A method for identifying the stiffness and stiffness difference of a rolling mill in a hot rolling production line based on the PSO algorithm

Abstract

The present invention provides a method for identifying the stiffness and stiffness difference of a rolling mill in a hot rolling production line based on the PSO algorithm, which relates to the technical field of rolling mill stiffness and includes the following steps: measuring the hot rolling production line by using a pressure measuring element that directly measures the rolling pressure to obtain the rolling force data on both sides of the rolling mill during the actual production process of the hot rolling production line; solving the dynamic response prediction model to obtain the loss function of the state space model on both sides of the rolling mill under the target rolling state; obtaining the stiffness on both sides of the rolling mill and the stiffness difference between both sides of the rolling mill under the target rolling state according to the parameter matrix to be estimated. This patent fully exploits the information contained in the actual vibration data and rolling data of the hot rolling unit to identify the stiffness and stiffness difference between the operating side and the drive side of the rolling mill, and on this basis, it can further analyze the influence degree of the strip on the stiffness and stiffness difference on both sides of the stand under different rolling widths and rolling speeds, etc., and dynamically track the changes in the stiffness and stiffness difference of the hot rolling unit.

Description

Technical Field

[0001] The present invention relates to the technical field of rolling mill stiffness, and more particularly, to a method for identifying the stiffness and stiffness difference of a hot rolling production line rolling mill based on the PSO algorithm. Background Art

[0002] For rolling mills, especially hot rolling mills, the stiffness and stiffness difference of the rolling mill are one of the important technical indicators for controlling the strip shape quality. However, in actual production, due to the manufacturing accuracy of the rolling mill housing and the influence of factors such as the need to connect an external drive system on the drive side of the rolling mill, the stiffness of the operating side and the working side of the rolling mill cannot be guaranteed to be exactly the same, that is, there is an asymmetry in the stiffness of the two sides of the rolling mill, and there is a certain stiffness difference. If the stiffness and stiffness difference of the rolling mill are within a certain range, the rolling mill stiffness is considered qualified. Otherwise, it is necessary to stop the machine for stiffness adjustment, such as replacing bearings and adjusting shims. The stiffness and stiffness difference of the rolling mill will directly affect the estimation accuracy of the springback equation in thickness control, and then affect the thickness control and strip shape quality of the strip. In order to improve the control accuracy of the strip thickness and ensure the good strip shape of the strip, the measurement accuracy of the rolling mill stiffness and stiffness difference should be improved as much as possible.

[0003] Currently, there are the following two methods for measuring the stiffness and stiffness difference of a rolling mill: The pressing method: Under the condition of ensuring a certain rolling speed of the rolls, perform pre-pressing so that the pre-pressing rolling force reaches a preset value, and then press against the rolls in a rolling force closed-loop mode at a set step size, record the rolling force and roll gap value at each step, and then perform regression analysis on the measured elastic curve to obtain the rolling mill stiffness. The plate rolling method: After the pre-pressing is completed, roll the plate according to the no-load roll gap set by the rolling schedule, measure the rolling force and the thickness of the rolled strip, repeat the experiment several times, perform linear regression on the obtained data, and calculate the estimated value of the rolling mill stiffness according to the springback equation.

[0004] However, the current methods for measuring the mill stiffness and stiffness difference have certain limitations. When using the press contact method, the two working rolls are in full roll surface contact. However, in the actual process of strip production, due to the different widths of the strip, the contact between the working roll and the strip is not full roll surface contact. Therefore, during rolling, the springback amount is greater than that during no-load press contact, and at the same time, it is impossible to evaluate the influence of different strip widths and different rolling speeds on the stiffness of both sides of the mill, and it cannot reflect the stiffness and stiffness difference characteristics under the dynamic behavior of the mill; for narrow-strip hot rolling mills, the rolling pressure is generally measured by measuring the hydraulic oil pressure with pressure sensors installed on both sides of the hydraulic cylinder piston, and the rolling force is indirectly calculated, rather than directly using the pressure measuring element for measuring the rolling pressure. However, there is a peak effect in the pressure inside the hydraulic cylinder, so only the rolling force measured at the static moment is the real rolling force, that is, there is a certain lag in the detection of the pressure sensor. Therefore, it is very difficult for this pressure sensor to detect the accurate rolling force. When using the rolling plate method, a suitable test plate needs to be prepared for a special test, which will take up production time and reduce production efficiency; secondly, due to the detection lag of the pressure sensor, the number of valid points of the rolling force under this method is small, and it may not be possible to relatively accurately regress the stiffness of the mill. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to propose a method for identifying the mill stiffness and stiffness difference in a hot rolling production line based on the PSO algorithm. This method deeply excavates the mill vibration data measured in the actual production process of the hot rolling production line and the rolling force data measured directly using the pressure measuring element for measuring the rolling pressure. It can reflect the stiffness and stiffness difference in the dynamic rolling process of the mill under different rolling states, improve the accuracy of obtaining the mill stiffness and stiffness difference, and at the same time, neither shutdown treatment nor special test experiments are required during the identification process, reducing the consumption of personnel and materials. To solve the technical problem that the existing methods for measuring the mill stiffness and stiffness difference cannot reflect the mill stiffness and stiffness difference under different rolling states.

