Molten steel segregation prediction method and related equipment

Abstract

The invention discloses a molten steel segregation prediction method and related equipment, and relates to the technical field of continuous casting, and the method comprises the following steps: obtaining process parameter data of electromagnetic stirring in a continuous casting process and corresponding molten steel segregation data; preprocessing the process parameter data and the molten steel segregation data to obtain a training data set for model training and a test data set for model verification; adopting a grey wolf algorithm to optimize hyper-parameters of the random forest model to obtain an optimized random forest model; training the optimized random forest model by using the training data set, and verifying the prediction performance of the optimized random forest model through the test data set; on the basis of the trained and verified optimized random forest model, real-time process parameter data are input, and a prediction result of the molten steel segregation degree is output; and optimizing and adjusting process parameter data of electromagnetic stirring according to the prediction result.

Description

Technical Field

[0001] The present application relates to the technical field of continuous casting, and particularly relates to a method for predicting molten steel segregation and related equipment. Background Art

[0002] During the continuous casting process, the segregation phenomenon of molten steel is one of the important factors affecting the quality of steel. Due to the uneven distribution of solute elements during solidification, concentration gradients are formed in different parts of the slab, resulting in macroscopic segregation and microscopic segregation. The electromagnetic stirring technology drives the flow of molten steel through an electromagnetic field, which can effectively reduce the segregation phenomenon and improve the microscopic structure and mechanical properties of the casting. However, the existing methods for optimizing electromagnetic stirring parameters mainly rely on empirical formulas and numerical simulations, making it difficult to predict and adjust process parameters in real time, resulting in inaccurate segregation control and large fluctuations in the quality of steel in actual production. In addition, traditional mathematical models are difficult to accurately describe the complex physical and chemical changes during continuous casting and lack the ability to predict the effect of electromagnetic stirring in real time, which limits the further application of electromagnetic stirring technology. Therefore, there is an urgent need for a method for predicting molten steel segregation to solve the above-mentioned problems. Summary of the Invention

[0003] A series of simplified concepts are introduced in the Summary of the Invention section, which will be further described in detail in the Detailed Description section. The Summary of the Invention section of the present application does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the protection scope of the claimed technical solution.

[0004] In a first aspect, the present application provides a method for predicting molten steel segregation, including:

[0005] Obtaining process parameter data of electromagnetic stirring and corresponding molten steel segregation data during continuous casting, where the process parameter data includes at least one of current intensity, frequency, coil position, coil shape, stirring time, and magnetic field direction, and the molten steel segregation data includes the average segregation index of molten steel;

[0006] Preprocessing the process parameter data and the molten steel segregation data to obtain a training data set for model training and a test data set for model verification;

[0007] Optimizing the hyperparameters of the random forest model using the grey wolf algorithm to obtain an optimized random forest model;

[0008] Training the optimized random forest model using the training data set and verifying the prediction performance of the optimized random forest model through the test data set;

[0009] Based on the optimized random forest model that has been trained and verified, by inputting real-time process parameter data, outputting a prediction result of the degree of molten steel segregation;

[0010] Optimize and adjust the process parameter data of electromagnetic stirring according to the prediction results.

[0011] In some embodiments, preprocess the process parameter data and the molten steel segregation data to obtain a training data set for model training and a test data set for model verification, including:

[0012] Detect outliers in the process parameter data and the molten steel segregation data, and if the error is within a preset range, correct it through regression analysis to obtain corrected data;

[0013] Supplement missing values in the corrected data, and fill in the missing values through interpolation to obtain supplemented data;

[0014] Standardize the supplemented data and divide it into a training data set and a test data set.

[0015] In some embodiments, use the Grey Wolf Algorithm to optimize the hyperparameters of the Random Forest model to obtain an optimized Random Forest model, including:

[0016] Initialize the Grey Wolf population and define the optimization objective function. Among them, the optimization objective function is to minimize the prediction error, and the prediction error includes the mean square error and the mean absolute error;

[0017] Update the Grey Wolf positions through iterative operations, and gradually approach the optimal hyperparameter combination during the iteration process. Among them, the hyperparameters include the number of decision trees, the maximum depth of the decision tree, and the minimum number of samples required for leaf nodes;

[0018] Output the obtained optimal hyperparameter combination, and use the optimal hyperparameter combination in the construction process of the Random Forest model to obtain an optimized Random Forest model.

[0019] In some embodiments, use the training data set to train the optimized Random Forest model, and verify the prediction performance of the optimized Random Forest model through the test data set, including:

[0020] Calculate the influence weight of each process parameter on the molten steel segregation based on the entropy value, and select the parameter with the highest influence weight as the root node of the decision tree;

[0021] Recursively divide the intermediate nodes until the stopping condition is met. The stopping conditions include that all samples in the current node belong to the same category, the attribute set is empty, or the sample set is empty;

[0022] Train the optimized Random Forest model based on the training data set to obtain the trained optimized Random Forest model;

[0023] Validate the trained optimized random forest model based on the test data set to obtain the test error and the test coefficient of determination. Among them, in the case where the test coefficient of determination is less than the first preset threshold or the test error is greater than the second preset threshold, re-optimize the hyperparameters and perform iterative training until the test coefficient of determination is greater than or equal to the first preset threshold and the test error is less than or equal to the second preset threshold.

[0024] In some embodiments, the average segregation index of the molten steel is determined based on the following formula, expressed as:

[0025]

[0026] where ASI is the average segregation index, C i is the actual concentration value at the i-th measurement point, C avg is the average concentration value of all measurement points, and n is the number of measurement points.

[0027] In some embodiments, the method further includes:

[0028] Package the optimized random forest model into a prediction software, where the prediction software is implemented in the Python language, and the prediction software includes functions of data dynamic update, model training, and visualization interface;

[0029] Receive real-time process parameter data through the visualization interface, and output the molten steel segregation prediction result and parameter adjustment suggestions to realize the monitoring and optimization of the continuous casting process.

