A dynamic optimization method for spread model coefficients of hot roughing mill

Abstract

The present application relates to the field of metallurgical rolling automation control, and discloses a dynamic optimization method for a hot continuous rolling rough rolling mill width expansion model coefficient, and the specific steps are as follows: a rough rolling mill width expansion mechanism model is constructed, the rough rolling mill width expansion mechanism model coefficient to be optimized and constraint conditions are determined; multi-source key process data are collected and preprocessed to construct a complete data set; a rough rolling mill width expansion mechanism model coefficient optimization objective function is constructed; the optimal solution of the optimization objective function is solved; the coefficient of the optimal rough rolling mill width expansion mechanism model is updated to the rough rolling mill control system in real time, the rough rolling mill width expansion mechanism model precision is verified by using actual production data, if the width deviation does not meet the requirement, the latest data are collected again for calculation, and the rough rolling mill width expansion mechanism model coefficient is adaptively and dynamically optimized. The present application can track the rolling condition change in real time, dynamically adjust the width expansion model coefficient, significantly improve the width expansion amount prediction precision, reduce the strip width deviation, and improve the yield rate and production stability.

Description

Technical Field

[0001] This invention relates to the field of automated control of metallurgical rolling, and specifically to a dynamic optimization method for the width expansion model coefficient of a hot continuous rolling roughing mill. Background Technology

[0002] In the hot strip mill roughing process, the accuracy of strip width prediction directly affects the quality of strip width control, yield, and production stability, making it a key technical challenge in the field of hot strip mill automatic control. Existing strip width prediction models for hot strip mill roughing mills generally employ fixed coefficient settings, failing to track dynamic changes in operating conditions during rolling in real time, such as steel grade switching, temperature fluctuations, vertical roll side pressure adjustments, and horizontal reduction changes. This results in significant prediction errors in the strip width prediction models, making it difficult to meet the demands of high-precision production. Furthermore, existing technologies often employ single-mechanism modeling or data-driven approaches to construct strip width models, which have limitations: single-mechanism modeling struggles to adapt to complex and changing production conditions, exhibiting poor adaptability and robustness; single-data-driven modeling lacks physical rationale support, resulting in insufficient model reliability.

[0003] Furthermore, traditional optimization algorithms suffer from slow convergence speed, premature convergence, and insufficient optimization accuracy when used for strip width model coefficient optimization. Moreover, most optimization methods are offline calibration modes, which cannot achieve real-time dynamic optimization, further restricting the improvement of strip width prediction accuracy. This results in strip width deviation, increased production losses, and low yield, making it difficult to adapt to the diverse production needs of various hot continuous rolling mills. Therefore, there is an urgent need for a technical solution that can dynamically optimize strip width model coefficients and improve prediction accuracy. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a dynamic optimization method for the width spread model coefficient of a hot strip mill roughing mill. This method solves the technical problems of fixed width spread model coefficients, inability to adapt to dynamic changes in working conditions, low prediction accuracy, and large strip width deviations in existing technologies. It achieves real-time dynamic optimization of the width spread model coefficient, thereby improving width control accuracy and yield.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A dynamic optimization method for the width expansion model coefficient of a hot strip roughing mill, comprising the following steps: Step 1: Construct a roughing mill width development mechanism model including a dog-bone recovery width development model and a horizontal rolling natural width development model. The expression for the dog-bone recovery width development model is as follows:

[0006] in, Correction factor for dog bone restoration; This is the vertical roller trough shape correction factor, with a value of 1.0; and Dimensionless intermediate variables characterizing the geometric features of the deformation region:

[0007] in, For the inlet thickness; The radius of the vertical roller; The slope coefficient is set to 1.0 to restore the slope. This refers to the side pressure of the vertical roller; The expression for the horizontal rolling natural width expansion model is:

[0008] in, This is a horizontal width correction factor; This refers to the width of the intermediate billet after vertical roll rolling; This refers to the reduction amount per horizontal pass; The radius of the horizontal working roller; Sedlaczek broadening factor; For the inlet thickness; Step 2: Using the dog bone recovery correction coefficient in the roughing mill width expansion mechanism model. and horizontal width correction factor To optimize the variables, the dog bone recovery correction coefficient is set in the roughing mill width expansion mechanism model. and the horizontal width correction coefficient Constraints:

[0009] in, , These are the lower and upper bounds of the dog bone restoration correction coefficient, respectively. , These are the lower and upper bounds of the horizontal width correction coefficient, respectively. Step 3: Collect multi-source process data in real time during the hot continuous rolling roughing process, and perform outlier removal processing on the collected multi-source process data to form a complete dataset; Step 4: Based on the multi-source process data after removing outliers, construct the dog bone recovery correction coefficient in the roughing mill width extension mechanism model. and horizontal width correction factor The optimization objective function; Step 5: Solve the optimization objective function to obtain the optimal dog bone recovery correction coefficient under the current rolling conditions. and horizontal width correction factor ; Step Six: Apply the optimal dog bone restoration correction coefficient obtained in Step Five. and horizontal width correction factor The model is updated in real time to reflect the width expansion mechanism of the roughing mill. Actual production data is used to verify the accuracy of the model. If the width expansion deviation does not meet the preset value, steps three through five are repeated to restore the correction coefficients of the dog-bone structure in the roughing mill width expansion mechanism model. and horizontal width correction factor Perform adaptive dynamic optimization.

[0010] Furthermore, in step three, multi-source process data are collected in real time during the hot continuous rolling roughing process. The multi-source process data includes: roughing mill inlet thickness, roughing mill inlet width, vertical roll side pressure, horizontal reduction, vertical roll diameter, horizontal work roll diameter, and measured roughing mill outlet width.

[0011] Furthermore, in step three, outlier removal processing is performed on the collected multi-source process data to form a complete dataset. The specific method for outlier removal processing is as follows: The 3σ criterion is used to remove outliers from the acquired data, as shown in the following formula: ,

[0012] in, This is the average value; Standard deviation These are measured values; Furthermore, the dog bone restoration correction coefficient in step four... and horizontal width correction factor The optimization objective function is to minimize the sum of squared deviations between the predicted value and the actual width expansion of the roughing mill, as expressed by:

[0013] in, N This represents the total number of strip steel samples in the dataset. For the first The measured width of the strip steel at the roughing mill exit, in mm; The predicted width value is in mm.

[0014] Furthermore, the dog-bone recovery correction coefficient of the roughing mill width expansion mechanism model in step six... and horizontal width correction factor The adaptive dynamic optimization process is as follows: The optimized dog bone restoration correction coefficient and horizontal width correction factor Substitute the values ​​into the roughing mill width expansion mechanism model to calculate the deviation between the predicted width and the actual width. like Then output the optimized dog bone restoration correction coefficient. and horizontal width correction factor As coefficients in the current roughing mill width expansion mechanism model. The preset allowable threshold for width deviation; like Then, re-collect the latest multi-source process data from the roughing production, and repeat the optimization process until... , This is the preset allowable threshold for width deviation.

[0015] The beneficial effects of this invention are as follows: The dynamic optimization method for the width spread model coefficients of hot strip mill roughing mills disclosed in this invention belongs to the field of automatic control technology for hot strip mills, effectively making up for the shortcomings of existing technologies. Its core advantage is that by dynamically adjusting the coefficients of the roughing mill width spread mechanism model, it solves the pain point of existing technologies where coefficients are fixed and cannot adapt to fluctuations in operating conditions. It can track changes in rolling conditions in real time, significantly improving the accuracy of width spread prediction. Verified in a 1780mm hot strip mill roughing mill, the average prediction deviation of the width spread model decreased from 6.59mm before optimization to 2.85mm, far below the preset allowable deviation (±4mm). The hit rate of silicon steel width increased from 63.85% to 80.96%, significantly reducing production losses. This invention integrates the advantages of mechanism modeling and data-driven approaches, taking into account the physical rationality and adaptability of the model, and is suitable for various roughing mills and different operating conditions. An improved LEE algorithm is used to solve for the optimal coefficients, achieving fast and stable convergence and solving the drawbacks of traditional algorithms. This method requires no major equipment modifications, does not affect production, significantly improves yield and production stability, creates economic benefits for enterprises, and has strong practicality and good engineering application value. Attached Figure Description

[0016] Appendix Figure 1 This is a flowchart of the dynamic optimization method for the width expansion model coefficient of a hot continuous rolling mill according to the present invention; Appendix Figure 2 This is a schematic diagram of the process for optimizing the coefficients of the broadened model provided in an embodiment of the present invention; Appendix Figure 3 A schematic diagram of the optimization iteration curve of the improved LEE algorithm provided in this embodiment of the invention; Appendix Figure 4 This is a schematic diagram showing the comparison of application effects before and after the optimization of the expansion model coefficients provided in an embodiment of the present invention. Detailed Implementation