[0006] The technical means adopted by the present invention are as follows:

[0007] A method for identifying the mill stiffness and stiffness difference in a hot rolling production line based on the PSO algorithm, the method comprising the following steps:

[0008] Measure the hot rolling production line using a pressure measuring element that directly measures the rolling pressure to obtain the rolling force data on both sides of the mill in the actual production process of the hot rolling production line;

[0009] Obtain the rolling force fluctuation function expression according to the rolling force data on both sides of the mill;

[0010] Measure the vibration data on both sides of the mill in the actual production process of the hot rolling production line using a vibration test platform;

[0011] Process the vibration data to obtain acceleration fluctuation data, velocity fluctuation data, and displacement fluctuation data;

[0012] Select the target rolling state, and filter the rolling force fluctuation function expression, acceleration fluctuation data, velocity fluctuation data, and displacement fluctuation data corresponding to the target rolling state according to time;

[0013] Establish a mathematical model of the dynamic system on both sides of the rolling mill according to the rolling force fluctuation function expression and the structural states on both sides of the rolling mill system;

[0014] Convert the mathematical model of the dynamic system into the state equation and control equation of the state space model according to the conversion principle of the state space model;

[0015] Solve the state equation and control equation, and obtain a dynamic response prediction model using the obtained parameter matrix;

[0016] Obtain the loss function of the state space model on both sides of the rolling mill under the target rolling state according to the acceleration fluctuation data, velocity fluctuation data, displacement fluctuation data, and the dynamic response prediction model;

[0017] Use the MATLAB software platform to build a virtual environment for the dynamic parameter identification model based on the particle swarm optimization algorithm;

[0018] Initialize the virtual environment, and optimize the loss function of the state space model on both sides of the rolling mill based on the PSO algorithm to obtain the parameter matrix to be estimated under the minimum loss function of the state space model on both sides of the rolling mill;

[0019] Obtain the stiffness, damping, and stiffness difference between both sides of the rolling mill under the target rolling state according to the parameter matrix to be estimated;

[0020] Furthermore, it also includes an optimization step, and the optimization step is as follows:

[0021] Divide the vibration data into three groups of data;

[0022] Identify the three groups of data respectively to obtain the stiffness, damping, and stiffness difference between both sides of the rolling mill for the three groups of data;

[0023] Average the stiffness, damping, and stiffness difference between both sides of the rolling mill for the three groups of data to obtain the optimized stiffness, damping, and stiffness difference data between both sides of the rolling mill.

[0024] Further, the expression of the rolling force fluctuation function obtained based on the rolling forces on both sides of the rolling mill includes: performing wavelet decomposition on the rolling force data, analyzing the detail components of the rolling force data to obtain the frequency of the rolling force fluctuation, and performing detrending processing on the rolling force data, analyzing the fluctuation amplitude of the detrended data to obtain the amplitude of the rolling force fluctuation.

[0025] Further, the processing of the vibration data includes wavelet decomposition, reconstruction, and denoising processing.

[0026] Further, the optimized stiffness, damping, and the stiffness difference data between both sides of the rolling mill include the dynamic rolling stiffness of the operating side and the drive side of the rolling mill and the stiffness difference between both sides in cases such as rolling strip materials with the same rolling speed, the same thickness, and different strip widths, and rolling strip materials with the same material, the same thickness, and the same strip width at different rolling speeds.