[0030] In some embodiments, optimize and adjust the process parameter data of electromagnetic stirring according to the prediction result, including:

[0031] Based on the prediction result of the optimized random forest model, generate an optimized adjustment strategy for electromagnetic stirring parameters. Among them, the optimized adjustment strategy includes adjusting the current intensity and frequency to enhance the electromagnetic stirring effect in the case where the predicted average segregation index of the molten steel exceeds the preset threshold; the optimized adjustment strategy also includes optimizing the coil position and stirring time to improve the flow uniformity of the molten steel in the case where the predicted average segregation index of the molten steel is lower than the preset threshold.

[0032] In a second aspect, the present application proposes a device for predicting molten steel segregation, including:

[0033] A continuous casting parameter acquisition unit, configured to acquire the process parameter data of electromagnetic stirring and the corresponding molten steel segregation data during continuous casting, where the process parameter data includes at least one of current intensity, frequency, coil position, coil shape, stirring time, and magnetic field direction, and the molten steel segregation data includes the average segregation index of the molten steel;

[0034] A data partitioning and processing unit for preprocessing process parameter data and molten steel segregation data to obtain a training data set for model training and a test data set for model verification;

[0035] A model hyperparameter tuning and optimization unit for optimizing the hyperparameters of a random forest model using the Grey Wolf algorithm to obtain an optimized random forest model;

[0036] A model training and testing unit for training the optimized random forest model using the training data set and verifying the prediction performance of the optimized random forest model through the test data set;

[0037] A model application and prediction unit, based on the optimized random forest model that has been trained and verified, outputs a prediction result of the degree of molten steel segregation by inputting real-time process parameter data;

[0038] A parameter optimization and adjustment unit for optimizing and adjusting the process parameter data of electromagnetic stirring according to the prediction result.

[0039] In a third aspect, an electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor is configured to implement the steps of the prediction method for molten steel segregation according to any one of the first aspects when executing the computer program stored in the memory.

[0040] In a fourth aspect, the present application also proposes a computer-readable storage medium having a computer program stored thereon. The computer program, when executed by a processor, implements the prediction method for molten steel segregation according to any one of the first aspects.

[0041] In summary, the present application obtains process parameter data and molten steel segregation data during continuous casting, optimizes the hyperparameters of a random forest model using the Grey Wolf algorithm, trains and verifies the model using a training data set, and finally predicts the degree of molten steel segregation in real time based on the optimized model. This method can effectively improve the prediction accuracy, guide the optimization and adjustment of electromagnetic stirring parameters in real time, reduce the phenomenon of molten steel segregation, and improve the quality and production efficiency of steel. At the same time, through data preprocessing and model optimization, this method significantly reduces the dependence of the model on sample data, enhances the generalization ability and robustness of the model, and provides reliable technical support for the intelligent control of the continuous casting process. Description of the Drawings

[0042] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to limit this specification. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0043] Figure 1 Schematic flow chart of a method for predicting molten steel segregation provided by an embodiment of the present application;

[0044] Figure 2 Schematic diagram of the physical model of electromagnetic stirring during continuous casting provided by an embodiment of the present application;

[0045] Figure 3 Screenshot of the software for predicting molten steel segregation during electromagnetic stirring in continuous casting provided by an embodiment of the present application;

[0046] Figure 4 Schematic structural diagram of a device for predicting molten steel segregation provided by an embodiment of the present application;

[0047] Figure 5 Schematic structural diagram of an electronic device for predicting molten steel segregation provided by an embodiment of the present application. Detailed implementation manners

[0048] In the description of the present application, the terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order different from that shown or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily limit to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments.

[0049] Please refer to Figure 1 , which is a schematic flow chart of a method for predicting molten steel segregation provided by an embodiment of the present application, and specifically may include:

[0050] S110. Obtain the process parameter data of electromagnetic stirring and the corresponding molten steel segregation data during continuous casting. Among them, the process parameter data includes at least one of current intensity, frequency, coil position, coil shape, stirring time, and magnetic field direction, and the molten steel segregation data includes the average segregation index of molten steel;

[0051] Exemplarily, during the continuous casting process, the electromagnetic stirring technology drives the flow of molten steel through an electromagnetic field, which can effectively reduce the phenomenon of molten steel segregation, improve the microstructure and mechanical properties of the casting. The effect of electromagnetic stirring mainly depends on the setting of process parameters, including current intensity, frequency, coil position, coil shape, stirring time, and magnetic field direction, etc. These parameters directly affect the distribution and intensity of the electromagnetic field, and further affect the flow state and solidification behavior of molten steel. By obtaining these process parameter data and their corresponding average segregation indices of molten steel, it can provide basic data support for subsequent model training and prediction, so as to achieve accurate prediction and optimized adjustment of the electromagnetic stirring effect.

[0052] The acquisition of molten steel segregation data is usually based on experimental measurements or industrial field collection, which reflects the uneven distribution degree of solute elements in molten steel. The average segregation index of molten steel is an important indicator to measure the segregation degree. By calculating the deviation between the concentration value of each measurement point and the average concentration, the segregation situation of molten steel can be quantified. Combining the process parameter data of electromagnetic stirring, a correlation model between process parameters and segregation degree can be established, providing a data basis for subsequent prediction and optimization. This step is the premise of subsequent model training and prediction, ensuring the integrity and effectiveness of the data.

[0053] S120. Preprocess the process parameter data and molten steel segregation data to obtain a training data set for model training and a test data set for model verification;

[0054] Exemplarily, preprocessing the process parameter data and molten steel segregation data is a key step to ensure data quality and model performance. In the actual production process, the data usually has problems such as missing values, anomalies, or noise, which may affect the training effect of the model. Therefore, the preprocessing process includes cleaning, correcting, and standardizing the data to ensure the accuracy and consistency of the input data. Through this process, the reliability of the training data set and the test data set can be effectively improved, making the subsequent model training more accurate.

