Method for efficiently optimizing continuous casting solidification process technological parameters based on generative model and related device

Abstract

The invention belongs to the field of intellectualization of a continuous casting process of ferrous metallurgy, and discloses a method and a related device for efficiently optimizing process parameters of a continuous casting solidification process based on a generative model, which utilize a trained simulation model of the continuous casting solidification process to process the obtained process parameters of the online real-time continuous casting process, so as to optimize the process parameters of the continuous casting solidification process. Obtaining predicted physical field data in the continuous casting solidification process; extracting feature data of the predicted physical field data; utilizing the characteristic data to calculate metallurgical indexes in the continuous casting process; whether the metallurgical indexes meet the requirements of metallurgical criteria or not is judged, and if the metallurgical indexes meet the requirements, the obtained technological parameters of the online real-time continuous casting process serve as technological parameters of the continuous casting solidification process; otherwise, optimizing the obtained process parameters of the online real-time continuous casting process to obtain optimized process parameters, and taking the optimized process parameters as the process parameters of the continuous casting solidification process. According to the method, optimization of technological parameters in the continuous casting solidification process can be rapidly achieved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent continuous casting process in iron and steel metallurgy, specifically involving a method and related apparatus for efficiently optimizing process parameters of continuous casting solidification process based on a generative model. Background Technology

[0002] Continuous casting is a crucial step in metallurgical production, and its complexity makes controlling billet quality defects and optimizing the production process difficult. For process optimization, numerical simulation is mainly used to construct a multi-field coupled model of the continuous casting process, including temperature, magnetic field, flow field, and stress field. This model can predict the solidification behavior of continuously cast billets and reveal the formation mechanism of billet quality defects.

[0003] To this end, those skilled in the art have designed various prediction algorithms for application in metallurgical processes, as illustrated below: Chinese patent publication number 116384160 A achieves collaborative management of parameter, algorithm, model, and result libraries for the simulation process through a unified user interface. It establishes a system architecture encompassing "equipment parameters, steel grade parameters, process parameters, model parameters, online / offline simulation prediction calculation, result display, quality monitoring / temperature control / process parameter optimization / casting machine capability development / new steel grade development," enabling simulation of temperature distribution in the casting machine and prediction of billet segregation and macroscopic microstructure characteristics. This provides analytical tools and data support for various fields such as process parameter optimization, internal quality monitoring, casting machine capability development, and new product development in continuous casting processes, demonstrating significant application prospects.

[0004] Chinese patent CN114626673A discloses a method for predicting the quality of continuously cast billets based on Bayesian networks. By analyzing the causal relationship between process parameters and billet quality, the method determines the evaluation indicators of parameters and quality, and constructs an evaluation system for the production process. Through parameter learning of the Bayesian network, qualitative analysis of each node in the continuous casting production process is achieved. By inputting each production parameter (or some production parameters) as evidence variables into the model and performing forward inference analysis through the Bayesian network, the method can predict the quality of the billets.

[0005] Chinese patent CN117009586A discloses a method for tracing the quality of continuously cast billets based on decision trees and knowledge graphs. It uses a decision tree model to summarize the process parameters involved in steel production, constructs a knowledge graph of steel quality defects based on the Neo4j graph database, and designs a model for quality tracing based on this knowledge graph. This reveals the potential and influencing factors leading to quality problems, helping to improve steel quality control and production efficiency.

[0006] Chinese patent CN115034554A discloses a method and system for predicting the quality of continuously cast billets based on time-series networks. This method solves the problem of real-time quality prediction for multivariate production data, enabling timely detection of continuously cast billets with quality issues. It prevents defective billets from entering the hot rolling production line, improves the level of intelligent manufacturing, and reduces economic losses caused by defective products.

[0007] Chinese patent CN115034554A discloses a method and system for predicting the quality of continuously cast billets based on time-series networks. This method solves the problem of real-time quality prediction for multivariate production data, enabling timely detection of continuously cast billets with quality issues. It prevents defective billets from entering the hot rolling production line, improves the level of intelligent manufacturing, and reduces economic losses caused by defective products.

