invention;

[0051] Figure 6 It is a flow chart of the model predictive control algorithm in an embodiment of the present invention;

[0052] Figure 7 It is a closed-loop control flow chart of the shape of the leveler established in an embodiment of the present invention;

[0053] Figure 8 It is a structural diagram of the cooperation between the MPC controller of the cold rolling strip leveler and the cold rolling site L1, L2, and L3 built in an embodiment of the present invention; Detailed Embodiment

[0054] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the accompanying drawings in the embodiments of this application. Apparently, the described embodiments are some, but not all, of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts shall fall within the protection scope of this application.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs; the terms used in the description of this application in the specification are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the description and claims of this application and the drawings are intended to cover non-exclusive inclusion.

[0056] Referring to "embodiments" herein means that the specific features, structures, or characteristics described in connection with the embodiments may be included in at least one embodiment of this application. The phrase "embodiment" appearing in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein may be combined with other embodiments.

[0057] First, briefly summarize that the technical steps of this embodiment mainly include:

[0058] 1. Identify the shape pattern coefficient of the strip steel at the outlet of the cold rolling mill:

[0059] The identification of the shape pattern coefficient is based on the measured strip shape of the steel plate, and uses Legendre orthogonal first, second, and fourth polynomials to fit the shape curve to obtain the shape pattern coefficient;

[0060] 2. Establish a data model of the cold rolling temper mill based on data-driven:

[0061] By cleaning, screening, and analyzing the historical production data of a cold rolling temper mill, use the method of system identification to establish a data model of the temper mill between the shape adjustment means and the shape pattern coefficient; among them, the model of the system identification is a discrete state space model; the estimation method for fitting the system parameters is the least squares method;

[0062] 3. Establish a shape leveling closed-loop MPC controller:

[0063] Based on the temper mill data model, use the model predictive control algorithm to establish a shape closed-loop controller; among them, the model predictive control algorithm is composed of model prediction, rolling optimization, and feedback correction processes based on the temper mill data model.

[0064] 4. Research on the Shape Closed-Loop Control System of the Cold Rolling Skin Pass Mill:

[0065] Based on the real-time shape mode coefficient of the strip steel at the outlet of the cold rolling mill, by taking the difference from the target shape mode coefficient, the shape deviation value is obtained and input into the MPC controller. The adjustment amount of the shape adjustment mechanism is obtained and sent to the actuator to complete the closed-loop control of the strip steel shape.

[0066] A data-driven method applicable to the shape closed-loop control of the cold rolling skin pass mill proposed in this embodiment is as Figure 7 shown, and specifically includes the following steps:

[0067] Step 1: Identify the shape mode coefficient of the strip steel at the outlet of the cold rolling mill;

[0068] Furthermore, the identification of the mode coefficient of the strip steel at the outlet of the cold rolling mill includes:

[0069] Scale the strip width to between [-1, 1]; obtain the strip steel shape by the shape meter at the outlet of the cold rolling mill; use the Legendre orthogonal polynomial to identify the actual shape mode coefficient of the strip steel.

[0070] Specifically, in this embodiment, the identification of the mode coefficient of the strip steel shape at the outlet of the cold rolling mill:

[0071] 1. In the actual production process, the width of each coil of strip steel is not fixed, that is, the number of effective acquisition points of the thickness gauge is different. When the strip width changes, the number of acquisition points in the middle part of the shape meter in contact with the strip steel will also change accordingly. In this regard, select the non-zero stress value in the middle of the shape meter and uniformly scale it to [-1, 1], that is, scale the strip width to between [-1, 1];

[0072] 2. Fit the measured shape through the Legendre polynomial. Based on the measured strip steel shape at the outlet of the cold rolling mill obtained by the shape meter, use the Legendre orthogonal polynomial to identify the shape mode coefficient of the strip steel at the outlet of the cold rolling mill. Among them, the measured shape, the Legendre orthogonal polynomial, and the shape mode coefficient are shown in the following formula:

[0073] Y(y) = c1p1(y) + c2p2(y) + c4p4(y)

[0074] Among them, Y(y) is the measured shape, c1, c2, and c4 are the shape mode coefficients, and p1(y), p2(y), and p4(y) are the Legendre first, second, and fourth polynomials respectively, as shown in the following formula:

[0075] p1(y) = y

[0076]

[0077] The diagrams of the Legendre orthogonal polynomials of the first, second, and fourth degrees are shown as Figure 2 follows. Among them, according to the positive and negative of the flatness pattern coefficients, each pattern coefficient can be divided into two, positive and negative; the measured flatness curve and the flatness curve fitted by the Legendre orthogonal polynomial are shown as Figure 3 follows.