[0055] After data preprocessing, the process parameter data and molten steel segregation data will be divided into a training data set and a test data set. The training data set is used to train the model, while the test data set is used to verify the generalization ability of the model. Through reasonable data division, the relative independence between the training set and the test set is ensured, thus avoiding the overfitting problem and ensuring the prediction ability of the model on unknown data. The optimization of the data preprocessing and division steps provides a solid foundation for subsequent model training and performance evaluation.

[0056] S130. Optimize the hyperparameters of the random forest model using the grey wolf algorithm to obtain an optimized random forest model;

[0057] Exemplarily, the Grey Wolf Algorithm is an optimization algorithm based on the hunting behavior of grey wolves in nature, which can effectively search for the optimal solution in the multi-dimensional parameter space. In the random forest model, hyperparameters such as the number of decision trees, the depth of the trees, and the minimum number of samples in the leaf nodes have a direct impact on the prediction ability of the model. Through the optimization of the Grey Wolf Algorithm, these hyperparameters can be effectively adjusted to minimize the error during the training process of the model, thereby improving the prediction accuracy of the model.

[0058] Optimizing the hyperparameters of the random forest through the Grey Wolf Algorithm can not only improve the accuracy of the model but also enhance the generalization ability of the model. Compared with traditional hyperparameter tuning methods, the Grey Wolf Algorithm can find a more ideal combination of hyperparameters within fewer iterations, avoiding the tediousness of manual parameter tuning and the computational cost. Therefore, the random forest model optimized by the Grey Wolf Algorithm can more accurately capture the relationship between process parameters and the degree of segregation when facing complex molten steel segregation prediction tasks.

[0059] S140. Use the training dataset to train the optimized random forest model and verify the prediction performance of the optimized random forest model through the test dataset;

[0060] Exemplarily, the optimized random forest model will continuously adjust the weights and decision rules inside the model according to the input process parameters and the corresponding molten steel segregation data in the training dataset, thereby gradually fitting the relationship between the process parameters and the molten steel segregation. This training process ensures that the model can achieve a lower error in the training data and form a reliable prediction pattern through multiple iterations.

[0061] After the training is completed, the test dataset is used to verify the performance of the optimized random forest model. The test dataset usually consists of data that did not participate in the training and is used to evaluate the prediction ability of the model on unknown data. By comparing the differences between the prediction results of the model and the actual test data, the prediction error and the coefficient of determination (R 2 value) of the model can be obtained, thereby verifying the effectiveness and accuracy of the model. If the test results show that the prediction performance of the model meets the expected standards, it can be confirmed that the model has successfully completed the training and has a high generalization ability, and can accurately predict the degree of molten steel segregation under real-time process parameters.

[0062] S150. Based on the optimized random forest model that has been trained and verified, by inputting real-time process parameter data, output the prediction results of the degree of molten steel segregation to optimize and adjust the process parameter data of electromagnetic stirring.

[0063] Exemplarily, based on the optimized random forest model that has been trained and passed verification, the system can take the process parameter data collected in real time as input, process it through the model, and obtain the prediction result of the degree of molten steel segregation. The core of this process is to use the trained model to quickly and accurately predict new process parameters, thereby helping the operator make timely decisions during the actual production process. This prediction result provides a basis for subsequent optimization and adjustment to ensure that the degree of molten steel segregation during the production process is controlled within the expected range.

[0064] S160. Optimize and adjust the process parameter data of electromagnetic stirring according to the prediction result.

[0065] Exemplarily, according to the prediction result of molten steel segregation output by the optimized random forest model, the process parameters of electromagnetic stirring can be further precisely adjusted. Through the dynamic feedback of real-time process parameter data, the system can identify the change trend of the segregation degree and suggest timely adjustment of relevant parameters, such as current intensity, frequency, or stirring time, etc., to optimize the fluidity and uniformity of molten steel. This real-time optimization and adjustment can not only improve production efficiency but also significantly improve the quality of molten steel, ensuring the uniformity and stability of the final casting product.

[0066] In summary, the molten steel segregation prediction method proposed in this application uses a random forest model optimized by the grey wolf algorithm. By inputting the process parameter data of electromagnetic stirring during the continuous casting process in real time, it can efficiently and accurately predict the degree of molten steel segregation. By optimizing the hyperparameters of the random forest model and combining the entropy method to evaluate the influence weights of process parameters, key parameters are selected for adjustment, significantly improving the prediction accuracy and stability. In addition, this application can automatically optimize process parameters through a real-time feedback and adaptive adjustment mechanism, achieve precise control of molten steel segregation, and ultimately improve the quality and production efficiency of steel.

[0067] In some instances, as Figure 2 shown, a physical model of electromagnetic stirring during continuous casting is presented, demonstrating the key production links of the steel billet in the secondary cooling zone. During continuous casting, the molten steel enters the secondary cooling zone after being initially solidified in the mold. There are three pairs of electromagnetic stirring rolls distributed in this area, namely the first pair, the second pair, and the third pair, which are arranged in sequence along the running direction of the steel billet and continuously apply a stirring effect to the moving steel billet.

[0068] From Figure 2On the right side, the layout of electromagnetic stirring rollers on the inner and outer arcs of the billet can be seen. Multiple groups of electromagnetic stirring rollers are provided on both the inner and outer arcs and are symmetrically distributed relative to the billet. This design utilizes the principle of electromagnetic induction. The electromagnetic stirring rollers generate an alternating magnetic field through the internal electromagnetic coils. When the billet passes through this alternating magnetic field region, induced current will be generated in the molten steel within the billet. According to the principle of electromagnetic induction, the induced current interacts with the alternating magnetic field to generate an electromagnetic force, prompting the molten steel to move and achieving the stirring effect. By reasonably setting parameters such as the current intensity, frequency, and magnetic field direction of the electromagnetic stirring rollers, the stirring intensity and method of the molten steel can be precisely controlled, reducing molten steel segregation, refining grains, and improving the quality of castings to meet the requirements of different steel grades and production processes.