[0027] Further, the steps for building the virtual environment include:

[0028] Setting initialization parameters, where the initialization parameters include the number of particle swarm parameters, the particle population size, the maximum number of iterations, the defined range of the parameters to be identified, the defined range of the particle optimization speed, and the weight learning factor;

[0029] Setting the initial position and speed;

[0030] Setting the update methods for the particle position and speed;

[0031] Setting the methods for setting the local optimum and the global optimum.

[0032] Further, the loss function is the sum of the squares of the differences between the predicted acceleration and the actual acceleration of each data node, the sum of the squares of the differences between the predicted speed and the actual speed of each data node, and the sum of the squares of the differences between the predicted displacement and the actual displacement of each data node, and the loss function is obtained based on the relative difference of the ideal solution.

[0033] Further, the relative difference of the ideal solution is:

[0034]

[0035] where V is the relative difference between the objective function and the ideal solution, is the extreme value of each objective design criterion, n is the number of objective functions, and Q i is the actual value of each objective loss function.

[0036] Compared with the prior art, the present invention has the following advantages:

[0037] Compared with the traditional pressing method, the measurement state of the present invention is the actual production state. By further mining the existing data on the rolling site, the utilization degree of the data is improved, and the measured stiffness can more truly reflect the actual stiffness on both sides of the rolling mill under different rolling states; during the rolling process of strip materials, the pressure measuring elements for measuring the rolling pressure are directly used, which also avoids the hysteresis of the old-fashioned pressure sensor detection and more accurately reflects the actual rolling force on site.

[0038] Compared with the traditional plate rolling method, the present invention does not require extra preparation of suitable strip materials for experiments. At the same time, the effective points during the rolling process are increased, and the accuracy of the rolling force is improved. At the same time, compared with the above methods, the hot rolling production line of the present invention does not need to stop production, nor does it need to carry out special test experiments, reducing material consumption and personnel consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0040] Figure 1 It is a flowchart for identifying the stiffness and stiffness difference of the present invention.

[0041] Figure 2 It is a wavelet decomposition diagram of the rolling force of the present invention.

[0042] Figure 3 It is a vibration displacement diagram with the same frequency as the rolling force fluctuation obtained by wavelet decomposition interception of the present invention.

[0043] Figure 4 It is a diagram of the identification result of the present invention.

[0044] Figure 5 It is a time-domain comparison diagram of the simulation response and the original response of the local identification result of the present invention.

[0045] Figure 6 It is a coherence function diagram of the simulation response and the original response of the identification result of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] In order to enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0047] It should be noted that the terms "first", "second", etc. in the description, claims and above-mentioned drawings of the present invention are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0048] As Figure 1 shown, the present invention provides a method for identifying the stiffness and stiffness difference of a rolling mill in a hot rolling production line based on the PSO algorithm. The method includes the following steps:

[0049] Step 1: Use a pressure measuring element that directly measures the rolling pressure to measure the rolling forces on the operator side and the drive side of the rolling mill during the actual production process of the hot rolling production line, and process the rolling forces on the operator side and the drive side respectively to obtain the function expressions of the rolling force fluctuations on the operator side and the drive side of the rolling mill during the actual production process of the hot rolling production line.

[0050] Select the rolling force fluctuation data on the operator side of a certain rolling mill when rolling a certain slab in a certain hot rolling production line of a certain steel plant, and perform a fast Fourier transform on the rolling force fluctuation data on the operator side of the rolling mill to obtain the frequency of the rolling force on the operator side of the rolling mill. Due to the good control of the rolling force of the rolling mill at the hot rolling site, the frequency of the rolling force of the rolling mill is approximately 0, but there is still a fluctuation phenomenon in the rolling force curve on the operator side of the rolling mill. Therefore, it is considered that the frequency of the rolling force fluctuation is the frequency corresponding to the detailed component of the rolling force on the operator side of the rolling mill.