[0078] Step 2: Establishment of a data-driven cold rolling temper mill data model

[0079] Furthermore, the establishment of a data-driven cold rolling temper mill data model includes:

[0080] First, clean, screen, and analyze the historical production data of a certain cold rolling temper mill; secondly, use the method of system identification to fit the temper mill data model between the main flatness control factors and the flatness pattern coefficients; among them, the model of the system identification is a discrete state space model; the estimation method of the fitting system parameters is the least squares method;

[0081] The historical production data includes strip width, strip inlet thickness, yield strength, inlet speed, outlet speed, target elongation, actual elongation, rolling force, inlet tension, outlet tension, roll tilt, intermediate roll bending force, and work roll bending force, etc.;

[0082] The data cleaning, screening, and analysis include: dividing the data by the interquartile range, deleting some moderately outlying values and extreme outlying values, and the violin distribution diagrams of some data are shown as Figure 1 follows;

[0083] The data analysis is to use the Pearson correlation coefficient to analyze the correlation degree between variables based on the dataset after cleaning and screening. Among them, the calculation formula of the Pearson correlation coefficient is:

[0084]

[0085] where X i is the sample point, is the sample average value, and n is the number of samples.

[0086] As Figure 4As shown, it is a Pearson correlation heatmap among various variables based on the historical production data. It can be seen that there is a relatively high correlation between roll tilt and the flatness pattern coefficient c1; there is a certain correlation between strip inlet thickness, yield strength, inlet speed, outlet speed, rolling force, outlet tension, intermediate roll bending force, and work roll bending force and the flatness pattern coefficient c2; there is a certain correlation between inlet speed, outlet speed, rolling force, outlet tension, work roll bending force, and intermediate roll bending force and the flatness pattern coefficient c4. In addition, it can be seen from the heatmap that the inlet speed and the outlet speed have an extremely high positive correlation. For the convenience of subsequent research, only the inlet speed among them is selected as one of the influencing factors. Considering the above comprehensively, the strip inlet thickness, outlet speed, rolling force, outlet tension, roll tilt amount, work roll bending force, and work roll bending force are selected as the inputs of the skin pass mill model, and the flatness pattern coefficients c1, c2, and c4 are used as the outputs of the skin pass mill data model.

[0087] Using system identification to fit the skin pass mill data model includes:

[0088] The system identification uses the least squares method as the parameter fitting method and the discrete state space model as the model. Among them, the input parameters of the model are strip inlet thickness, strip inlet speed, rolling force, outlet tension, roll tilt amount, intermediate roll bending force, and work roll bending force, and the output parameters of the model are the flatness pattern coefficients c1, c2, and c4; the expression of the discrete state space model is as follows:

[0089]

[0090] In the formula: x(k) represents the state quantity at time k; y(k) represents the output quantity at time k; A represents the system state matrix; B represents the system input matrix; C represents the system output matrix.

[0091] Step 3: Establishment of the MPC controller for strip flatness

[0092] Based on the skin pass mill data model, a strip shape closed-loop controller is established using the model predictive control algorithm; among them, the model predictive control algorithm is composed of model prediction, rolling optimization, and feedback correction processes based on the skin pass mill data model;

[0093] The model predictive control is based on the prediction of the future behavior of the system and determines the optimal input by solving the optimal solution of the loss function within a finite number of steps. The mechanism of the model predictive control is as Figure 6 shown, including three important links: prediction model, rolling optimization, and feedback correction. Their respective main contents can be summarized as:

[0094] 1. Prediction model: That is, the skin pass mill data model;

[0095] 2. Rolling Optimization: At the next moment, it is necessary to measure the state variables at this moment again and use them as the initial adjustment to re-optimize and solve.