[0069] In some instances, preprocess the process parameter data and molten steel segregation data to obtain a training data set for model training and a test data set for model verification, including:

[0070] Perform outlier detection on the process parameter data and molten steel segregation data. If the data error is within the preset range, correct it through regression analysis to obtain corrected data;

[0071] Supplement missing values for the corrected data, fill in the missing values through interpolation method to obtain supplemented data;

[0072] Standardize the supplemented data and divide it into a training data set and a test data set.

[0073] Exemplarily, during the prediction process of molten steel segregation, the quality and accuracy of data are key factors affecting the model performance. Perform outlier detection on the process parameter data and molten steel segregation data to identify and exclude data points that do not conform to physical laws or statistical characteristics. Outliers are usually caused by sensor failures, data transmission errors, or external environmental interferences. If not corrected, they will seriously affect the training and prediction accuracy of the model. Therefore, in the preprocessing stage, screen the data by setting a preset range to ensure the elimination of these abnormal data, thereby ensuring the reliability of the data set.

[0074] For data with errors within the preset range, use regression analysis for correction to restore the rationality of the data. Regression analysis is a commonly used statistical method that can utilize existing relevant data to infer the reasonable range of missing values or outliers. The corrected data obtained through correction can, to the greatest extent, retain the original trend of the data, avoiding excessive deviations or misleading the model training process. This process helps to improve the accuracy and consistency of the data, providing more reliable input for subsequent data analysis and model training.

[0075] Fill in the missing values in the corrected data. Data missing is a common phenomenon, especially in industrial production, where sensors may not be able to continuously provide complete real-time data. By using interpolation methods to fill in the missing values, these gaps can be effectively filled, avoiding incomplete or biased model training due to missing data. Interpolation methods usually include linear interpolation, spline interpolation, etc., and a suitable filling method is selected according to the characteristics of the data. To improve the efficiency and accuracy of model training, all the supplemented data will be standardized. Standardization converts data with different scales and dimensions into a unified standard range, ensuring that the impacts of various features on model training are balanced, thereby further improving the prediction accuracy of the model. After standardization, the dataset will be divided into a training dataset and a test dataset, providing effective support for subsequent model training and verification.

[0076] It should be noted that the ratio of the training dataset to the test dataset is 8:2.

[0077] In some instances, the grey wolf algorithm is used to optimize the hyperparameters of the random forest model to obtain an optimized random forest model, including:

[0078] Initialize the grey wolf population and define the optimization objective function. Among them, the optimization objective function is to minimize the prediction error, and the prediction error includes the mean square error and the mean absolute error;

[0079] Update the grey wolf positions through iterative operations, and gradually approach the optimal hyperparameter combination during the iteration process. Among them, the hyperparameters include the number of decision trees, the maximum depth of the decision tree, and the minimum number of samples required for leaf nodes;

[0080] Output the obtained optimal hyperparameter combination, and use the optimal hyperparameter combination in the construction process of the random forest model to obtain an optimized random forest model.

[0081] Exemplarily, the Grey Wolf Algorithm simulates the predation behavior of a grey wolf population to perform optimization search. When optimizing the hyperparameters of a random forest model, the grey wolf population is first initialized. Each grey wolf individual represents a combination of hyperparameters of a random forest model in the search space, such as combinations of different values of the number of decision trees, the maximum depth of decision trees, and the minimum number of samples required for leaf nodes. By initializing a certain number of grey wolf individuals, an initial search population of hyperparameter combinations is formed, providing a starting point for subsequent optimization search. At the same time, the optimization objective function is defined as minimizing the prediction error, where the prediction error includes the mean squared error and the mean absolute error. The mean squared error calculates the average of the squares of the differences between the predicted values and the true values, giving greater weight to larger errors and being able to more sensitively reflect the deviation between the predicted values and the true values; the mean absolute error is the average of the absolute values of the differences between the predicted values and the true values, relatively more intuitively measuring the average size of the prediction error. Taking the minimization of these two errors as the optimization objective aims to make the prediction results of the random forest model as close as possible to the true values and improve the prediction accuracy of the model.

[0082] After initializing the grey wolf population and defining the optimization objective function, the grey wolf positions are updated through iterative operations to gradually approach the optimal hyperparameter combination. In each iteration, the grey wolf population simulates the hierarchical system and cooperative behavior in the actual predation process. The alpha wolf, beta wolf, and delta wolf in the leading positions lead the population to move towards the direction of the possible optimal solution, while the omega wolf follows their actions. In the search space, the update of the grey wolf positions is calculated based on the current positions and the position information of the alpha wolf, beta wolf, and delta wolf. By continuously adjusting their own positions, grey wolf individuals explore different regions of hyperparameter combinations in the search space. As the number of iterations increases, the grey wolf population gradually converges to the region where the optimal hyperparameter combination is located in the search space, just like gradually approaching the prey in the predation process. This iterative update mechanism enables the Grey Wolf Algorithm to efficiently find the optimal solution in the complex hyperparameter search space and continuously optimize the hyperparameters of the random forest model.

[0083] After multiple rounds of iteration, when certain stopping conditions are met (such as reaching a preset maximum number of iterations or the value of the optimized objective function converges to a certain extent), the Grey Wolf Algorithm outputs the obtained optimal hyperparameter combination. This optimal hyperparameter combination represents the parameter settings that are most likely to enable the random forest model to achieve the best prediction performance under the current search space and optimization objective. Apply this optimal hyperparameter combination to the construction process of the random forest model. When constructing the model, determine key parameters such as the number of decision trees, the maximum depth of each decision tree, and the minimum number of samples required for leaf nodes based on these optimal parameters. By using the optimized hyperparameters to construct the random forest model, the model can better capture the internal laws in the data when processing process parameter data and molten steel segregation data during the continuous casting process, improve the generalization ability and prediction accuracy of the model, and thus obtain an optimized random forest model that can more accurately predict the degree of molten steel segregation.