[0008] Chinese patent CN108256260A discloses a method for predicting the quality of continuously cast billets based on Extreme Learning Machines (ELM). This method first collects data related to actual production of continuously cast billets to identify influencing factors affecting billet quality. Then, it preprocesses the selected sample data and determines the number of input and output nodes for the ELM based on this data. Next, it inputs the training dataset into the ELM to complete its training. Finally, it inputs the remaining sample data to classify the defect levels of the continuously cast billets. This method offers fast training speed, high prediction accuracy, and good adaptability, significantly improving both prediction accuracy and computation speed compared to other continuously cast billet quality prediction models based on statistical methods, expert systems, and backpropagation neural networks. This allows for timely and accurate assessment of the quality of continuously cast billets.

[0009] Chinese patent CN112418538A discloses a method for predicting inclusions in continuously cast billets based on random forest classification. First, it collects data on factors affecting inclusion quality in the continuously cast billet during production to construct an initial feature sample set. Then, it preprocesses the initial feature sample set, dividing it into a training set and a test set. Based on the training set data, it uses a sequence backward selection algorithm combined with a random forest algorithm to obtain the optimal feature sample dataset or optimal feature set. Next, it uses a genetic algorithm to optimize the parameters of the random forest algorithm to obtain the optimal random forest classification prediction model. Finally, it predicts inclusions in the continuously cast billet. Compared to statistical methods, expert systems, support vector machines, BP neural networks, and other continuous casting billet inclusion prediction models, this method has higher prediction accuracy and faster running speed, which is beneficial for timely and accurate determination of the quality of continuously cast billets.

[0010] Chinese patent CN115860211A discloses a method for predicting billet quality based on local online modeling, which can solve the problems of poor adaptability and unstable prediction accuracy of traditional offline global models when predicting billet quality. The method uses a support vector regression model as a local model for local online modeling to predict billet quality during continuous casting. It selects the historical dataset most relevant to the data to be tested using a similarity function. Based on the traditional Euclidean distance, it introduces process parameter feature weights and sample time feature weights to enhance the correlation between sample data and the data to be tested. Based on the sample similarity, it selects the most relevant data from the historical dataset to construct a sample dataset for local modeling.

[0011] However, the numerical simulation process of the aforementioned existing technologies often suffers from problems such as high resource consumption, long calculation time, and model instability, which frequently lead to abnormal calculation results and make it impossible to quickly provide process optimization references. Summary of the Invention

[0012] The purpose of this invention is to address the issues of slow computation speed and high resource consumption in existing technologies by providing a method and related apparatus for efficiently optimizing process parameters in continuous casting solidification based on a generative model. This invention can achieve optimization of process parameters in continuous casting solidification relatively quickly.

[0013] To achieve the above objectives, the present invention adopts the following technical solution: A method for efficiently optimizing process parameters in continuous casting solidification based on a generative model includes the following steps: Using a pre-trained simulation model of the continuous casting solidification process, the acquired process parameters of the online real-time continuous casting process are processed to obtain the predicted physical field data during the continuous casting solidification process. Extract the feature data from the predicted physical field data; The metallurgical parameters of the continuous casting process are calculated using the aforementioned feature data; Determine whether the metallurgical indicators meet the requirements of metallurgical standards. If they do, use the acquired process parameters of the online real-time continuous casting process as the process parameters of the continuous casting solidification process. Otherwise, optimize the acquired process parameters of the online real-time continuous casting process to obtain optimized process parameters, and use the optimized process parameters as the process parameters of the continuous casting solidification process.

[0014] Preferably, the process of establishing the simulation model for the continuous casting solidification process includes: In the TensorFlow deep learning framework, process parameters are mapped to the latent code layer of a convolutional autoencoder through a parametric encoder to establish a simulation model of the continuous casting solidification process with process parameters as input and physical field data of the continuous casting process as output.

[0015] Preferably, the training process of the simulation model for the continuous casting solidification process includes: A multiphysics coupling model of the continuous casting process is established; this multiphysics coupling model of the continuous casting process can be established using Ansys software. The multiphysics coupling model is used to calculate the physical field data of the continuous casting process under different process parameters, and the physical field data is used to construct a numerical simulation calculation database. The simulation model of the continuous casting solidification process is trained using the numerical simulation calculation database.