[0051] As Figure 2 shown, use a db2 wavelet filter to decompose the signal of the rolling force on the operator side of the rolling mill. Wavelet decomposition of the signal means using a wavelet filter to decompose the original signal into an approximate signal and a detailed signal. Perform a fast Fourier transform on the first-layer detailed signal to obtain the frequency w of the rolling force fluctuation on the operator side of the rolling mill; perform a detrending process on the rolling force on the operator side of the rolling mill to obtain the amplitude F of the rolling force fluctuation on the operator side of the rolling mill. Further, write the rolling force fluctuation function expression as shown in Equation (1):

[0052] ΔP = F × sin(2π × w × t) ······· (1)

[0053] In the formula: ΔP is the rolling force fluctuation value, and t is the sampling time;

[0054] Step 2: Use the vibration test platform to measure the vibration data of the operator side and the drive side of the rolling mill during the actual production process of the hot rolling production line (the sensor we use is an acceleration sensor, and the obtained vibration data is acceleration data). Integrate this data to obtain the acceleration fluctuation data, speed fluctuation data, and displacement fluctuation data of the operator side and the drive side of the rolling mill during the actual production process of the hot rolling production line;

[0055] Perform filtering and noise reduction on the vibration data of the operator side of the rolling mill, as well as the speed fluctuation data and displacement fluctuation data after integration processing. Considering that the frequency of the rolling force fluctuation during the rolling of this slab is w, perform wavelet decomposition on the vibration data of the operator side of the rolling mill, as well as the speed fluctuation data and displacement fluctuation data after integration processing, as Figure 3 shown, decompose the vibration data of the operator side of the rolling mill with the data fluctuation frequency around w, as well as the speed fluctuation data and displacement fluctuation data with the data fluctuation frequency around w;

[0056] Step 3: Screen the vibration data after processing on the operator side and the rolling force fluctuation function expression during the rolling of the same slab, and the vibration data after processing on the drive side and the rolling force fluctuation function expression;

[0057] Step 4: Establish a mathematical model of the dynamic system for the operator side and the drive side of the rolling mill respectively. Convert the mathematical models of the dynamic systems of the operator side and the drive side of the rolling mill into corresponding state equations and control equations according to the conversion principle of the state space model method to characterize the correlation between the model input data and output data. Taking the operator side of the rolling mill as an example, it specifically includes:

[0058] a. Considering that the roll in direct contact with the strip is the work roll, and the vibration of the work roll has the most direct impact on the strip, establish the mathematical models of the dynamic systems of the upper work roll and the lower work roll on the operator side of the rolling mill as shown in Equations (2) and (3):

[0059]

[0060] In Equations (2) and (3): m OS上 is the mass of the upper work roll on the operator side of the rolling mill, is the vibration acceleration of the upper work roll on the operator side of the rolling mill, c OS上 is the damping of the upper work roll on the operator side of the rolling mill, is the vibration speed of the upper work roll on the operator side of the rolling mill, k OS上 is the stiffness of the upper work roll on the operator side of the rolling mill, x OS上 is the vibration displacement of the upper work roll on the operator side of the rolling mill, ΔP OS is the rolling force fluctuation on the operator side of the rolling mill, m OS下 is the mass of the lower work roll on the operator side of the rolling mill, is the vibration acceleration of the lower working roll on the operator side of the rolling mill, c OS下 is the damping of the lower working roll on the operator side of the rolling mill, is the vibration velocity of the lower working roll on the operator side of the rolling mill, k OS下 is the stiffness of the lower working roll on the operator side of the rolling mill, x OS下 is the vibration displacement of the lower working roll on the operator side of the rolling mill;

[0061] The identification method of the stiffness k of the upper working roll on the operator side of the rolling mill OS上 is the same as that of the stiffness k of the lower working roll on the operator side of the rolling mill OS下 Here, only the identification method of the stiffness k of the upper working roll on the operator side of the rolling mill OS上 is introduced. Finally, the stiffness k of the operator side of the rolling mill OS is shown as in Equation (4):

[0062]

[0063] b. Establish the state - space model as shown in Equation (5):

[0064]

[0065] In the formula: x(t) is the state variable of the state - space model, which links the model input and output; u(t) is the input of the rolling force fluctuation on the operator side of the rolling mill at time t, N; y(t) is the measured output of the acceleration (or velocity, displacement) on the operator side of the rolling mill at time t, m / s 2 (or m / s, m); e 1 (t) and e 2 (t) are the model error values at time t; A, B, C, D are the parameter matrices to be estimated of the state model;