[0084] In some instances, the optimized random forest model is trained using the training dataset, and the prediction performance of the optimized random forest model is verified through the test dataset, including:

[0085] Calculate the influence weight of each process parameter on molten steel segregation based on entropy value, and select the parameter with the highest influence weight as the root node of the decision tree;

[0086] Recursively divide the intermediate nodes until the stopping conditions are met. The stopping conditions include that all samples of the current node belong to the same category, the attribute set is empty, or the sample set is empty;

[0087] Train the optimized random forest model based on the training dataset to obtain the trained optimized random forest model;

[0088] Verify the trained optimized random forest model based on the test dataset to obtain the test error and the test determination coefficient. Among them, in the case where the test determination coefficient is less than the first preset threshold or the test error is greater than the second preset threshold, re-optimize the hyperparameters and iterate the training until the test determination coefficient is greater than or equal to the first preset threshold and the test error is less than or equal to the second preset threshold.

[0089] Exemplarily, when training the optimized random forest model using the training dataset, the influence weights of each process parameter on the segregation of molten steel are calculated based on entropy values. The principle of this operation stems from information theory. Entropy is a measure of uncertainty. In this context, by calculating the entropy value of each process parameter, the amount of information it contains in describing the segregation of molten steel can be measured. The lower the entropy value of a process parameter, the higher the certainty in distinguishing different segregation states of molten steel, that is, the greater the influence weight on the segregation of molten steel. For example, if the entropy value of the process parameter of current intensity is relatively lower than other parameters, it indicates that it is more certain and discriminative in reflecting the segregation of molten steel. Selecting the parameter with the highest influence weight as the root node of the decision tree is because the selection of the root node is crucial for the construction of the decision tree. It determines the starting direction for the decision tree to partition the data. Using the parameter with the highest influence weight as the root node enables the decision tree to start classifying and regressing the data from the most discriminative dimension, thereby more effectively mining the potential relationship between process parameters and the segregation of molten steel.

[0090] After determining the root node, recursively partition the intermediate nodes until the stopping condition is met. This process is based on the construction logic of the decision tree. The decision tree gradually refines the sample space by continuously partitioning the data to achieve accurate prediction of the target variable (molten steel segregation index). During the recursive partitioning process, each partition aims to make the samples within the child nodes more consistent in terms of the segregation state of molten steel. The stopping conditions include that all samples in the current node belong to the same category, the attribute set is empty, or the sample set is empty. When all samples in the current node belong to the same category, it means that there is no need to further partition at this node because the segregation state of the samples has been completely determined; an empty attribute set means that there are no available process parameters for further partitioning; an empty sample set means that there are no more samples available for partitioning. Through such recursive partitioning and stopping condition judgment, the decision tree can construct a reasonable hierarchical structure to accurately model the relationship between process parameters and the segregation of molten steel in the training data.

[0091] The optimized random forest model is trained based on the training dataset. By learning the features and patterns in the training data, the model constructs a set of decision trees to form the ability to predict the relationship between process parameters and molten steel segregation. Subsequently, the trained optimized random forest model is verified based on the test dataset, and the prediction performance of the model is evaluated by calculating the test error and the test coefficient of determination. The test error reflects the deviation between the predicted value and the actual value of the model, while the test coefficient of determination measures the goodness of fit of the model to the test data. The closer its value is to 1, the better the fitting effect of the model to the data. When the test coefficient of determination is less than the first preset threshold or the test error is greater than the second preset threshold, it indicates that the prediction performance of the model does not meet the expectations, and there may be overfitting or underfitting problems. At this time, re-optimizing the hyperparameters and iterative training are carried out to adjust the structure and parameters of the model to make it better adapt to the data characteristics, improve the generalization ability and prediction accuracy of the model until the test coefficient of determination is greater than or equal to the first preset threshold and the test error is less than or equal to the second preset threshold, ensuring that the model can reliably predict the degree of molten steel segregation in practical applications.

[0092] It should be noted that in the embodiments of the present application, the first preset threshold can be set to 90%, and the second preset threshold can be set to 10%.

[0093] In some examples, the average segregation index of molten steel is determined based on the following formula, expressed as:

[0094]

[0095] Where ASI is the average segregation index, C i is the actual concentration value at the i-th measurement point, C avg is the average concentration value of all measurement points, and n is the number of measurement points.

[0096] Exemplarily, there are differences in the solute element concentrations at different positions inside the molten steel, and this concentration difference leads to the occurrence of segregation phenomena. This formula comprehensively reflects the distribution of solute elements in the molten steel by collecting the actual concentration values of multiple measurement points. The average concentration value of all measurement points represents the overall concentration level of the molten steel and is a reference value. Taking the absolute value of the difference between the actual concentration value of each measurement point and the average concentration value can intuitively reflect the degree of deviation of this measurement point from the overall average level. Summing up and averaging this degree of deviation for all measurement points, that is, dividing by the number of measurement points, the obtained average segregation index can comprehensively reflect the overall situation of molten steel segregation. The larger the index, the greater the difference between the concentrations of each measurement point in the molten steel and the average concentration, that is, the more serious the degree of molten steel segregation; on the contrary, the smaller the index, the lighter the degree of segregation.