[0016] Preferably, the process also includes the following: The evaluation indicators, such as the mean square error, mean absolute error, or relative error, of the predicted physical field data and the measured physical field data during the online real-time continuous casting process are calculated. When the evaluation indicators meet the preset requirements, the simulation model of the continuous casting solidification process at this time is taken as the final trained simulation model of the continuous casting solidification process; otherwise, the simulation model of the continuous casting solidification process is retrained until the evaluation indicators of the simulation model of the continuous casting solidification process meet the preset requirements.

[0017] Preferably, the training process for the simulation model of the continuous casting solidification process includes: K-fold cross-validation was used to divide the numerical simulation database into training and validation sets. Hyperparameter tuning methods were employed to optimize the hyperparameters of the continuous casting solidification process simulation model. An early stopping strategy was introduced, where training was stopped when the validation set loss no longer decreased, resulting in a well-trained continuous casting solidification process simulation model. The hyperparameter tuning method employed either a random search method or a Bayesian optimization method. The hyperparameters of the continuous casting solidification process simulation model included the learning rate, batch size, and network layers.

[0018] Preferably, the process parameters of the acquired online real-time continuous casting process are optimized using a genetic algorithm to obtain optimized process parameters.

[0019] Preferably, the process parameters include casting speed, specific water volume, secondary cooling water temperature, secondary cooling water pressure, crystallizer electromagnetic stirring parameters, crystallizer water volume, crystallizer water temperature difference, tundish temperature, and continuous casting billet shell temperature. Physical field data includes temperature field, stress field, and solidification field; Metallurgical parameters include reheat temperature, liquid phase cavity length, air-cooled section temperature, and straightening section outlet temperature.

[0020] This invention also provides a system for efficiently optimizing process parameters of continuous casting solidification based on a generative model, used to implement the above-mentioned method of this invention, comprising: Prediction module: Used to process the acquired process parameters of the online real-time continuous casting process using a pre-trained simulation model of the continuous casting solidification process, and obtain the predicted physical field data of the continuous casting solidification process. Feature extraction module: used to extract feature data from the predicted physical field data; Calculation module: used to calculate metallurgical indicators of the continuous casting process using the aforementioned feature data; Optimization module: Used to determine whether the metallurgical indicators meet the requirements of metallurgical standards. If the requirements are met, the process parameters of the acquired online real-time continuous casting process are used as the process parameters of the continuous casting solidification process; otherwise, the process parameters of the acquired online real-time continuous casting process are optimized to obtain optimized process parameters, which are then used as the process parameters of the continuous casting solidification process.

[0021] The present invention also provides an electronic device, comprising: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method described above for efficiently optimizing process parameters of continuous casting solidification process based on generative models.

[0022] The present invention also provides a storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the method described above for efficiently optimizing process parameters of continuous casting solidification process based on a generative model.

