Martensitic stainless steel finish rolling thickness control optimization method

Abstract

The invention belongs to the technical field of industrial manufacturing, and provides a martensitic stainless steel finish rolling thickness control optimization method. Comprising the following steps: collecting and sorting data, cleaning and preprocessing the collected and sorted data, screening variables by adopting an L1 regularization algorithm, establishing a regression model, and providing an optimization scheme by applying the regression model in combination with process conditions in production; according to the method, key variables influencing thickness changes are screened out through L1 regularization, parameter setting in the finish rolling process is optimized through the regression model, the phenomenon that the head of the strip steel is thin and the tail of the strip steel is thick can be remarkably reduced, and therefore the thickness consistency and precision of a final finished product are improved, the product rejection rate caused by the fact that the thickness does not reach the standard is avoided, and the product quality is improved. Therefore, the production efficiency is improved, and the production cost is reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of industrial manufacturing, and relates to an optimization method for finishing rolling control, in particular to an optimization method for finishing rolling thickness control of martensitic stainless steel. Background Art

[0002] Due to its unique physical properties and structure, martensitic stainless steel is widely used in various industrial manufacturing, especially in fields that require high-strength and corrosion-resistant materials such as chemical industry, nuclear power, aerospace, etc. In the production and manufacturing process of martensitic stainless steel, especially in the finishing stage, an important challenge is how to accurately control its thickness. In particular, when the thickness is less than 3.0 mm, the thickness control of the head and tail of the finished strip often deviates, resulting in substandard thickness of the finished product, affecting subsequent processing and use. Figure 1 As shown in the figure, the actual thickness curve of the strip shows obvious deviation in thickness over time.

[0003] The specific phenomenon is that the thickness of the head is thinner and the thickness of the tail is thicker. This phenomenon is not only related to the temperature, pressure, and plastic deformation of the material, but also affected by many factors such as the response speed of AGC (automatic thickness control system) and gain setting during the finishing rolling process.

[0004] To address these problems, traditional thickness control methods such as increasing the AGC gain and manually adjusting the tail compensation coefficient are not very effective. Summary of the invention

[0005] The purpose of the present invention is to provide a method for optimizing the thickness control of martensitic stainless steel finish rolling. Based on the L1 regularization algorithm, a linear regression model is constructed to establish a quantitative relationship between variables, thereby screening out the variables with the greatest impact on thickness change, optimizing the control parameters of each link in the finish rolling process, and ensuring that the thickness of the final product meets the target requirement.

[0006] The technical solution adopted by the present invention to achieve the above-mentioned purpose is: A method for optimizing the thickness control of martensitic stainless steel finish rolling, comprising the following steps: S1. Data collection and processing, including: Pressure changes at different stages during rolling; by monitoring load pressure data, analyze the impact of different pressures on stainless steel finishing thickness; AGC output data; collect AGC total output, absolute value AGC output, relative AGC output and monitor AGC output data, and analyze the actual impact of the data on thickness control; Temperature data: collect real-time data of stainless steel finishing inlet and outlet temperatures; Thickness gauge data; relationship between the thickness gauge's outlet thickness data and other variables.

[0007] S2. Clean and preprocess the collected data to delete noise data and outliers.

[0008] S3. Use the L1 regularization algorithm to screen the variables. By introducing penalty terms, the coefficients of unimportant variables are turned to zero, thereby screening out variables that have a significant impact on the results. The specific steps are as follows: S31. Establishing a preliminary model: Preliminarily establishing the relationship between thickness and multiple variables through a multiple linear regression model; S32. Introduce L1 regularization term: Introduce L1 regularization term in the regression model to control the complexity of the model and reduce the impact of multicollinearity; S33. Variable screening: Through model training, variables with non-zero coefficients are screened out. These variables are the key variables that have a greater impact on thickness changes.

[0009] S4. Establish a regression model; after screening out the key variables, construct a linear regression model to determine the quantitative relationship between each variable and the stainless steel finish rolling thickness. The specific steps are as follows: S41. Data division: Divide the collected data into training set and test set to ensure the generalization ability of the model; S42, model training: using the data in the training set, training the linear regression model and determining the regression coefficient of each variable; S43, Model validation: Use the test set data to validate the model and evaluate the prediction accuracy and reliability of the model; S44. Result analysis: By analyzing the size and sign of the regression coefficient, the positive and negative effects of different variables on thickness changes can be determined.