[0017] The present invention will now be further described with reference to the accompanying drawings and specific embodiments: This embodiment uses a 1780mm hot strip roughing mill as the application object, and employs the dynamic optimization method of this invention to optimize the width spread model coefficients, such as... Figure 1 The diagram shown is a flowchart illustrating a specific implementation of the present invention, which includes the following steps: S1. Construct a mechanism model for the width expansion of the roughing mill and determine the coefficients to be optimized and the constraints. S11, the Exit width after horizontal rolling The mathematical expression is:

[0018] in, The amount of width generated during the recovery of the dog's bones; This refers to the natural width spread resulting from horizontal rolling. This refers to the width of the slab after vertical roll rolling; S12, in the roughing control system, the dog-bone recovery width expansion model uses a standard simplified model based on the geometric characteristics of the deformation zone width-to-thickness ratio and contact arc length to predict the lateral flow behavior of metal, and its expression is as follows:

[0019] in, For the dog bone restoration correction factor, This is the vertical roller trough shape correction factor (with a value of 1.0). and Dimensionless intermediate variables characterizing the geometric features of the deformation region:

[0020] in, For the inlet thickness; The radius of the vertical roller; To restore the slope coefficient (with a value of 1.0); This refers to the side pressure of the vertical roller; S13, the horizontal rolling natural width spread model adopts the Sedlaczek formula based on the law of minimum resistance and the assumption of an ideal rigid-plastic body, and introduces a horizontal width spread correction factor. Perform a strong correction, as shown in the following formula:

[0021] in, This refers to the width of the intermediate billet after vertical roll rolling; This refers to the reduction amount per horizontal pass; The radius of the horizontal working roller; Sedlaczek broadening factor; For the inlet thickness; The dog bone recovery correction coefficient in the roughing mill width expansion mechanism model and horizontal width correction factor To optimize the variables, the constraints are set as follows: , .

[0022] S2. Real-time acquisition and preprocessing of multi-source process data during hot continuous rolling roughing process; Specifically, the rolling data obtained in this embodiment include the following process parameters: entry thickness (57-234mm), entry width (1031-1375mm), vertical roll side pressure (0.1-64mm), horizontal reduction (18-45), vertical roll diameter (1159-1200mm), horizontal work roll diameter (1142-1282mm), and measured width of roughing mill exit (1035-1372mm). Specifically, in this embodiment, the S1 data processing described above uses the 3σ criterion to remove outliers from the acquired data, as shown in the following formula: ,

[0023] in, This is the average value; Standard deviation; These are measured values; After the above preprocessing process, this embodiment finally obtains 3,000 complete and valid rolling sample data, providing reliable data support for subsequent coefficient optimization.

[0024] S3. Construct a broadened model coefficient optimization objective function; Specifically, the objective function is constructed with the goal of minimizing the root mean square error (RMSE) between the predicted and measured values ​​of the extended model. The constraint condition is the range of the coefficients to be optimized set in step S1, ensuring that the optimization result meets the actual working condition requirements. The objective function expression is optimized as follows:

[0025] in, N This represents the total number of strip steel samples in the dataset. For the first The measured width of the strip steel at the roughing mill exit, in mm; The predicted width value is in mm.

[0026] S4. The improved LEE algorithm is used to optimize the objective function and find the optimal dog bone restoration correction coefficient. and horizontal width correction factor ;like Figure 2The diagram shown is a schematic of the process for optimizing the coefficients of the broadening model using the improved LEE algorithm provided in an embodiment of the present invention. S41, the dog-bone recovery correction coefficient in the roughing mill width extension mechanism model. and horizontal width correction factor The position of each particle in the population is randomly initialized within the constrained space, and each individual in the population is represented as... , representing the solution to an optimization problem in a 2-dimensional space, with the initial position of the particle defined by the following formula:

[0027] in, and Indicate the question number j The lower and upper bounds of a dimension, rand This represents a random number within the interval [0,1]. S42 uses the optimization objective function as the fitness index to select candidate coefficient values ​​carried by the particles. Substituting the model of the roughing mill width expansion mechanism, the predicted width is calculated and the fitness value of each particle is obtained. The particle with the lowest fitness is selected and its position is marked as the global optimal solution. And record the corresponding minimum error value. ; S43 achieves optimization by dynamically adjusting the speed and position of particles, setting relevant parameters to balance global search and local optimization capabilities, and ensuring optimization efficiency and accuracy.