[0097] From a fundamental level of analysis, the design purpose of this formula is to quantify the complex segregation phenomenon in molten steel. Through precise concentration measurement and mathematical calculation, the degree of segregation that is difficult to visually evaluate is converted into specific numerical values. In actual continuous casting production, this numerical value provides an objective measurement standard. With the help of this standard, the influence degree of different process parameters (such as electromagnetic stirring parameters like current intensity and frequency) on the segregation of molten steel can be clearly understood. When studying the effect of electromagnetic stirring on the segregation of molten steel, researchers can compare the average segregation index of molten steel under different electromagnetic stirring parameters, so as to determine the optimal combination of process parameters to achieve the goal of reducing the segregation of molten steel and improving the quality of castings.

[0098] In some examples, the method further includes:

[0099] Encapsulating the optimized random forest model into a prediction software, where the prediction software is implemented through the Python language, and the prediction software includes functions of data dynamic update, model training, and visualization interface;

[0100] Receiving real-time process parameter data through the visualization interface, and outputting the prediction result of molten steel segregation and parameter adjustment suggestions to achieve the monitoring and optimization of the continuous casting process.

[0101] Exemplarily, encapsulating the optimized random forest model into a prediction software and implementing it using the Python language is based on the strong support of the Python language for data analysis and machine learning libraries. Python has rich toolkits such as Scikit-learn, which can conveniently call the trained optimized random forest model and encapsulate it. Integrating the data dynamic update function is considering that the process parameters and the segregation situation of molten steel in the continuous casting production process may change with time and production conditions. By obtaining new data in real time, the model can continuously update the knowledge it has learned to adapt to the new production situation and maintain the accuracy of prediction. The integration of the model training function allows users to retrain and optimize the model when there is a large amount of new data accumulation or the model performance deteriorates, improving the adaptability and prediction ability of the model. The visualization interface function is for facilitating the interaction between the operator and the software. It presents complex data and model prediction results in an intuitive form, such as Figure 3 the interface of the prediction software for the segregation of molten steel stirred by electromagnetic stirring in the continuous casting process as shown, and the operator can clearly see the parameters to be input and the results to be obtained.

[0102] The visualization interface plays a key role in connecting the operator and the model in the software. From Figure 3It can be seen that an input parameter area is provided on the interface, where operators can input the real-time process parameter data of electromagnetic stirring during continuous casting, including frequency, current intensity, coil position, coil shape, stirring time, magnetic field direction, etc. These parameter data are transmitted to the encapsulated optimized random forest model through the interface inside the software. Based on the relationship between the process parameters learned during its training and the segregation of molten steel, the model analyzes and predicts the input data, generating a prediction result of molten steel segregation. The prediction result is then fed back to the visualization interface through the output module of the software and presented to the operator in an intuitive form.

[0103] By receiving process parameter data in real time, outputting prediction results and parameter adjustment suggestions, the prediction software realizes the monitoring and optimization of the continuous casting process. In terms of monitoring, operators can, according to the prediction result of molten steel segregation output by the software, grasp the segregation status of molten steel during continuous casting in real time and promptly discover potential quality problems. In terms of optimization, when the prediction result shows that the segregation of molten steel may exceed the target range, the parameter adjustment suggestions provided by the software can guide operators to adjust the process parameters of electromagnetic stirring, such as changing the current intensity, frequency, etc., to improve the stirring effect of molten steel, reduce segregation and improve the quality of castings. By continuously cycling through this process of data input, prediction, suggestion and adjustment, the software can continuously optimize the continuous casting process, ensuring the stability of the production process and the reliability of product quality.

[0104] In some instances, the process parameter data of electromagnetic stirring are optimized and adjusted according to the prediction result, including:

[0105] Based on the prediction result of the optimized random forest model, an optimization adjustment strategy for electromagnetic stirring parameters is generated. Among them, the optimization adjustment strategy includes adjusting the current intensity and frequency to enhance the electromagnetic stirring effect when the predicted average segregation index of molten steel exceeds the preset threshold; the optimization adjustment strategy also includes optimizing the coil position and stirring time to improve the flow uniformity of molten steel when the predicted average segregation index of molten steel is lower than the preset threshold.

[0106] Exemplarily, the average segregation index of molten steel is a key quantitative index to measure the segregation degree of molten steel, which reflects the uniformity of the distribution of solute elements in molten steel. Through learning various process parameter data and molten steel segregation data during continuous casting, the optimized random forest model can accurately predict the average segregation index of molten steel. When the predicted average segregation index of molten steel exceeds the preset threshold, it means that the segregation degree of molten steel is relatively serious, and at this time the distribution of solute elements in molten steel is extremely uneven. Adjusting the current intensity and frequency can change the magnetic field intensity and variation frequency generated by electromagnetic stirring, thereby enhancing the stirring effect of electromagnetic force on molten steel. Stronger electromagnetic stirring can promote the more thorough mixing of molten steel, break the aggregation area of solute elements, and make solute elements more evenly distributed in molten steel, thus effectively reducing the segregation degree of molten steel.

[0107] When the predicted average segregation index of the molten steel is lower than the preset threshold, it indicates that the degree of molten steel segregation is relatively light, but there is still room for optimization to further improve the uniformity of molten steel flow. The coil position determines the distribution of the magnetic field generated by electromagnetic stirring in the molten steel. By optimizing the coil position, the action area and intensity distribution of the magnetic field can be adjusted, so that each part of the molten steel receives a more reasonable electromagnetic force during solidification, promoting the uniform flow of the molten steel and reducing the possible micro-segregation in local areas. The stirring time directly affects the duration of the stirring action on the molten steel. Appropriately adjusting the stirring time can allow the molten steel to have more sufficient time for mixing and homogenization, ensuring that solute elements are more fully diffused and distributed in the molten steel, further improving the uniformity of the molten steel, and achieving the purpose of improving the quality of castings.

[0108] The technical solution of the present application will be further described in detail below through specific embodiments.