[0010] S5. Use regression model and process conditions in production to propose optimization solutions.

[0011] Beneficial effects of the method of the present invention: 1. The method of the present invention screens out the key variables that affect the thickness change through L1 regularization, and uses the regression model to optimize the parameter setting in the finishing rolling process, which can significantly reduce the phenomenon of thinness at the head of the strip and thickness at the tail, thereby improving the thickness consistency and accuracy of the final finished stainless steel; avoiding the product scrap rate caused by substandard thickness, thereby improving production efficiency and reducing production costs; 2. Based on data analysis and modeling results, the method of the present invention improves the response speed and gain setting of the finishing rolling AGC system, so that the finishing rolling system can respond to thickness changes faster and more accurately, reducing thickness deviations in the production process; 3. The method of the present invention introduces data analysis and machine learning technology to make the setting of finishing rolling process parameters more automated and intelligent, reducing the reliance on manual intervention and experience-based adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0013] The method of the present invention is further described below in conjunction with the accompanying drawings and embodiments: The method of the present invention is applied to a hot rolling 2250 finishing mill, and the finishing thickness control of martensitic stainless steel is optimized based on L1 regularization and regression model under the background of big data.

[0014] The optimization method includes the following steps: S1. Data collection and collation: In the stainless steel finishing process, there are a large number of variables that will affect the thickness of the final stainless steel strip. In order to comprehensively analyze these variables, it is necessary to first collect a large amount of production variable data, including: 1) Pressure changes at different stages during the rolling process, such as pressure feedback from F1 to F7; by monitoring the load pressure data, analyze the impact of different pressures on the thickness of stainless steel finish rolling. Pressure changes at different stages during the rolling process have a direct impact on the plastic deformation of the material and are an important variable affecting the thickness; 2) AGC output data: The automatic adjustment function of the AGC system plays a key role in thickness control. Collect data such as AGC total output, absolute value AGC output, relative AGC output and monitoring AGC output, and analyze the actual impact of the data on thickness control; 3) Temperature data: collect real-time data of stainless steel finishing inlet and outlet temperatures; temperature is an important factor affecting material deformation behavior. During the finishing process, changes in inlet and outlet temperatures will have a direct impact on the thickness of the strip; 4) Thickness gauge data; Relationship between the thickness gauge outlet data and other variables; The outlet thickness data of the F7 thickness gauge is used as the response variable. By analyzing the relationship with other variables, the main factors affecting the thickness change can be found.

[0015] S2. After data collection is completed, the collected and organized data is cleaned and preprocessed to delete noise data and outliers to ensure the accuracy and reliability of the data.

[0016] S3. Use the L1 regularization algorithm to screen the variables. By introducing penalty terms, the coefficients of unimportant variables are turned to zero, thereby screening out variables that have a significant impact on the results. The specific steps are as follows: S31. Establishing a preliminary model: Preliminarily establishing the relationship between thickness and multiple variables through a multiple linear regression model; S32. Introduce L1 regularization term: Introduce L1 regularization term in the regression model to control the complexity of the model and reduce the impact of multicollinearity; S33, variable screening: through model training, screen out variables with non-zero coefficients, which are the key variables that have a greater impact on thickness change; The variables selected by L1 regularization can significantly simplify the model structure while improving the model's interpretability and predictive ability.

[0017] S4, establish regression model; After screening out the key variables, a linear regression model was constructed to determine the quantitative relationship between each variable and the stainless steel finishing thickness. The specific steps are as follows: S41. Data division: Divide the collected data into training set and test set to ensure the generalization ability of the model; S42, model training: using the data in the training set, training the linear regression model and determining the regression coefficient of each variable; S43, Model validation: Use the test set data to validate the model and evaluate the prediction accuracy and reliability of the model; S44. Result analysis: By analyzing the size and sign of the regression coefficient, the positive and negative effects of different variables on thickness changes can be determined.

[0018] S5. Propose optimization solutions based on the process conditions in actual production.

[0019] By establishing a regression model, we can identify factors that have a significant impact on the problem of thin head and thick tail, thus providing a basis for adjusting process parameters.