[0066] Furthermore, the parameter matrices A, B, C, D to be estimated of the model are obtained through the subspace algorithm, and their expressions are shown as in Equation (6):

[0067]

[0068] Step 5: According to the state equations and control equations established for the operator side and the drive side of the rolling mill, solve the dynamic response under the measured rolling force (this response can be the displacement, velocity or acceleration data of the dynamic model under this rolling force, and the specific situation is related to the parameter matrix C. If C = [1 0], the dynamic response is the displacement fluctuation; if C = [0 1], the dynamic response is the velocity fluctuation; if C = [0 1] and differential processing is performed, the dynamic response is the acceleration fluctuation) prediction model, and solve the loss function of the state - space models of the operator side and the drive side of the rolling mill under this rolling state, specifically including:

[0069] Obtain the expression of the transfer function H of the input u and the output y in the measured state space model, as shown in Equation (7):

[0070] H = C(ωI - A) -1 B + D ········· (7)

[0071] Where: ω is the spatial frequency, m -1 ; I is the identity matrix;

[0072] Furthermore, after obtaining the transfer function of the state space model, the dynamic response output can be predicted by using the rolling force fluctuation input u(t) on the operating side of the rolling mill The calculation formula of the prediction model is as shown in Equation (8):

[0073]

[0074] Where: is the predicted dynamic response output, is the estimated transfer function of the state space model;

[0075] Furthermore, define the loss function of the state space model, and minimize the loss function of the state space model to obtain the parameter matrices A, B, C, D to be estimated of the state space model, and obtain the stiffness on the operating side of the rolling mill. The expression of the loss function of the state space model is as shown in Equation (9):

[0076]

[0077] Where, N is the total number of data samples participating in the calculation of the loss function;

[0078] Furthermore, there are three objective loss functions of acceleration, velocity and displacement in the system. However, when one objective loss function is optimal, the other two objective loss functions are not necessarily optimal. Therefore, different objective loss functions may obtain different optimal solutions, and the operator needs to select the objective loss function most suitable for the situation of this production line to calculate the optimal solution, or obtain the relative optimal solution according to the relative difference of the ideal solution;

[0079] Step Six: Use the MATLAB software platform to build a virtual environment for the dynamic parameter identification model based on the particle swarm optimization algorithm (hereinafter simplified as the PSO algorithm), specifically including:

[0080] a. Initialize the parameters. The number of particle swarm parameters (L), the particle population size (M), the maximum number of iterations (T), the defined range of the parameters to be identified [X min X max , the defined range of the particle optimization speed [v min v max , the weight learning factor c 1and c 2 ;

[0081] b. The weight coefficient has a great influence on the PSO algorithm. When the objective values of each particle tend to be the same or tend to be locally optimal, the inertia weight coefficient is increased to enable the algorithm to jump out of the local optimum and perform global search. When the objective values of each particle are more dispersed, the inertia weight coefficient is decreased to enable the algorithm to perform local search. The expression of the adaptive weight coefficient ω is shown in Equation (10):

[0082]

[0083] where: ω is the inertia weight coefficient, ω min is the minimum value of the inertia weight coefficient, ω max is the maximum value of the inertia weight coefficient, is the objective value of the current particle, Obj max is the maximum value of the particles among all current particles, is the average value of the particles of all current particles, s T is the current iteration step, T is the number of maximum iterations;

[0084] c. Randomly initialize the positions and velocities of the parameters of each particle swarm, and use this position as the initial local optimal position and global optimal position. The random initialization formula for the position is shown in Equation (11), and the random initialization formula for the velocity is shown in Equation (12):

[0085] X = X min +(X max -X min ) × rand ········· (11)

[0086] v = v min +(v max -v min ) × rand ········· (12)

[0087] where: rand is a random number between 0 and 1, X is the initialized population particle, and v is the initialized population particle velocity;

[0088] d. In order to enhance the global search ability of the algorithm in the initial stage and the ability to quickly converge to the global optimum in the later stage, the weight learning factor is optimized and updated as follows. The specific implementation method is shown in Equation (13):

[0089]

[0090] where: is the optimized value of the weight learning factor, c 1,f is c 1 the final iteration value of c2,f is the iteration end value for c 2 , and s T is the number of iteration steps;