[0028] in, For quality; For damping; For random disturbance terms; a For acceleration; The damping coefficient; and These are two random solutions for the current position; The current temperature; Initial temperature; Current speed; Current position; For the individual optimal solution; This is the globally optimal solution; For dynamic inertia weights; This is the acceleration adjustment coefficient; it This represents the number of iterations. MaxIt This represents the maximum number of iterations. These are the weighting coefficients; S44. To prevent the population from converging prematurely to a local optimum and to enhance the depth of solution space exploration, a novel diversity promotion operator (DP) is introduced. This operator integrates the solution generated by the LEE search mechanism with the current solution whose position has been adjusted by the step size, which can effectively guarantee population diversity and solution space exploration capability. Its expression is as follows:

[0029] in, This represents the final updated position; σ is a random value. Step size; UL The coefficient of difference between the upper and lower bounds; The optimal solution found; It is a constant positive number (with a value of 0.3); These are random numbers distributed according to a standard normal distribution. Current position; S45, the Local Escape Operator (LEO) is a key enhancement operator for improving the LEE algorithm. Based on the random fluctuation principle of the Langevin equation, it guides candidate solutions to escape local optima in a controlled manner, avoiding premature convergence and accelerating the optimization convergence speed. Its specific formula is as follows:

[0030] in, This represents the probability (ranging from 0.01 to 0.05). The escape intensity coefficient (range 0.05~0.1); S46. If the number of consecutive iterations is ≥100 or the change in fitness value is <ε (ε is the preset convergence threshold), then terminate the iteration and output the optimal broadened model coefficients; if not, return to step S43 to continue the iteration.

[0031] Specifically, based on the results of multiple preliminary experiments, the key control parameters of the LEE algorithm are set as follows: population size. Maximum number of iterations Initial system temperature Initial damping coefficient Acceleration adjustment coefficient Weighting coefficient ; fundamental probability constant Preset convergence threshold .

[0032] Under the above parameter settings, the improved LEE algorithm exhibits high optimization performance, reaching a stable convergence state only in the 11th iteration, as shown below. Figure 3As shown, after the algorithm converges, the root mean square error (RMSE) of the extended model decreases from 7.00 before optimization to 3.80, a reduction of 45.7%, significantly improving the prediction accuracy of the extended model. The model parameters and performance indicators before and after optimization are shown in Table 1. Table 1 Model parameters and performance indicators

[0033] S5. Real-time updates of the mill system width expansion coefficient and continuous dynamic optimization; The optimal coefficients [0.35, 0.47] obtained through optimization are updated in real time to the spread model of the roughing mill control system, completing one optimization. Then, return to step S2, continuously collect new rolling process data, and repeat steps S2-S5 to achieve continuous dynamic optimization of the spread model coefficients.

[0034] Application results show that the prediction accuracy of the strip width prediction model is significantly improved after adopting the dynamic optimization method of this invention. The average prediction deviation is only 2.85 mm, far less than the preset allowable deviation (±4 mm), fully meeting the stringent requirements for strip width prediction accuracy in hot strip roughing production. This effectively reduces the problem of strip width deviation caused by excessive strip width prediction errors. Simultaneously, in actual production applications, the accuracy rate of silicon steel width prediction is significantly improved, increasing from 63.85% before optimization to 80.96%, significantly reducing the generation of strips with unqualified widths, reducing production losses, and fully verifying the practicality and superiority of the method of this invention. Figure 4 This is a schematic diagram showing the comparison of application effects before and after the optimization of the expansion model coefficients provided in an embodiment of the present invention.

[0035] In summary, the optimization method for the vertical roll width spread model coefficient of hot strip roughing mill proposed in this embodiment achieves accurate optimization of the width spread model coefficient of hot strip roughing mill without affecting production and with minimal equipment modification. It effectively improves the accuracy of width spread prediction and the level of strip width control, and has good engineering application value.

[0036] Furthermore, it should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0037] Finally, it should be noted that the above description represents a preferred embodiment of the present invention. It should be pointed out that although preferred embodiments have been described, those skilled in the art, once they understand the basic inventive concept of the present invention, can make various improvements and modifications without departing from the principles described herein. These improvements and modifications should also be considered within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present invention.