[0023] The present invention has the following beneficial effects: This invention utilizes a pre-trained simulation model of the continuous casting solidification process to process online real-time process parameters and acquire predicted physical field data. This eliminates the need for complex multi-field coupled numerical simulations, significantly reducing computational resource consumption, shortening data acquisition time, and avoiding anomalies caused by real-time calculation instability. This ensures the timeliness and reliability of process optimization references. By extracting physical field data features and calculating metallurgical indicators, this invention can quickly focus on core indicators affecting billet quality. By determining whether these metallurgical indicators meet metallurgical criteria, it can rapidly identify whether current process parameters are compliant, achieving efficient evaluation of process rationality. This solves the problem of traditional methods failing to promptly determine process applicability due to computational lag. Furthermore, this invention optimizes parameters and determines final process parameters in cases where criteria are not met, forming an efficient "real-time processing - rapid judgment - dynamic optimization" process. This effectively improves the controllability of the continuous casting solidification process, reducing billet quality defects and avoiding resource waste through precise optimization, achieving a balance between high-quality production and high efficiency. In summary, this invention accurately solves the problems of high resource consumption, long calculation time, and model instability in existing numerical simulations that prevent the rapid provision of process optimization references. Attached Figure Description Figure 1 This is a temperature field and solidification field prediction route diagram for the solidification process of continuously cast billets in an embodiment of the present invention; Figure 2 This is a flowchart of step (4) in an embodiment of the present invention; Figure 3 This is a temperature field cloud map obtained in an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0025] The overall concept of this invention is as follows: For different steel grades, Ansys software is used to establish a multi-physics coupling model covering electromagnetic stirring, molten steel flow, and heat transfer. This model accurately simulates the distribution of the physical field (including temperature field, stress field, and solidification field) of the continuously cast billet under different process parameters, constructing a high-confidence simulation database (i.e., a numerical simulation database). This database is then imported into the TensorFlow framework, where a convolutional autoencoder extracts the spatiotemporal features of the physical field. Simultaneously, a parametric encoder embeds process parameters (including casting speed, specific water volume, secondary cooling water temperature, secondary cooling water pressure, crystallizer electromagnetic stirring parameters, crystallizer water volume, crystallizer water temperature difference, and tundish temperature) into the latent code layer, establishing a mapping model from process parameters to key physical fields (i.e., a continuous casting solidification process simulation model). A K-fold algorithm is then used. Cross-validation was used to divide the dataset, and Bayesian optimization was used to adjust the network hyperparameters. An early-stop mechanism was introduced to train the continuous casting solidification process simulation model to ensure its robustness. A real-time sensor system was deployed to collect process parameter data (including billet and shell temperature data) online, and the prediction accuracy of the continuous casting solidification process simulation model was verified by indicators such as mean square error. Based on the physical field (including temperature field and solute field) data output by the continuous casting solidification process simulation model, metallurgical indicators (including recovery temperature and liquidus cavity length) were calculated. When the metallurgical indicators deviated from the metallurgical criteria, a genetic algorithm was triggered to dynamically optimize the process parameters, and the optimized process parameters were used as the actual continuous casting solidification process parameters. When the metallurgical indicators met the metallurgical criteria, the process parameters measured by the sensor system were used as the actual continuous casting solidification process parameters.

[0026] Ultimately, the model trained by this invention (i.e., the simulation model of the continuous casting solidification process) can be connected to the real-time data stream of the production line to output process optimization schemes that conform to metallurgical standards online, forming a "prediction-evaluation-optimization" closed-loop control system.

[0027] Specifically, refer to Figure 1 and Figure 2 The present invention provides a method for efficiently optimizing process parameters in continuous casting solidification based on a generative model, comprising the following steps: Step (1): Ansys software is used to establish a multi-physics coupling model integrating "electromagnetic stirring - molten steel flow - heat transfer" for the entire continuous casting process. The multi-physics coupling model is used to simulate and calculate the physical field data (including temperature field, stress field and solidification field) under multiple sets of different process parameters (including casting speed, specific water volume, secondary cooling water temperature, secondary cooling water pressure, crystallizer electromagnetic stirring parameters, crystallizer water volume, crystallizer water temperature difference, tundish temperature). The process parameters and physical field data are used to construct a numerical simulation calculation database.

[0028] Step (2) involves applying the constructed numerical simulation database to the TensorFlow deep learning framework. Combining a convolutional autoencoder and a parametric encoding model, the spatiotemporal characteristics of the physical field data are captured. Process parameters such as casting speed, specific water volume, secondary cooling water temperature, secondary cooling water pressure, crystallizer electromagnetic stirring parameters, crystallizer water volume, crystallizer water temperature difference, and tundish temperature are mapped to the latent code layer of the convolutional autoencoder via the parametric encoder. This enables the prediction of the temperature field, stress field, and solidification field of the continuous casting process using the input process parameters. In summary, this invention utilizes the numerical simulation database within the TensorFlow deep learning framework to map process parameters to the latent code layer of the convolutional autoencoder, establishing a continuous casting solidification process simulation model with process parameters as input and physical field data of the entire continuous casting process as output.

[0029] During the training of the simulation model for the continuous casting solidification process, K-fold cross-validation is used to divide the numerical simulation database into training and validation sets to avoid overfitting. Hyperparameter tuning methods such as random search or Bayesian optimization are employed to optimize hyperparameters such as the learning rate, batch size, and number of network layers in the continuous casting solidification process simulation model. Simultaneously, an early stopping strategy is introduced, stopping training when the validation set loss no longer decreases, thereby improving training efficiency and ensuring the predictive accuracy and robustness of the continuous casting solidification process simulation model.