[0091] e. Evaluate the fitness of each particle in each iteration. Compare the fitness value of each particle with the local optimal position it has passed through. If it is better, then take the particle position as the current local optimal position, and then compare it with the global optimal position it has passed through. If it is better, then take it as the current global optimal position. Considering that when the particle velocity is small, the particle will wander too much within a local range, and when the particle velocity is large, it may cause the particle to converge to the local minimum prematurely. Therefore, to improve the above problems, the optimization speed of the particle is updated according to the formula to ensure the efficiency and accuracy of the PSO algorithm. The specific formula is shown in Equation (14):

[0092]

[0093] In the formula: is the velocity value after the particle updates its velocity, X B is the current optimal individual, and BestS is the global optimal individual;

[0094] f. Update the position of the particle according to the velocity update formula. The specific formula is shown in Equation (15):

[0095]

[0096] In the formula: is the position value after the particle updates its position;

[0097] g. Iterate step by step until the loss function Q of the state space model reaches the minimum value and converges, or the number of iteration steps reaches the set maximum number of iteration steps;

[0098] Step 7: Initialize the virtual environment. Optimize the loss function of the state space models established for the operating side and the drive side of the rolling mill based on the PSO algorithm, and respectively obtain the parameter matrices A, B, C, D to be estimated under the minimum loss functions of the state space models of the operating side and the drive side of the rolling mill. Further, obtain the dynamic parameters on both sides of the rolling mill, that is, the stiffness, damping, etc. of the operating side and the drive side of the rolling mill. Further, the stiffness difference between the two sides of the rolling mill can be obtained through calculation. The specific formula is shown in Equation (16):

[0099] Δk = |k OS - k DS | ········· (16)

[0100] In the formula: Δk is the stiffness difference between the two sides of the rolling mill, k OS is the stiffness of the operating side of the rolling mill, and k DS is the stiffness of the drive side of the rolling mill;

[0101] Step 8: Due to various factors, uncontrollable errors may occur in the identification of the stiffness and stiffness difference on both sides of the rolling mill in the hot rolling production line based on the PSO algorithm. We divide the vibration data of rolling the same slab into three groups, perform identification on each group of data respectively, and average the results of the three groups of data to obtain a relatively accurate stiffness of the operator side and the drive side of the rolling mill and the stiffness difference between the two sides. The specific formula is shown in Equation (17). Taking the operator side of the rolling mill as an example, the identification result is as Figure 4 shown. By simulating the response of the identification result, the error comparison between the simulated response and the measured vibration response in the time domain and frequency domain can be obtained, as shown in Figure 5 and Figure 6 shown;

[0102]

[0103] In the formula: k OS1 , k OS2 , k OS3 are the stiffnesses of the operator side of the rolling mill obtained by dividing the vibration data of one slab into three groups and performing identification respectively. k DS1 , k DS2 , k DS3 are the stiffnesses of the drive side of the rolling mill obtained by dividing the vibration data of the same slab into three groups and performing identification respectively;

[0104] Furthermore, organize and summarize the rolling performance, and identify and analyze the dynamic rolling stiffness of the operator side and the drive side of the rolling mill and the stiffness difference between the two sides in cases such as rolling strip with the same material, the same thickness, and different widths at the same rolling speed, and rolling strip with the same material, the same thickness, and the same width at different rolling speeds;