[0020] Based on the analysis results of the regression model, the parameter settings of the finishing rolling AGC system are optimized, and the following specific optimization suggestions are proposed for the AGC response speed, gain setting and tail compensation coefficient: Increase the AGC response speed; by increasing the response speed of the AGC system, the frequency of thickness deviation can be reduced; Adjust the tail compensation coefficient; To address the problem of thick tail, optimize the tail compensation coefficient to ensure that the thickness remains within the target range during the tail rolling process; Enhanced AGC gain setting: Based on the traditional gain setting, the gain value of the AGC system is further improved to ensure that the system can respond to thickness changes quickly and effectively.

[0021] The above steps improve the response speed and gain setting of the AGC system based on data analysis and modeling results, allowing the system to respond to thickness changes faster and more accurately, reducing thickness deviations during production.

[0022] After the model and parameter optimization were completed, the effect of the optimization scheme was verified through field tests. By applying the optimized AGC parameter settings on the actual production line, the significant improvement of the head thinning and tail thickening phenomena was verified. The consistency and accuracy of the final thickness control were evaluated through comparative analysis of thickness gauge data.

[0023] The method of the present invention screens out the key variables that affect the thickness change through L1 regularization, and uses the regression model to optimize the parameter setting in the finishing rolling process, which can significantly reduce the phenomenon of thinness at the head of the strip and thickness at the tail, thereby improving the thickness consistency and accuracy of the final product, avoiding the product scrap rate due to thickness not meeting the standard, thereby improving production efficiency and reducing production costs.

[0024] The method of the present invention introduces data analysis and machine learning technology to make the parameter setting in the finishing rolling process more automated and intelligent, reducing the reliance on manual intervention and experience adjustment. The method of the present invention is not only applicable to the finishing rolling process of martensitic stainless steel, but can also be promoted to the thickness control and optimization of other types of stainless steel or metal materials, and has broad industrial application prospects.

Claims

1. A method for optimizing the thickness control of martensitic stainless steel finish rolling, characterized in that: The steps include: S1, data collection and compilation; S2. Clean and preprocess the collected data to delete noise data and outliers; S3, using L1 regularization algorithm to screen variables, by introducing penalty terms, the coefficients of unimportant variables are turned to zero, thereby screening out variables that have a significant impact on the results; S4. Establish a regression model; After screening out the key variables, a linear regression model was constructed to determine the quantitative relationship between each variable and the stainless steel finish rolling thickness; S5. Use regression model and process conditions in production to propose optimization solutions.

2. The method for optimizing the thickness control of martensitic stainless steel finish rolling according to claim 1, characterized in that: In S1, the data collected and sorted include: 1) Pressure changes at different stages during rolling; by monitoring load pressure data, analyze the impact of different pressures on the thickness of stainless steel finish rolling; 2) AGC output data: collect AGC total output, absolute value AGC output, relative AGC output and monitoring AGC output data, and analyze the actual impact of the data on thickness control; 3) Temperature data: collect real-time data of stainless steel finishing inlet and outlet temperatures; 4) Thickness gauge data; relationship between the thickness gauge's outlet thickness data and other variables.

3. The method for optimizing the thickness control of martensitic stainless steel finish rolling according to claim 1, characterized in that: The specific steps of S3 include: S31. Establishing a preliminary model: Preliminarily establishing the relationship between thickness and multiple variables through a multiple linear regression model; S32. Introduce L1 regularization term: Introduce L1 regularization term in the regression model to control the complexity of the model and reduce the impact of multicollinearity; S33. Variable screening: Through model training, variables with non-zero coefficients are screened out. These variables are the key variables that have a greater impact on thickness changes.

4. The method for optimizing the thickness control of martensitic stainless steel finish rolling according to claim 1, characterized in that: The specific steps of S4 include: S41. Data division: Divide the collected data into training set and test set to ensure the generalization ability of the model; S42, model training: using the data in the training set, training the linear regression model and determining the regression coefficient of each variable; S43, Model validation: Use the test set data to validate the model and evaluate the prediction accuracy and reliability of the model; S44. Result analysis: By analyzing the size and sign of the regression coefficient, the positive and negative effects of different variables on thickness changes can be determined.

Images