Claims

1. A dynamic optimization method for the width expansion model coefficient of a hot continuous rolling mill roughing mill, characterized in that, The specific steps are as follows: Step 1: Construct a roughing mill width development mechanism model including a dog-bone recovery width development model and a horizontal rolling natural width development model. The expression for the dog-bone recovery width development model is as follows: in, Correction factor for dog bone restoration; This is the vertical roller trough shape correction factor, with a value of 1.0; and Dimensionless intermediate variables characterizing the geometric features of the deformation region: in, For the inlet thickness; The radius of the vertical roller; The slope coefficient is set to 1.0 to restore the slope. This refers to the side pressure of the vertical roller; The expression for the horizontal rolling natural width expansion model is: in, This is a horizontal width correction factor; This refers to the width of the intermediate billet after vertical roll rolling; This refers to the reduction amount per horizontal pass. The radius of the horizontal working roller; Sedlaczek broadening factor; For the inlet thickness; Step 2: Using the dog bone recovery correction coefficient in the roughing mill width expansion mechanism model. and horizontal width correction factor To optimize the variables, the dog bone recovery correction coefficient is set in the roughing mill width expansion mechanism model. and the horizontal width correction coefficient Constraints: in, , These are the lower and upper bounds of the dog bone restoration correction coefficient, respectively. , and are the lower and upper bounds of the horizontal width correction coefficient, respectively; Step 3: Collect multi-source process data in real time during the hot continuous rolling roughing process, and perform outlier removal processing on the collected multi-source process data to form a complete dataset; Step 4: Based on the multi-source process data after removing outliers, construct the dog bone recovery correction coefficient in the roughing mill width extension mechanism model. and horizontal width correction factor The optimization objective function; Step 5: Solve the optimization objective function to obtain the optimal dog bone recovery correction coefficient under the current rolling conditions. and horizontal width correction factor ; Step Six: Apply the optimal dog bone restoration correction coefficient obtained in Step Five. and horizontal width correction factor The model is updated in real time to reflect the width expansion mechanism of the roughing mill. Actual production data is used to verify the accuracy of the model. If the width expansion deviation does not meet the preset value, steps three through five are repeated to restore the correction coefficients of the dog-bone structure in the roughing mill width expansion mechanism model. and horizontal width correction factor Perform adaptive dynamic optimization.

2. The dynamic optimization method for the width expansion model coefficient of a hot continuous rolling mill as described in claim 1, characterized in that, In step three, multi-source process data are collected in real time during the hot continuous rolling roughing process. The multi-source process data includes: roughing mill inlet thickness, roughing mill inlet width, vertical roll side pressure, horizontal reduction, vertical roll diameter, horizontal work roll diameter, and measured roughing mill outlet width.

3. The dynamic optimization method for the width expansion model coefficient of a hot continuous rolling mill as described in claim 1, characterized in that, In step three, outlier removal processing is performed on the collected multi-source process data to form a complete dataset. The specific method for outlier removal processing is as follows: The 3σ criterion is used to remove outliers from the acquired data, as shown in the following formula: in, This is the average value; Standard deviation, These are measured values.

4. The dynamic optimization method for the width expansion model coefficient of a hot continuous rolling mill as described in claim 1, characterized in that, The dog bone restoration correction coefficient in step four and horizontal width correction factor The optimization objective function is to minimize the sum of squared deviations between the predicted value and the actual width expansion of the roughing mill, as expressed by: in, N This represents the total number of strip steel samples in the dataset. For the first The measured width of the strip steel at the roughing mill exit, in mm; The predicted width value is in mm.

5. The dynamic optimization method for the width expansion model coefficient of a hot continuous rolling mill as described in claim 1, characterized in that, The dog-bone recovery correction coefficient of the roughing mill breadth mechanism model in step six. and horizontal width correction factor The adaptive dynamic optimization process is as follows: The optimized dog bone restoration correction coefficient and horizontal width correction factor Substitute the values ​​into the roughing mill width expansion mechanism model to calculate the deviation between the predicted width and the actual width. like Then output the optimized dog bone restoration correction coefficient. and horizontal width correction factor As coefficients in the current roughing mill width expansion mechanism model. The preset allowable threshold for width deviation; like Then, re-collect the latest multi-source process data from the roughing production, and repeat the optimization process until... , This is the preset allowable threshold for width deviation.

Images