[0105] Furthermore, analyze the identification situation when the rolling mill rolls slabs prone to vibration and there are quality defects in the strip products, so as to summarize a threshold of stiffness and stiffness difference for the rolling process of the production line, providing a reasonable support basis for the adjustment of the rolling mill equipment on the production line.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying the stiffness and stiffness difference of a rolling mill in a hot rolling production line based on the PSO algorithm, characterized in that, the method comprises the following steps: Measuring the hot rolling production line by using a pressure measuring element that directly measures the rolling pressure to obtain the rolling force data on both sides of the rolling mill during the actual production process of the hot rolling production line; Obtaining an expression for the rolling force fluctuation function based on the rolling force data on both sides of the rolling mill; Measuring the vibration data on both sides of the rolling mill during the actual production process of the hot rolling production line by using a vibration test platform; Processing the vibration data to obtain acceleration fluctuation data, velocity fluctuation data, and displacement fluctuation data; Selecting a target rolling state, and screening the rolling force fluctuation function expression, acceleration fluctuation data, velocity fluctuation data, and displacement fluctuation data corresponding to the target rolling state according to time; Establishing a mathematical model of the dynamic system on both sides of the rolling mill based on the rolling force fluctuation function expression and the structural states on both sides of the rolling mill system; Converting the mathematical model of the dynamic system into the state equation and control equation of the state space model according to the conversion principle of the state space model; Solving the state equation and control equation, and obtaining a dynamic response prediction model by using the obtained parameter matrix; Based on the acceleration fluctuation data, velocity fluctuation data, displacement fluctuation data, and the dynamic response prediction model, obtaining the loss function of the state space model on both sides of the rolling mill under the target rolling state; Using the MATLAB software platform to build a virtual environment for a dynamic parameter identification model based on the particle swarm optimization algorithm; Initializing the virtual environment, and optimizing the loss function of the state space model on both sides of the rolling mill based on the PSO algorithm to obtain the parameter matrix to be estimated under the minimum loss function of the state space model on both sides of the rolling mill; Obtaining the stiffness, damping, and stiffness difference between both sides of the rolling mill under the target rolling state according to the parameter matrix to be estimated.

2. The method for identifying the stiffness and stiffness difference of a rolling mill in a hot rolling production line based on the PSO algorithm according to claim 1, characterized in that, it further comprises an optimization step, and the optimization step is: Dividing the vibration data into three groups of data; Identifying the three groups of data respectively to obtain the stiffness, damping, and stiffness difference between both sides of the rolling mill for the three groups of data; Performing an averaging process on the stiffness, damping, and stiffness difference between both sides of the rolling mill for the three groups of data to obtain optimized stiffness, damping, and stiffness difference data between both sides of the rolling mill.

3. The method for identifying the stiffness and stiffness difference of a rolling mill in a hot rolling production line based on the PSO algorithm according to claim 1, characterized in that, the obtaining of the rolling force fluctuation function expression based on the rolling force on both sides of the rolling mill includes: performing wavelet decomposition on the rolling force data, analyzing the detailed components of the rolling force data to obtain the frequency of the rolling force fluctuation, and by performing a detrending process on the rolling force data, analyzing the fluctuation amplitude of the detrended data to obtain the amplitude of the rolling force fluctuation.

4. The method for identifying the stiffness and stiffness difference of a rolling mill in a hot rolling production line based on the PSO algorithm according to claim 1, characterized in that, processing the vibration data includes wavelet decomposition, reconstruction, and denoising processing.

5. The method for identifying the stiffness and stiffness difference of a rolling mill in a hot rolling production line based on the PSO algorithm according to claim 2, It is characterized in that The optimized stiffness, damping and stiffness difference data on both sides of the rolling mill include the dynamic rolling stiffness on the operating side and the drive side of the rolling mill and the stiffness difference between the two sides in cases such as rolling strip with the same material, the same thickness and different strip widths at the same rolling speed, and rolling strip with the same material, the same thickness and the same strip width at different rolling speeds.

6. The method for identifying the stiffness and stiffness difference of a hot rolling line rolling mill based on the PSO algorithm according to claim 1 It is characterized in that The steps for building the virtual environment include Setting initialization parameters, which include the number of particle swarm parameters, the particle population size, the maximum number of iterations, the defined range of parameters to be identified, the defined range of particle optimization speed and the weight learning factor Setting the initial position and speed Setting the update method of particle position and speed Setting the setting method of local optimum and global optimum 7. The method for identifying the stiffness and stiffness difference of a hot rolling line rolling mill based on the PSO algorithm according to claim 1 It is characterized in that The loss function is the sum of the squares of the differences between the predicted acceleration and the actual acceleration of each data node, the sum of the squares of the differences between the predicted speed and the actual speed of each data node, or the sum of the squares of the differences between the predicted displacement and the actual displacement of each data node, and the loss function is obtained based on the relative difference of the ideal solution 8. The method for identifying the stiffness and stiffness difference of a hot rolling line rolling mill based on the PSO algorithm according to claim 7 It is characterized in that The relative difference of the ideal solution is Among them, V is the relative difference between the objective function and the ideal solution, is the extreme value of each objective design criterion, n is the number of objective functions, and Q i is the actual value of each objective loss function.

Images