Plate flexible rolling process optimization method
Through the combination of AI prediction model and process decision model, the problems of narrow steel grade range, complex process adjustment and high cost in flexible rolling technology are solved, and the intelligent and efficient flexible rolling technology is realized to meet diversified market demands. .
Patent Information
- Application Number
- CN202411611922.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-12
- Publication Date
- 2025-01-10
AI Technical Summary
The existing flexible rolling technology is unable to effectively deal with diversified market demands in terms of narrow steel range, complex process adjustments and high production costs, which affects production efficiency and competitiveness.
Use AI prediction models to predict real-time performance, build process decision models, output process adjustment suggestions, reduce traditional sampling and detection links, reduce process adjustment complexity, and improve production efficiency.
It realizes the intelligence and efficiency of flexible rolling technology, shortens feedback time, reduces production costs, improves product quality and production efficiency, and meets the needs of diversified steel grades.
Smart Images
Figure CN119281816A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer-aided engineering, and in particular to a method for optimizing a flexible rolling process for plates. Background Art
[0002] With the continuous development of the global steel industry, the market demand for different steel grades is increasing. Traditional rolling technology is weak in dealing with diverse steel grades and complex management requirements. Especially in the production process of plates, flexible rolling technology, as an emerging manufacturing process, aims to improve production efficiency, shorten production cycle, and effectively respond to diverse market demands. However, the existing flexible rolling technology has encountered a series of problems in practical applications, and its technological progress needs to be broken through to meet the needs of modern steel production.
[0003] Current flexible rolling technology has significant limitations in the following areas:
[0004] 1. The range of steel types covered is narrow. Existing flexible rolling technology usually requires strict restrictions on the chemical composition of steel types and can only operate within a very narrow range. This limitation makes the production line unable to flexibly adapt to diverse market demands, affecting the competitiveness of enterprises;
[0005] 2. The process of process adjustment is complicated. During the production process, the adjustment of process parameters usually requires multiple tests to achieve the desired effect. After each adjustment, the production team needs to wait for the mechanical properties test results of the final product to judge the effectiveness of the adjustment. This long feedback process leads to a decrease in production efficiency and an inability to respond to changes in the production environment in a timely manner;
[0006] 3. Significant increase in production costs. Due to the continuous adjustment of the process confirmation process, the production process is inevitably accompanied by a large amount of waste and loss of test samples, which not only increases the cost of process adjustment, but also involves the related expenses of fine processing and performance testing of samples. This situation is particularly obvious in the flexible rolling process, thereby raising the overall production cost. Summary of the invention
[0007] The embodiment of the present application provides a method for optimizing the flexible rolling process of plates. The present application uses an AI prediction model to make real-time predictions on product performance during the production process, thereby effectively eliminating the traditional sampling and detection links and significantly shortening the feedback time. Based on the prediction results of the prediction model and combined with the process measures that have been implemented, a process decision model is constructed to output subsequent process adjustment suggestions in real time. This innovative method aims to reduce the complexity of process adjustments, improve production efficiency, and ultimately realize the intelligence and efficiency of flexible rolling technology, thereby better meeting the market demand for diversified steel grades and solving the problem that the process adjustment process of flexible rolling technology is complex and the production cost increases significantly during the adjustment process.
[0008] The embodiment of the present application provides a method for optimizing a plate flexible rolling process, comprising:
[0009] Data collection, to obtain high-quality data collection;
[0010] Constructing a prediction model, and training the prediction model through the high-quality data collection to obtain an optimal prediction model;
[0011] Constructing a process decision model, inputting first data into the process decision model, and obtaining unimplemented process adjustment measures;
[0012] Inputting the second data including the unimplemented process adjustment measures into the optimal prediction model, the optimal prediction model outputting a final mechanical property prediction value;
[0013] If the predicted value of the mechanical property reaches the acceptable range of the target performance, the unimplemented process adjustment measure is adopted.
[0014] In a feasible implementation, if the predicted value of the mechanical property does not reach the acceptable range of the target performance, the step of inputting the first data into the process decision model to obtain the step of not implementing the process adjustment measures is repeated.
[0015] In a feasible implementation, during the interaction between the process decision model and the optimal prediction model, after the first data is input into the process decision model, the process decision model outputs a plurality of the unimplemented process adjustment measures;
[0016] One or more of the unimplemented process adjustment measures are selected and input into the optimal prediction model for verification, thereby evaluating the quality of the input unimplemented process.