[0109] In the experiment of predicting molten steel segregation during continuous casting, specific experimental conditions and objectives were set. The experimental conditions cover electromagnetic stirring related parameters such as current intensity, frequency, coil position, coil shape, stirring time, and magnetic field direction. The observation objective of the experiment is the average segregation index of the molten steel, which is a key indicator to measure the degree of molten steel segregation. During the process of predicting element segregation, it is stipulated that when the relative error between the predicted value and the experimental value of the element segregation index is less than 15%, it is considered an effective prediction. Through 300 groups of test data, it is found that 286 groups of test groups have effective predictions, indicating that the prediction accuracy rate of the algorithm reaches 95.333%. After debugging the algorithm parameters, predictions are made before production. By adjusting the input parameters, the segregation index of a specific steel grade will be made to reach an ideal value. For example, the average segregation index of carbon element is 0.3. When the segregation index is greater than 0.3, it indicates that the carbon concentration at each measurement point is too large, that is, the more serious the segregation of the molten steel. Therefore, the predicted segregation value should be lower than 0.3. If the predicted value does not reach the ideal index, the production parameters will be adjusted according to the influence degree of the input conditions on the prediction result. If the predicted value deviates too much, the magnitude of the current intensity will be adjusted; if the predicted value deviates too little, the adjustment frequency and stirring time, etc. will be finely adjusted. The production parameters are adjusted through the predicted data, and the real-time prediction method assists in adjusting the parameters of industrial production, avoiding serious segregation of the steel samples after production is completed.

[0110] By using the entropy value attribute analysis method, the influence degree of each experimental condition parameter on the prediction result is quantitatively evaluated. The results show that the influence degree ratios of current intensity, frequency, coil position, coil shape, stirring time, and magnetic field direction are 24.6%, 17.3%, 10.1%, 13.6%, 19.63%, and 14.77% respectively. It can be seen that the current intensity and stirring time have the most significant influence on the degree of molten steel segregation, playing a decisive role, and the influence degree of frequency is the second.

[0111] Furthermore, the random forest prediction algorithm optimized by the grey wolf algorithm is used to predict the molten steel segregation, and relevant data results are obtained through training. This model shows a very high prediction accuracy rate of 95.375%, which means that in a large number of prediction samples, the proportion of the model that can accurately predict the molten steel segregation situation is relatively high. In addition, the mean absolute error (MAE) value of the model is 89.372%, and the coefficient of determination value is 92.714%. MAE reflects the average absolute value of the error between the model prediction value and the actual value. A lower MAE value indicates that the model prediction result is closer to the actual value; the 2 R value measures the goodness of fit of the model to the data, and an R value close to 1 indicates that the model can fit the data well. 2

[0112] At the same time, this algorithm is compared with other algorithms such as linear regression, regression decision tree, and support vector machine. In terms of the proportion of data entries within the range of the fitting sample set error ±5%, for linear regression it is 63.627%, for regression decision tree it is 74.543%, and for support vector machine it is 51.612%. The random forest algorithm model improved based on the grey wolf algorithm has an advantage in prediction ability and can more accurately predict the molten steel segregation, providing support for optimizing the electromagnetic stirring process parameters and controlling the degree of molten steel segregation during the continuous casting production process.

[0113] Please refer to Figure 4 for the structural schematic diagram of a prediction device for molten steel segregation provided by an embodiment of this application, including:

[0114] A continuous casting parameter acquisition unit 21, configured to acquire the process parameter data of electromagnetic stirring and the corresponding molten steel segregation data during continuous casting. Among them, the process parameter data includes at least one of current intensity, frequency, coil position, coil shape, stirring time, and magnetic field direction, and the molten steel segregation data includes the average segregation index of molten steel;

[0115] A data division and processing unit 22, configured to preprocess the process parameter data and the molten steel segregation data to obtain a training data set for model training and a test data set for model verification;

[0116] The model parameter tuning and optimization unit 23 is used to optimize the hyperparameters of the random forest model by using the grey wolf algorithm to obtain an optimized random forest model;

[0117] The model training and testing unit 24 is used to train the optimized random forest model by using the training data set and verify the prediction performance of the optimized random forest model through the test data set;

[0118] The model application and prediction unit 25, based on the optimized random forest model that has been trained and verified, outputs the prediction result of the degree of molten steel segregation by inputting real-time process parameter data;

[0119] The parameter optimization and adjustment unit 26 is used to optimize and adjust the process parameter data of electromagnetic stirring according to the prediction result.

[0120] Please refer to Figure 5 , this application embodiment also provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored on the memory 310 and operable on the processor. When the processor 320 executes the computer program 311, it implements the steps of any method of the molten steel segregation prediction device.

[0121] Since the electronic device introduced in this embodiment is the device used to implement a molten steel segregation prediction device in this application embodiment, based on the method introduced in this application embodiment, those skilled in the art can understand the specific implementation manners and various variations of the electronic device in this embodiment. Therefore, the specific implementation of how this electronic device implements the method in this application embodiment will not be described in detail here. As long as the device used by those skilled in the art to implement the method in this application embodiment belongs to the scope protected by this application.

[0122] In the specific implementation process, when the computer program 311 is executed by the processor, it can implement any implementation manner in the corresponding embodiment of the first aspect.

[0123] It should be noted that in the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0124] Those skilled in the art should understand that the embodiments of this application can provide methods, systems, or computer program products. Therefore, this application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can adopt the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-readable program code.

[0125] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of flows and / or blocks in the flowchart and / or block diagram can also be implemented. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0126] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0127] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0128] Embodiments of the present application also provide a computer program product, which includes computer software instructions. When the computer software instructions run on a processing device, the processing device is caused to execute Figure 1 the process of a method for predicting molten steel segregation in a corresponding embodiment.

[0129] A computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, they implement, wholly or partly, the processes or functions in accordance with the embodiments of this application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired means (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless means (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be stored by a computer or a data storage device such as a server or a data center that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0130] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0131] In several embodiments provided in this application, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be indirect couplings or communication connections through some interfaces, devices, or units, and can be in electrical, mechanical, or other forms.