[0096] 3. Feedback Correction: The difference between the state variable and the target variable is used as feedback information to form an overall closed-loop control;

[0097] The rolling optimization generally consists of 3 steps at time k:

[0098] 1. Estimate or measure the current system state x k ;

[0099] 2. Based on u k 、u k+1 、u k+2 、…、u k+N Find the optimal solution of the loss function for optimal control, where u is the input at each moment and N is the prediction horizon;

[0100] The loss function is:

[0101]

[0102] In the formula, e is the error between the output value and the reference value; Q is the error weight coefficient; u is the input; R is the input weight coefficient; F is the weight coefficient of the terminal state variable; (k + i|k) represents the quantity when predicting i steps forward from time k;

[0103] 3. Only execute the current u at time k k ;

[0104] The MPC controller will calculate the optimal solution of the loss function in each control step. Among them, the prediction horizon, control horizon, error weight, and input weight will greatly affect the control effect of the MPC controller. The prediction horizon will affect the controller's prediction ability for the future. The control horizon will affect the controller's control ability. In addition, the ratio of the error weight to the input weight is also the key to adjusting the controller's effect. Increasing the ratio of the error weight will make the controller pay more attention to the adjustment effect, and vice versa, it will pay more attention to reducing the control quantity. The rolling optimization needs to be carried out under certain constraint conditions;

[0105] Taking everything into consideration, the parameters and constraints of the MPC controller are shown in Table 1.

[0106] Table 1 MPC controller parameters and simulation constraints Tabel 1 MPC controller parameters andsimulation constraints

[0107]

[0108]

[0109] Step 4: Research on the shape closed-loop control system of the cold rolling temper mill;

[0110] Based on the real-time shape pattern coefficient of the strip steel at the outlet of the cold rolling mill, by taking the difference from the target shape pattern coefficient, the shape deviation value is obtained and input into the MPC controller. The adjustment amount of the shape adjustment mechanism is obtained and sent to the actuator to complete the closed-loop control of the strip steel shape.

[0111] Build the shape closed-loop control system of the temper mill as Figure 7 shown. At a certain operating moment, the temper mill will identify the shape pattern coefficient using the Legendre orthogonal polynomial according to the measurement results of the current shape meter to obtain the current shape pattern coefficient. By taking the difference from the target shape pattern coefficient, the pattern coefficient of the shape deviation can be obtained.

[0112] Based on the pattern coefficient of the current shape deviation, the current strip steel inlet thickness, rolling force, outlet tension and inlet speed, the MPC controller will perform an optimal solution with a finite number of steps on the loss function to obtain the adjustment amount of the shape control mechanism.

[0113] The said shape control mechanism includes: roll tilt, intermediate roll bending and work roll bending;

[0114] Send the adjustment amount of the shape adjustment mechanism calculated by the MPC controller to each shape adjustment mechanism to complete the shape closed-loop adjustment process of the temper mill.

[0115] Combine the shape closed-loop control system of the temper mill with the cold rolling site L1, L2 and L3, and its structural flow chart is as Figure 8 shown. When the shape closed-loop system of the temper mill is put into use, it will read the target shape of the current strip steel from the L2 and HMI screens, and then take the difference from the actual shape read through the L1 communication to obtain the real-time strip steel shape deviation. Subsequently, the shape deviation is put into the shape closed-loop MPC controller to obtain the adjustment values of the shape control means, namely the roll tilt amount, intermediate roll bending force and work roll bending force, and sent to the L1 communication module and the HMI screen module to finally realize the automatic closed-loop adjustment of the shape flatness.

[0116] In this embodiment, by identifying the shape pattern coefficient of the strip steel at the outlet of the cold rolling mill, the real-time shape pattern coefficient is obtained; then, the adjustment amount of the shape control means of the temper mill is calculated by the shape MPC controller established based on the data-driven temper mill data model. Finally, the adjustment amount is sent to the shape control mechanism to complete the shape closed-loop control of the temper mill. This comprehensive control idea effectively combines data analysis, prediction model and real-time control technology, providing a scientific and intelligent solution for the shape control of the cold rolling temper mill.

[0117] The embodiments of the present invention described above do not constitute a limitation on the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the claims of the present invention.