[0017] In a feasible implementation, the second data also includes customized specific process measures;
[0018] The second data including the customized specific process measures are input into the optimal prediction model, and the effect of the customized specific process measures is evaluated according to the final mechanical property prediction value output by the optimal prediction model.
[0019] In a feasible implementation, the second data also includes chemical composition information of the slab and information on implemented process measures.
[0020] In a feasible implementation, the data collection to obtain a high-quality data collection specifically includes:
[0021] Data collection, to obtain a collection of production data;
[0022] The production data collection is preprocessed to obtain the high-quality data collection; the preprocessing includes missing value processing, outlier processing, cluster analysis and feature engineering.
[0023] In a feasible implementation, the method of constructing a prediction model and training the prediction model through the high-quality data collection to obtain the best prediction model includes:
[0024] Use a machine learning algorithm or a deep learning algorithm to construct a prediction model; the machine learning algorithm includes a random forest regression model, an LGBM regression model, and an XGBoost regression model, and the deep learning algorithm includes a fully connected neural network algorithm;
[0025] The prediction model is trained based on the high-quality data collection, and the model with the best effect is selected through hyperparameter optimization to obtain the best prediction model.
[0026] In a feasible implementation, the method of training the prediction model based on the high-quality data collection and selecting the best model through hyperparameter optimization to obtain the best prediction model includes:
[0027] Divide the high-quality data collection into a training set and a test set according to a certain ratio, with 70% of the data used for model training and 30% of the data used for model verification;
[0028] Adjust the hyperparameters of the regression model, including learning rate, tree depth, regularization parameters, etc., to optimize model performance;
[0029] After the model training is completed, cross-validation is used to ensure the reliability of the model, and the mean square error and determination coefficient of the prediction are calculated as performance evaluation indicators.
[0030] In a feasible implementation, the first data includes chemical composition information of the current slab, information on implemented process measures, and target mechanical properties information.
[0031] In a feasible implementation, the unimplemented process adjustment measures include controlling the rolling temperature, controlling the rolling compression ratio, the post-rolling cooling rate and the red-returning temperature.
[0032] The overall idea of this application is as follows:
[0033] The overall implementation idea of this application is based on the AI prediction model and guided by the intelligent decision-making model, aiming to reduce the complexity of plate flexible rolling process adjustment, improve production efficiency and product quality, and ultimately realize the intelligence and efficiency of flexible rolling technology, so as to better meet the market demand for diversified steel grades. The specific process is divided into the following main steps:
[0034] Step 1: Data Collection:
[0035] The first stage involves comprehensive collection of production data from a steel mill in recent years. The collected data includes chemical composition data of slabs, continuous casting production data, steel rolling furnace data, real-time rolling data, post-rolling cooling data, and mechanical properties data (including yield strength, tensile strength, elongation after fracture, and impact energy).
[0036] All data are summarized and integrated according to furnace number, slab number and sample number. Then, through a variety of data analysis and processing techniques, such as missing value processing, outlier processing, cluster analysis and feature engineering, the final data is ensured to be of high quality and high availability.
[0037] Step 2: Build a prediction model:
[0038] After the data integration is completed, a machine learning algorithm (such as random forest regression, LGBM regression, XGBoost regression) or a deep learning algorithm (such as fully connected neural network) is used to build a prediction model. Relying on the high-quality data obtained in step 1, the model is trained, and the best model is selected through hyperparameter optimization to achieve accurate prediction of the mechanical properties of the plate.
[0039] Step 3: Build a process decision model:
[0040] After completing the construction of the prediction model, the process decision model is established using multi-objective intelligent optimization algorithms (such as particle swarm optimization, simulated annealing algorithm) and adversarial production algorithms.
[0041] The input of the model includes the chemical composition information of the current slab, the information of the implemented process measures and the target mechanical properties, and the output is the unimplemented process adjustment measures, mainly including the controlled rolling temperature, the controlled rolling compression ratio, the cooling rate after rolling, the red-returning temperature, etc. This decision-making model will optimize the key process parameters in the rolling process to ensure that the performance of the final product meets the requirements.
[0042] Step 4: Realize intelligent flexible rolling:
[0043] The output of the process decision model is used as one of the inputs of the prediction model. The input of the prediction model should also include the chemical composition information of the slab and the information of the implemented process measures. After calculation, the prediction model will output the final mechanical properties of the slab under the process measures recommended by the current process decision model. If the predicted value of the mechanical properties output by the prediction model reaches the acceptable range of the target performance, the process measures output by the decision model are directly adopted as a reference for the next process adjustment. If the predicted value fails to meet the preset acceptable standard, the process decision model will continue to optimize until the requirements are met.