[0030] Step (3): The continuous casting process parameters, such as casting speed, specific water volume, secondary cooling water temperature, secondary cooling water pressure, crystallizer electromagnetic stirring parameters, crystallizer water volume, crystallizer water temperature difference, and tundish temperature, are acquired online in real time through sensors. These process parameter data are processed through the continuous casting solidification process simulation model obtained in step (2) to obtain the predicted temperature field, stress field, and solidification field. The temperature prediction results of the temperature measurement point calculated by the continuous casting solidification process simulation model (referred to as the predicted temperature value of the temperature measurement point) are compared with the actual measurement data of the temperature measurement point (referred to as the actual measured temperature value of the temperature measurement point). The mean square error, average absolute error, or relative error of the temperature measurement point prediction value and the actual measured temperature value of the temperature measurement point are calculated to evaluate the performance and reliability of the model. If the evaluation indicators meet the requirements, step (4) is performed. Otherwise, the continuous casting solidification process simulation model is retrained until the evaluation indicators of the continuous casting solidification process simulation model meet the requirements.

[0031] Step (4): By combining the radius of the continuous casting machine and the coordinate points of the simulation model of the continuous casting solidification process, extract the characteristic data of the temperature field, solidification field, and stress field predicted by the simulation model of the continuous casting solidification process in step (3). Calculate the reheat temperature, liquid cavity length, and measured air-cooling section temperature and straightening section outlet temperature of the entire continuous casting process using the aforementioned characteristic data. Determine whether the metallurgical indicators meet the requirements of the metallurgical criteria. If they do, use the process parameter data obtained online in step (3) through sensors, such as casting speed, superheat, crystallizer vibration, and continuous casting billet shell temperature, as the process parameters of the continuous casting machine. Otherwise, optimize the process parameter data obtained online in step (3) through sensors, such as casting speed, superheat, crystallizer vibration, and continuous casting billet shell temperature, as the process parameters of the continuous casting machine. See [link to relevant documentation]. Figure 2 Specifically, the process involves setting target values ​​for metallurgical indicators such as return temperature, liquid phase cavity length, and target temperature at the measuring point. Then, a genetic algorithm is used to optimize parameters including casting speed, specific water volume, secondary cooling water temperature, secondary cooling water pressure, crystallizer electromagnetic stirring parameters, crystallizer water volume, crystallizer water temperature difference, and tundish temperature. These optimized parameters are then used as the process parameters for the continuous casting machine. The optimized process parameters are then used to adjust the continuous casting machine's process parameters online. When using a genetic algorithm to optimize continuous casting process parameters, the first step is to define the process parameter search bounding box, initialize the genetic algorithm population, evaluate the fitness of individual populations, and determine if the maximum number of iterations has been reached. If not, the genetic algorithm calculation is performed; if it has, the optimal combination of process parameters is obtained.

[0032] Example This embodiment describes a method for efficiently optimizing process parameters in continuous casting solidification based on a generative model, comprising the following steps: For 20# steel, with a continuous casting machine radius of 12m and a billet diameter of 210mm, 310 sets of process data from a steel plant over the past year were collected to calculate thermophysical parameters, followed by numerical simulation calculations. The specific range of continuous casting process parameters is shown in Table 1. Table 1