Claims

1. A data-driven closed-loop shape control method for cold rolling temper mill, characterized in that: It includes the following steps: S1. Identify the pattern coefficients of the strip shape at the outlet of the cold rolling mill; S2. By cleaning, screening, and analyzing historical production data, use the method of system identification to fit the leveler data model between the strip shape control means and the strip shape pattern coefficients; S3. Through the leveler data model, use the model predictive control algorithm to establish a strip shape closed-loop controller; among them, the model predictive control algorithm is based on the discrete state space model and consists of the model prediction, rolling optimization, and feedback correction processes; S4. Through the strip shape pattern coefficients at the outlet of the cold rolling mill, by taking the difference from the target strip shape pattern coefficients, obtain the strip shape deviation value, input it into the strip shape closed-loop controller, obtain the adjustment amount of the strip shape adjustment mechanism, and send it to the actuator to complete the closed-loop adjustment of the strip shape.

2. The shape closed-loop control method of a cold rolling temper mill based on data driving according to claim 1, wherein: In step S1, the identification of the strip shape pattern coefficients of the strip at the outlet of the cold rolling mill includes: Scale the bandwidths of different strips to between [-1, 1]; Based on the strip shape distribution of the strip at the outlet of the cold rolling mill obtained by the shape meter, use the Legendre orthogonal polynomial to identify the strip shape pattern coefficients of the strip at the outlet of the cold rolling mill. The relationship between the measured strip shape, the Legendre orthogonal polynomial, and the strip shape pattern coefficients is shown in the following formula: Y(y) = c1p1(y) + c2p2(y) + c4p4(y) Where, Y(y) is the measured strip shape, c1, c2, and c4 are the strip shape pattern coefficients, and p1(y), p2(y), and p4(y) are the Legendre first, second, and fourth polynomials respectively, as shown in the following formula: p1(y) = y 3. A data-driven shape closed-loop control method for a cold rolling temper mill according to claim 1, characterized in that: In step S2, the cleaning, screening, and analysis of historical production data include: Historical production data, including strip width, strip inlet thickness, yield strength, inlet speed, outlet speed, target elongation, actual elongation, rolling force, inlet tension, outlet tension, roll tilt amount, intermediate roll bending force, and work roll bending force. Perform quartile range division on the historical production data and delete some moderate outliers and extreme outliers; Data analysis is based on the dataset after cleaning and screening, and use the Pearson correlation coefficient to analyze the correlation degree between variables.

4. A shape closed-loop control method for a cold rolling temper mill based on data driving according to claim 3, characterized in that: The calculation formula of the Pearson correlation coefficient in step S2 is: Among them, X i is a sample point, is the sample average value, and n is the number of samples.

5. A shape closed-loop control method for a cold rolling temper mill based on data driving according to claim 1, characterized in that: In step S2, the system identification leveler model includes: System identification uses the least squares method as the parameter fitting method and the discrete state space as the model. The input parameters of the model are strip inlet thickness, strip inlet speed, rolling force, outlet tension, roll tilt amount, intermediate roll bending force, and work roll bending force. The output parameters of the model are the strip shape pattern coefficients c1, c2, and c4; the expression of the discrete state space model is as follows: In the formula: x(k) represents the state quantity at time k; y(k) represents the output quantity at time k; A represents the system state matrix; B represents the system input matrix; C represents the system output matrix.

6. A data-driven closed-loop shape control method for a cold rolling temper mill according to claim 1, characterized in that: In step S3, model predictive control is based on the prediction of the future behavior of the model and determines the optimal input by solving the optimal solution of the loss function within a finite number of steps.

7. A data-driven shape closed-loop control method for a cold rolling temper mill according to claim 6, characterized in that: In step S3, at time k, model predictive control is divided into the following three steps: S3.

1. Estimate or measure the current system state x k ; S3.

2. Based on u k , u k+1 , u k+2 , …, u k+N Find the optimal solution of the loss function for optimal control, where u is the input at each moment and N is the prediction step length; The loss function is: Wherein, e is the error between the output value and the reference value; Q is the error weight coefficient; u is the input; R is the input weight coefficient; F is the weight coefficient of the terminal state quantity; (k+i|k) represents the quantity when predicting i steps forward at time k; S3.

3. Only execute the current u at time k k。 8. A data-driven closed-loop flatness control method for a cold rolling temper mill according to claim 1, characterized in that: The regulation of the shape control mechanism in step S4 includes roll tilt, intermediate roll bending, and work roll bending.

Images