[0044] Step 5: Interaction process and operational flexibility:
[0045] In the interaction process between the decision model and the prediction model, the operator can flexibly choose to input the process measures output by the decision model into the prediction model for verification, and then evaluate the pros and cons of the process measures. At the same time, the operator can also customize specific process measures and input them into the prediction model for effect evaluation. In addition, the operator can adjust the acceptable range of the target performance in real time to achieve recalculation and optimization of the decision model, ensuring the flexibility and adaptability of the production process.
[0046] This application uses AI prediction models to make real-time predictions of product performance during the production process, thereby effectively eliminating the traditional sampling and testing process and significantly shortening the feedback time. Based on the prediction results of the prediction model and combined with the process measures that have been implemented, a process decision model is constructed to output subsequent process adjustment suggestions in real time. This innovative method aims to reduce the complexity of process adjustments, improve production efficiency, and ultimately realize the intelligence and efficiency of flexible rolling technology, thereby better meeting the market demand for diversified steel grades. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 It is a flow chart of the steps of the plate flexible rolling process optimization method provided by the present application;
[0048] Figure 2 It is a work flow chart of the plate flexible rolling process optimization method;
[0049] Figure 3 It is a logical flow chart of the plate flexible rolling process optimization method. DETAILED DESCRIPTION
[0050] In order to enable those skilled in the art to better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present application.
[0051] Flexible rolling process is an advanced rolling technology. Its main feature is that it can roll steel or other metal materials of different specifications (thickness, width, material, etc.) on the same rolling production line by quickly adjusting rolling parameters such as rolling force, rolling speed, rolling temperature, etc.
[0052] For example, when rolling steel plates of different thicknesses, the flexible rolling process can accurately control the gap and rolling force of the rollers, and quickly adjust the parameters from rolling thicker steel plates to rolling thinner steel plates, greatly improving the adaptability and production efficiency of the rolling production line. At the same time, it can also effectively reduce production costs and reduce operations such as frequent replacement of production line equipment when producing products of different specifications.
[0053] The current flexible rolling technology has significant limitations in the following aspects: 1. The range of steel grades covered is narrow. Existing flexible rolling technology usually requires strict restrictions on the chemical composition of steel grades and can only operate within a very narrow range. This limitation makes the production line unable to flexibly adapt to diverse market demands and affects the competitiveness of enterprises; 2. The process adjustment process is complicated. During the production process, the adjustment of process parameters usually requires multiple tests to achieve the desired effect. After each adjustment, the production team needs to wait for the mechanical property test results of the final product to judge the effectiveness of the adjustment. This long feedback process leads to a decrease in production efficiency and cannot respond to changes in the production environment in a timely manner; 3. Significant increase in production costs. Due to the continuous adjustment of the process confirmation process, the production process is inevitably accompanied by a large amount of waste and loss of test samples, which not only increases the cost of process adjustment, but also involves the related expenses of fine processing and performance testing of samples. This situation is particularly obvious in the flexible rolling process, thereby raising the overall production cost.
[0054] The present application provides a method for optimizing the flexible rolling process of plates. The present application uses an AI prediction model to make real-time predictions on product performance during the production process, thereby effectively eliminating the traditional sampling and testing process and significantly shortening the feedback time. Based on the prediction results of the prediction model and combined with the process measures that have been implemented, a process decision model is constructed to output subsequent process adjustment suggestions in real time. This innovative method aims to reduce the complexity of process adjustments, improve production efficiency, and ultimately realize the intelligence and efficiency of flexible rolling technology, thereby better meeting the market demand for diversified steel grades and solving the problem that the process adjustment process of flexible rolling technology is complex and the production cost increases significantly during the adjustment process.
[0055] The specific structure of the plate flexible rolling process optimization method provided in this application is described in detail below in conjunction with the accompanying drawings.
[0056] Reference Figure 1 As shown, the embodiment of the present application provides a method for optimizing a plate flexible rolling process, comprising:
[0057] S1: Data collection, obtaining high-quality data collection;
[0058] S2: construct a prediction model, and train the prediction model through the high-quality data collection to obtain the best prediction model;
[0059] S3: constructing a process decision model, inputting first data into the process decision model, and obtaining unimplemented process adjustment measures;
[0060] S4: inputting the second data including the unimplemented process adjustment measures into the optimal prediction model, and the optimal prediction model outputs the final mechanical property prediction value;
[0061] If the predicted value of the mechanical property reaches the acceptable range of the target performance, the unimplemented process adjustment measure is adopted.