[0033] 310 sets of numerical calculation data were obtained using a high-performance computer, and the process parameters and physical field data were used to construct a numerical simulation calculation database. The algorithm in step (2) was used to train 310 sets of data to obtain a deep learning-based simulation model of the continuous casting solidification process. The reliability of the deep learning-based simulation model was verified using data not used in the model training. Specific input parameters included: casting speed 1.30 m / min, specific water volume 0.3 L / kg, secondary cooling water temperature 29℃, secondary cooling water pressure 0.9 MPa, crystallizer electromagnetic stirring parameters 200A, 2HZ, and crystallizer water volume 110 m³ / min. 3With a crystallizer water temperature difference of 6.5℃ and an tundish temperature of 1540℃, the predicted outlet temperatures of the air-cooling and straightening sections differ from the actual measured values ​​by 6℃ and 5.2℃, respectively, meeting production requirements. The error cloud diagram obtained from the numerical simulation is shown below. Figure 3 As shown, the two-dimensional continuous casting process differs by 8.5℃, and the crystallizer outlet temperature differs by 7.4℃, verifying the reliability of the deep learning-based simulation model for the continuous casting solidification process. By combining the continuous casting machine radius and the coordinates of the simulation model, the reheat temperature is extracted to be 42℃, the liquidus cavity length is 9m, and the measured outlet temperatures of the air-cooling section and straightening section are 1020℃ and 905℃ respectively, which meet the metallurgical criteria. If the metallurgical criteria are not met, such as... Figure 2 As shown in the technical roadmap, a genetic algorithm is used to calculate and find the optimal process parameters for the following parameters: casting speed, specific water volume, secondary cooling water temperature, secondary cooling water pressure, crystallizer electromagnetic stirring parameters, crystallizer water volume, crystallizer water temperature difference, and tundish temperature.

[0034] Furthermore, embodiments of the present invention also provide a system for implementing the above-described method for efficiently optimizing process parameters of continuous casting solidification based on a generative model, the system comprising: Prediction module: Used to process the acquired process parameters of the online real-time continuous casting process using a pre-trained simulation model of the continuous casting solidification process, and obtain the predicted physical field data of the continuous casting solidification process. Feature extraction module: used to extract feature data from the predicted physical field data; Calculation module: used to calculate metallurgical indicators of the continuous casting process using the aforementioned feature data; Optimization module: Used to determine whether the metallurgical indicators meet the requirements of metallurgical standards. If the requirements are met, the process parameters of the acquired online real-time continuous casting process are used as the process parameters of the continuous casting solidification process; otherwise, the process parameters of the acquired online real-time continuous casting process are optimized to obtain optimized process parameters, which are then used as the process parameters of the continuous casting solidification process.

[0035] The embodiments of the present invention also provide corresponding electronic devices and computer-readable storage media for implementing the solutions provided in the embodiments of the present invention.

[0036] The electronic device includes a storage device and one or more processors. The storage device stores instructions or code, and the processors execute the instructions or code to enable the device to perform the method for efficiently optimizing process parameters of continuous casting solidification based on a generative model, as described in any embodiment of this application.

[0037] The storage medium stores a computer program, which, when executed by a processor, implements the method for efficiently optimizing process parameters of continuous casting solidification based on a generative model, as described in any embodiment of this application.

[0038] As can be seen from the above scheme, this invention collaboratively constructs a multi-physics coupling model and a deep learning framework, realizing the prediction and dynamic optimization of key physical fields in the continuous casting process. The convolutional autoencoder model trained on Ansys high-fidelity simulation data can accurately map the complex nonlinear relationship between process parameters and physical field data; the model's prediction accuracy is verified in real time using online sensor data, and key metallurgical indicators are automatically calculated. When the predicted value deviates from metallurgical criteria, a genetic algorithm is used to dynamically optimize the process parameters, forming a "prediction-evaluation-optimization" closed-loop control system. This improves the controllability of the continuous casting billet solidification process, effectively reducing defects such as internal cracks and segregation caused by temperature gradient runaway, while intelligently optimizing process parameters to reduce production energy consumption, providing reliable technical support for high-quality continuous casting billet production. In summary, this invention can accurately assess and predict the continuous casting solidification process, achieving real-time monitoring of the continuous casting process, aiming to improve the intelligence level and production efficiency of the continuous casting process.

[0039] Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort should fall within the scope of protection of the present invention.

[0040] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for efficiently optimizing process parameters in continuous casting solidification based on a generative model, characterized in that, The process includes the following: Using a pre-trained simulation model of the continuous casting solidification process, the acquired process parameters of the online real-time continuous casting process are processed to obtain the predicted physical field data during the continuous casting solidification process. Extract the feature data from the predicted physical field data; The metallurgical parameters of the continuous casting process are calculated using the aforementioned feature data; Determine whether the metallurgical indicators meet the requirements of the metallurgical criteria. If they do, then the process parameters of the online real-time continuous casting process that have been obtained shall be used as the process parameters of the continuous casting solidification process. Otherwise, the process parameters of the acquired online real-time continuous casting process are optimized to obtain optimized process parameters, which are then used as the process parameters for the continuous casting solidification process.