[0132] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0133] In addition, in each embodiment of the present application, each functional unit can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0134] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in each embodiment of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0135] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

[0136] Although the preferred embodiments of this specification have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments and all changes and modifications that fall within the scope of this specification.

[0137] Obviously, those skilled in the art can make various changes and deformations to this specification without departing from the spirit and scope of this specification. Thus, if these modifications and deformations of this specification fall within the scope of the claims of this specification and their equivalent technologies, this specification is also intended to include these changes and deformations.

Claims

1. A method for predicting molten steel segregation, characterized in that: The method comprises: Acquire process parameter data of electromagnetic stirring and corresponding molten steel segregation data during continuous casting, wherein the process parameter data includes at least one of current intensity, frequency, coil position, coil shape, stirring time and magnetic field direction, and the molten steel segregation data includes an average segregation index of the molten steel; Preprocessing the process parameter data and the molten steel segregation data to obtain a training data set for model training and a test data set for model verification; The gray wolf algorithm is used to optimize the hyperparameters of the random forest model to obtain the optimized random forest model; The optimized random forest model is trained using the training data set, and the prediction performance of the optimized random forest model is verified using the test data set; Based on the optimized random forest model that has been trained and verified, by inputting real-time process parameter data, outputting the prediction result of the segregation degree of molten steel; The process parameter data of electromagnetic stirring is optimized and adjusted according to the prediction results.

2. The method according to claim 1, characterized in that The preprocessing of the process parameter data and the molten steel segregation data to obtain a training data set for model training and a test data set for model verification includes: Performing abnormal value detection on the process parameter data and the molten steel segregation data, and correcting them through regression analysis if the errors are within a preset range to obtain corrected data; Supplementing missing values ​​of the revised data by interpolation to obtain supplementary data; The supplementary data are standardized and divided into a training data set and a test data set.

3. The method according to claim 1, characterized in that The gray wolf algorithm is used to optimize the hyperparameters of the random forest model to obtain an optimized random forest model, including: Initialize the gray wolf population and define an optimization objective function, wherein the optimization objective function is to minimize the prediction error, and the prediction error includes a mean square error and a mean absolute error; The position of the gray wolf is updated through iterative operations, and the optimal hyperparameter combination is gradually approached during the iteration process, wherein the hyperparameters include the number of decision trees, the maximum depth of the decision tree, and the minimum number of samples required for the leaf node; The obtained optimal hyperparameter combination is output, and the optimal hyperparameter combination is used in the construction process of the random forest model to obtain an optimized random forest model.

4. The method according to claim 1, characterized in that: The step of training the optimized random forest model using the training data set and verifying the prediction performance of the optimized random forest model using the test data set includes: The influence weight of each process parameter on the segregation of molten steel is calculated based on the entropy value, and the parameter with the highest influence weight is selected as the root node of the decision tree; Recursively divide the intermediate nodes until the stopping condition is met, wherein the stopping condition includes that all samples of the current node belong to the same category, the attribute set is empty, or the sample set is empty; Training the optimized random forest model based on the training data set to obtain the trained optimized random forest model; The trained optimized random forest model is verified based on the test data set to obtain a test error and a test determination coefficient, wherein, when the test determination coefficient is less than a first preset threshold or the test error is greater than a second preset threshold, the hyperparameters are re-optimized and the training is iterated until the test determination coefficient is greater than or equal to the first preset threshold and the test error is less than or equal to the second preset threshold.

5. The method according to claim 1, characterized in that The average segregation index of the molten steel is determined based on the following formula, expressed as: Among them, ASI is the average segregation index, C i is the actual concentration value of the i-th measurement point, C avg is the average concentration value of all measurement points, and n is the number of measurement points.

6. The method according to claim 1, characterized in that The method further comprises: Encapsulating the optimized random forest model into prediction software, wherein the prediction software is implemented in Python language and includes data dynamic update, model training and visualization interface functions; The real-time process parameter data is received through the visual interface, and the prediction results of molten steel segregation and parameter adjustment suggestions are output to realize the monitoring and optimization of the continuous casting process.

7. The method according to claim 1, characterized in that The optimizing and adjusting the process parameter data of electromagnetic stirring according to the prediction result comprises: Based on the prediction results of the optimized random forest model, an optimization adjustment strategy for electromagnetic stirring parameters is generated, wherein the optimization adjustment strategy includes adjusting the current intensity and frequency to enhance the electromagnetic stirring effect when the predicted average segregation index of the molten steel exceeds a preset threshold; the optimization adjustment strategy also includes optimizing the coil position and stirring time to improve the uniformity of the molten steel flow when the predicted average segregation index of the molten steel is lower than a preset threshold.

8. A device for predicting molten steel segregation, characterized in that: include: A continuous casting parameter acquisition unit, used to acquire process parameter data of electromagnetic stirring and corresponding molten steel segregation data during the continuous casting process, wherein the process parameter data includes at least one of current intensity, frequency, coil position, coil shape, stirring time and magnetic field direction, and the molten steel segregation data includes an average molten steel segregation index; A data partitioning processing unit, used for preprocessing the process parameter data and the molten steel segregation data to obtain a training data set for model training and a test data set for model verification; The model parameter optimization unit is used to optimize the hyperparameters of the random forest model using the gray wolf algorithm to obtain an optimized random forest model; A model training and testing unit, used to train the optimized random forest model using the training data set, and verify the prediction performance of the optimized random forest model using the test data set; A model application prediction unit, based on the optimized random forest model that has been trained and verified, outputs a prediction result of the segregation degree of molten steel by inputting real-time process parameter data; The parameter optimization and adjustment unit is used to optimize and adjust the process parameter data of electromagnetic stirring according to the prediction result.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to implement the steps of the method for predicting molten steel segregation as described in any one of claims 1 to 7 when executing the computer program stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting molten steel segregation according to any one of claims 1 to 7 is implemented.

Images