[0062] Reference Figure 2 and Figure 3 As shown, this embodiment proposes a flexible rolling method combining intelligent prediction and process decision model for the production process of medium and thick plates in a large steel plant. The method optimizes rolling process parameters through real-time data analysis and machine learning technology, thereby improving the mechanical properties and production efficiency of the product.
[0063] Step 1: Data collection and integration:
[0064] 1. Data source:
[0065] 1) Chemical composition data: In the continuous casting MES system, the composition data of batch steel is collected according to the steelmaking furnace number, including the content of elements such as carbon (C), silicon (Si), manganese (Mn), phosphorus (P), and sulfur (S).
[0066] 2) Steel rolling heating furnace data: In the steel rolling MES system, the slab heating data is obtained according to the slab number, including charging temperature, heating time, heating temperature, discharge temperature and other data.
[0067] 3) Rolling process data: In the secondary rolling system, real-time data of the rolling process is obtained according to the slab number, including rolling temperature, rolling speed, reduction, etc.
[0068] 4) Cooling data: obtained from the secondary system of steel rolling, which records various temperatures and cooling rates during the post-rolling cooling stage.
[0069] 5) Inspection and testing data: obtained from the MES system of the inspection and testing center, including the mechanical properties test data of each batch of plates, such as yield strength, tensile strength, elongation and impact energy, as well as the corresponding test conditions data.
[0070] 2. Data integration and preprocessing:
[0071] 1) Integrate the data of the same slab at different stages according to the furnace number, slab number and test sample number.
[0072] 2) For the missing values and outliers in the data set, they are merged according to similar steel grades and similar rolling process measures;
[0073] 3) By analyzing the relationship between data and mechanical properties, the feature selection technology based on random forest is used to screen out key features with high correlation with performance.
[0074] Step 2: Construct a mechanical properties prediction model:
[0075] 1. Model selection:
[0076] Use a machine learning algorithm or a deep learning algorithm to construct a prediction model; the machine learning algorithm includes a random forest regression model, an LGBM regression model, and an XGBoost regression model, and the deep learning algorithm includes a fully connected neural network algorithm;
[0077] This embodiment selects a machine learning algorithm for prediction, and finally determines to use the XGBoost algorithm because it performs well in processing nonlinear relationships and high-dimensional data, and supports the use of GPU to accelerate the training process.
[0078] 2. Model training:
[0079] 1) Divide the cleaned data into training set and test set according to a certain ratio, with 70% of the data used for model training and 30% of the data used for model verification.
[0080] 2) Adjust the hyperparameters of the XGBoost model, including learning rate, tree depth, regularization parameters, etc., to optimize model performance.
[0081] 3) After the model training is completed, the reliability of the model is ensured through cross-validation, and the mean square error (MSE) and determination coefficient (R 2 ) as the performance evaluation index.
[0082] Step 3: Build a process decision model:
[0083] 1. Model establishment:
[0084] Based on production objectives and adjustable process parameters, a multi-objective particle swarm optimization (PSO) algorithm is constructed to implement the process decision model.
[0085] 2. Input parameter settings:
[0086] Input parameters: Input data used for decision-making models, usually including but not limited to:
[0087] 1) Chemical composition data of rolled slab: such as the content of carbon (C), manganese (Mn), silicon (Si), nickel (Ni) and other elements.
[0088] 2) Heating furnace data of slab before rolling: such as charging temperature, time in the furnace, discharge temperature and other key process parameters.
[0089] 3) Completed rolling data: such as starting rolling temperature and slab thickness before entering the rolling mill.
[0090] 4) Ambient cooling conditions: such as the temperature and flow rate of the cooling medium.
[0091] 5) Target parameters of the product: such as target thickness, target yield strength, target tensile strength, target elongation, target impact energy, etc.
[0092] 3. Setting of output parameters;
[0093] Output parameters: The optimized process conditions generated by the decision model after calculation, usually including:
[0094] 1) Controlled rolling temperature: During the steel rolling process, the corresponding temperature is used to ensure the material properties.
[0095] 2) Controlled rolling compression ratio: The required compression ratio during the rolling process to ensure the appropriate amount of deformation.