2. The method for efficiently optimizing process parameters of continuous casting solidification process based on a generative model according to claim 1, characterized in that, The process of establishing the simulation model for the continuous casting solidification process includes: In the TensorFlow deep learning framework, process parameters are mapped to the latent code layer of a convolutional autoencoder through a parametric encoder to establish a simulation model of the continuous casting solidification process with process parameters as input and physical field data of the continuous casting process as output.

3. The method for efficiently optimizing process parameters of continuous casting solidification process based on generative model according to claim 1, characterized in that, The training process of the simulation model for the continuous casting solidification process includes: Establish a multi-physics coupling model for the continuous casting process; The multiphysics coupling model is used to calculate the physical field data of the continuous casting process under different process parameters, and the physical field data is used to construct a numerical simulation calculation database. The simulation model of the continuous casting solidification process is trained using the numerical simulation calculation database.

4. The method for efficiently optimizing process parameters of continuous casting solidification process based on a generative model according to claim 3, characterized in that, It also includes the following processes: The evaluation indicators, such as the mean square error, mean absolute error, or relative error, of the predicted physical field data and the measured physical field data during the online real-time continuous casting process are calculated. When the evaluation indicators meet the preset requirements, the simulation model of the continuous casting solidification process at this time is taken as the final trained simulation model of the continuous casting solidification process; otherwise, the simulation model of the continuous casting solidification process is retrained until the evaluation indicators of the simulation model of the continuous casting solidification process meet the preset requirements.

5. A method for efficiently optimizing process parameters of continuous casting solidification process based on a generative model, as described in claim 3 or 4, characterized in that, The training process for the simulation model of the continuous casting solidification process includes: K-fold cross-validation was used to divide the numerical simulation database into training and validation sets. Hyperparameter tuning methods were employed to optimize the hyperparameters of the continuous casting solidification process simulation model. An early stopping strategy was introduced, where training was stopped when the validation set loss no longer decreased, resulting in a well-trained continuous casting solidification process simulation model. The hyperparameter tuning method employed either a random search method or a Bayesian optimization method. The hyperparameters of the continuous casting solidification process simulation model included the learning rate, batch size, and network layers.

6. The method for efficiently optimizing process parameters of continuous casting solidification process based on generative model according to claim 1, characterized in that, The process parameters of the acquired online real-time continuous casting process are optimized using a genetic algorithm to obtain optimized process parameters.

7. The method for efficiently optimizing process parameters of continuous casting solidification process based on a generative model according to claim 1, characterized in that, The process parameters include casting speed, specific water volume, secondary cooling water temperature, secondary cooling water pressure, crystallizer electromagnetic stirring parameters, crystallizer water volume, crystallizer water temperature difference, tundish temperature, and continuous casting billet shell temperature. Physical field data includes temperature field, stress field, and solidification field; Metallurgical parameters include reheat temperature, liquid phase cavity length, air-cooled section temperature, and straightening section outlet temperature.

8. A system for efficiently optimizing process parameters of continuous casting solidification process based on a generative model, characterized in that, include: Prediction module: Used to process the acquired process parameters of the online real-time continuous casting process using a pre-trained simulation model of the continuous casting solidification process, and obtain the predicted physical field data of the continuous casting solidification process. Feature extraction module: used to extract feature data from the predicted physical field data; Calculation module: used to calculate metallurgical indicators of the continuous casting process using the aforementioned feature data; Optimization module: used to determine whether the metallurgical indicators meet the requirements of metallurgical standards. If the requirements are met, the process parameters of the acquired online real-time continuous casting process are used as the process parameters of the continuous casting solidification process. Otherwise, the process parameters of the acquired online real-time continuous casting process are optimized to obtain optimized process parameters, which are then used as the process parameters for the continuous casting solidification process.

9. An electronic device, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for efficiently optimizing process parameters of continuous casting solidification process based on a generative model as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, It stores a computer program, wherein when the computer program is executed by a processor, it implements the method for efficiently optimizing process parameters of continuous casting solidification process based on a generative model as described in any one of claims 1-7.

Images