[0096] 3) Post-rolling cooling rate: The cooling rate after rolling affects the microstructure and properties of the final product.
[0097] 4) Red-back temperature: The appropriate red-back temperature ensures stress elimination and performance improvement in subsequent processes.
[0098] 4. Decision output:
[0099] The process decision model calculates and simulates the input parameters and outputs the optimal combination of the above parameters to facilitate real-time adjustments during the rolling process.
[0100] Step 4: Implementation of intelligent flexible rolling:
[0101] 1. Actual operation process:
[0102] 1) During the real-time rolling process, the operator inputs the effective rolling parameters into the process decision model based on the output of the prediction model.
[0103] 2) The prediction model calculates the current mechanical properties to monitor the quality of the plate after rolling in real time.
[0104] 2. Feedback adjustment mechanism:
[0105] 1) When the predicted performance value (such as yield strength) is within the target range, the operator can directly use the output parameters of the process decision model for production.
[0106] 2) If the predicted value does not meet the preset standard, the operator can readjust the process parameters based on the feedback information and perform recalculation and optimization.
[0107] Step 5: System interaction and flexible adjustment:
[0108] 1. Interactive operation:
[0109] 1) Operators can independently choose to input the process measures output into the decision-making model into the prediction model for verification and evaluate the reliability of promotion.
[0110] 2) At the same time, operators can adjust the acceptable range of target performance in real time based on market changes and customer needs.
[0111] 2. Real-time update and optimization:
[0112] 1) The system has a real-time update function and can continuously optimize the decision-making model based on various parameters in the rolling process.
[0113] 2) Feedback information will be continuously used in the model’s training data to improve the model’s prediction accuracy and reliability in subsequent production.
[0114] Through the effective implementation of the above steps, the mechanical properties of the steel plant's medium plate products have been significantly improved, and the yield strength and tensile strength have reached the expected standards. In addition, due to the optimization of the entire rolling process and continuous data feedback, production efficiency has increased by 20%, and the pass rate has increased to more than 95%.
[0115] The technical solution provided in this application can realize intelligent optimization management of the steel plate rolling process by constructing a mechanical property prediction model and a process decision model, thereby improving production efficiency and product quality. The specific effects are analyzed as follows:
[0116] 1. Optimization of data cleaning and feature selection:
[0117] This application first cleans and extracts features from a large amount of data collected during the steel production process. By removing redundant and noisy data, the accuracy and reliability of the model input data are ensured. This process enables the subsequent mechanical property prediction model to achieve higher accuracy during the training process. Accurate input data directly improves the credibility of the prediction results, making the decision in production more scientific and reasonable.
[0118] 2. Multi-model integration and optimization selection:
[0119] This application integrates a variety of algorithm models, including XGBoost, random forest, LGBM, etc., for mechanical property prediction. This multi-model integration method allows the system to dynamically evaluate the prediction effect of each model based on historical data and real-time feedback. By selecting the best performing model, the accuracy and stability of the prediction can be effectively improved. This flexibility means that under different production conditions and data characteristics, the system can automatically select the most appropriate algorithm to ensure high-precision prediction of the yield strength and tensile strength of the rolled product. This feature not only improves the adaptability of the model, but also greatly reduces the production risks caused by improper model selection.
[0120] 3. Dynamically adaptive process decision-making mechanism:
[0121] In the process decision model, the particle swarm optimization algorithm is used, combined with other optimization strategies (such as genetic algorithm and simulated annealing, etc.) to achieve the collaborative work of multiple optimization algorithms. By real-time monitoring of various key parameters in the rolling process, the system can evaluate and select the optimal algorithm under the current environment to adjust and optimize the process parameters. This dynamic adaptation mechanism ensures the precise adjustment of various process parameters (such as temperature, compression ratio and cooling rate) during the rolling process, so that product quality and production efficiency are maximized. At the same time, this mechanism improves the flexibility of the system, can quickly respond to changes in market demand and production conditions, and reduce production costs and resource waste.
[0122] 4. Real-time feedback and optimization mechanism:
[0123] This application designs a feedback adjustment mechanism that can monitor the performance of the rolled plate in real time and make dynamic adjustments based on real-time data. This feature ensures the adaptability of the model, can flexibly respond to possible deviations in the production process, and quickly respond to changes in market demand, thereby improving production flexibility and response speed.
[0124] 5. Improved system integration and operability:
[0125] By combining mechanical property prediction with process decision-making to form a closed-loop control system, operators in the rolling process can more intuitively understand the relationship between process requirements and product performance. This integrated system design improves the convenience and intuitiveness of operation, increases production efficiency, and reduces quality problems caused by human errors.
[0126] In summary, the present invention forms a close connection between each technical feature and the final effect by clearly defining input and output parameters, applying advanced algorithms, designing a flexible feedback adjustment mechanism and effective system integration, which can improve the intelligence level of the plate rolling process.
[0127] It is easy to understand that those skilled in the art can combine, split, reorganize, etc. the embodiments of the present application to obtain other embodiments based on the several embodiments provided in the present application, and these embodiments do not exceed the protection scope of the present application.
[0128] The above specific implementation methods further explain in detail the purpose, technical solutions and beneficial effects of the embodiments of the present application. It should be understood that the above are only specific implementation methods of the embodiments of the present application and are not used to limit the protection scope of the embodiments of the present application. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solutions of the embodiments of the present application should be included in the protection scope of the embodiments of the present application.
Claims
1. A plate flexible rolling process optimization method, characterized in that: include; Data collection, to obtain high-quality data collection; Constructing a prediction model, and training the prediction model through the high-quality data collection to obtain an optimal prediction model; Constructing a process decision model, inputting first data into the process decision model, and obtaining unimplemented process adjustment measures; Inputting the second data including the unimplemented process adjustment measures into the optimal prediction model, the optimal prediction model outputting a final mechanical property prediction value; If the predicted value of the mechanical property reaches the acceptable range of the target performance, the unimplemented process adjustment measure is adopted.
2. The plate flexible rolling process optimization method according to claim 1, characterized in that: If the predicted value of the mechanical property does not reach the acceptable range of the target performance, the step of inputting the first data into the process decision model to obtain the step of not implementing the process adjustment measure is repeated.
3. The plate flexible rolling process optimization method according to claim 2, characterized in that: During the interaction between the process decision model and the optimal prediction model, after the first data is input into the process decision model, the process decision model outputs a plurality of the unimplemented process adjustment measures; One or more of the unimplemented process adjustment measures are selected and input into the optimal prediction model for verification, thereby evaluating the quality of the input unimplemented process.
4. The plate flexible rolling process optimization method according to claim 3 is characterized in that: The second data also includes customized specific process measures; The second data including the customized specific process measures are input into the optimal prediction model, and the effect of the customized specific process measures is evaluated according to the final mechanical property prediction value output by the optimal prediction model.
5. The plate flexible rolling process optimization method according to claim 4, characterized in that: The second data also includes chemical composition information of the slab and information on implemented process measures.
6. The plate flexible rolling process optimization method according to claim 1, characterized in that: The data collection to obtain a high-quality data collection specifically includes: Data collection, to obtain a collection of production data; The production data collection is preprocessed to obtain the high-quality data collection; the preprocessing includes missing value processing, outlier processing, cluster analysis and feature engineering.
7. The plate flexible rolling process optimization method according to claim 1, characterized in that: The method of constructing a prediction model and training the prediction model through the high-quality data collection to obtain the best prediction model includes: Use a machine learning algorithm or a deep learning algorithm to construct a prediction model; the machine learning algorithm includes a random forest regression model, an LGBM regression model, and an XGBoost regression model, and the deep learning algorithm includes a fully connected neural network algorithm; The prediction model is trained based on the high-quality data collection, and the model with the best effect is selected through hyperparameter optimization to obtain the best prediction model.
8. The plate flexible rolling process optimization method according to claim 1, characterized in that: The method of training the prediction model based on the high-quality data collection and selecting the best model by hyperparameter optimization to obtain the best prediction model includes: Divide the high-quality data collection into a training set and a test set according to a certain ratio, with 70% of the data used for model training and 30% of the data used for model verification; Adjust the hyperparameters of the regression model, including learning rate, tree depth, regularization parameters, etc., to optimize model performance; After the model training is completed, cross-validation is used to ensure the reliability of the model, and the mean square error and determination coefficient of the prediction are calculated as performance evaluation indicators.
9. The plate flexible rolling process optimization method according to claim 1, characterized in that: The first data includes chemical composition information of the current slab, information on implemented process measures, and target mechanical properties information.
10. The plate flexible rolling process optimization method according to claim 1, characterized in that: The unimplemented process adjustment measures include controlling the rolling temperature, controlling the rolling reduction ratio, the post-rolling cooling rate and the red-returning temperature.