Liquid level periodic fluctuation prediction method and device, storage medium and computer equipment

Abstract

The invention discloses a liquid level periodic fluctuation prediction method and device, a storage medium and computer equipment, and the method comprises the steps: generating a training data set based on the roller diameter, the roller spacing, the pulling speed and the element content of key chemical components, and training a plurality of target evaluation prediction models based on the training data set, the target evaluation prediction model comprises a random forest model, a support vector machine model and a genetic algorithm optimization back propagation neural network model; comprehensive performance evaluation values of all the target evaluation and prediction models under multiple performance evaluation indexes are calculated, the target evaluation and prediction model corresponding to the maximum comprehensive performance evaluation value is determined as the liquid level periodic fluctuation prediction model, and the performance evaluation indexes comprise the mean absolute error index, the mean square error index and the F1 score index. The prediction model is trained in combination with a metallurgy mechanism and big data, the trained model can predict the periodic abnormal fluctuation of the liquid level of the molten steel, and the continuous casting efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of steel smelting, and in particular to a method and device for predicting periodic fluctuations of liquid level, a storage medium, and a computer device. Background Art

[0002] As the core link of modern steel production, continuous casting has an irreplaceable position and role. However, due to the complexity of the continuous casting process, there are many surface defects in the continuous casting billet. Many studies have shown that the periodic fluctuation of the crystallizer liquid level determines the surface quality of the continuous casting billet. The increase in the fluctuation range will lead to a significant increase in the probability of subcutaneous slag inclusions and surface cracks in the continuous casting billet. Summary of the invention

[0003] In view of this, the present application provides a method and device for predicting periodic fluctuations of liquid level, a storage medium, and a computer device, which are applied to a continuous casting machine, wherein the continuous casting machine includes a crystallizer and a roller. When the continuous casting machine is used to cast molten steel, the molten steel flows out of the crystallizer and is transmitted through the roller, and a plurality of rollers are installed on the roller. A training data set is generated based on the roller diameter, roller spacing, drawing speed, and element content of key chemical components, and multiple target evaluation prediction models are trained based on the training data set, wherein the target evaluation prediction model includes a random forest model, a support vector machine model, and a genetic algorithm optimized back propagation neural network model. The comprehensive performance evaluation value of each target evaluation prediction model under a variety of performance evaluation indicators is calculated respectively, and the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value is determined as the liquid level periodic fluctuation prediction model, wherein the performance evaluation indicators include the mean absolute error indicator, the mean square error indicator, and the F1 score indicator. Combining the metallurgical mechanism with the big data training prediction model, the trained model can predict the periodic abnormal fluctuations of the molten steel level, which can improve the continuous casting efficiency.

[0004] According to one aspect of the present application, a method for predicting periodic fluctuations of a liquid level is provided, the method comprising:

[0005] Based on the micro-segregation model, determining the key chemical components that affect the abnormal periodic fluctuation of the target molten steel level, wherein the abnormal fluctuation of the target molten steel level changes periodically;

[0006] Acquiring continuous casting parameter data of each of the plurality of continuous casting machines, wherein the continuous casting parameter data includes roller diameter, roller spacing and casting speed;

[0007] Generate a training data set based on the roller diameter, the roller spacing, the drawing speed and the element content of the key chemical components, and train multiple target evaluation prediction models based on the training data set, wherein the target evaluation prediction models include a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model;

[0008] The comprehensive performance evaluation values ​​of each target evaluation prediction model under multiple performance evaluation indicators are calculated respectively, and the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value is determined as the liquid level periodic fluctuation prediction model, wherein the performance evaluation indicators include the mean absolute error indicator, the mean square error indicator and the F1 score indicator.

[0009] Optionally, the determining of the key chemical components that affect the periodic abnormal fluctuation of the target molten steel level based on the micro-segregation model includes:

[0010] A microsegregation model is established based on the solute redistribution of various alloying elements in the target steel liquid during solidification, wherein various alloying elements do not contain carbon and are composed of different chemical compositions;

[0011] The solidification process of target molten steel with different chemical compositions is simulated based on the microsegregation model, and the uneven distribution of each group of chemical compositions at the grain scale is obtained in various solidification processes. Based on the various uneven distributions obtained, the key chemical composition causing the periodic abnormal fluctuation of the target molten steel level is determined.

[0012] Optionally, the comprehensive performance evaluation values ​​of each target evaluation prediction model under multiple performance evaluation indicators are calculated separately, including:

[0013] For any target evaluation prediction model, calculate the mean absolute error index value, mean square error index value and F1 score index value of the target evaluation prediction model;

[0014] Based on the mean absolute error index value, the mean square error index value, the F1 score index value and the preset evaluation weights of various performance evaluation indicators, a comprehensive performance evaluation value of the target evaluation prediction model is obtained.

[0015] Optionally, calculating the mean absolute error index value of the target evaluation prediction model includes:

[0016] According to the mean absolute error index value calculation formula, the mean absolute error index value of the target evaluation prediction model is calculated, wherein the mean absolute error index value calculation formula is:

[0017]

[0018] MAE is the mean absolute error indicator value, n is the total number of data input into the target evaluation prediction model, and y p Evaluate the predicted value of the prediction model for the target, y i is the actual value of the i-th input data.

[0019] Optionally, calculating the mean square error index value of the target evaluation prediction model includes:

[0020] According to the mean square error index value calculation formula, the mean square error index value of the target evaluation prediction model is calculated, wherein the mean square error index value calculation formula is:

[0021]

[0022] MSE is the mean square error index value, n is the total number of data input into the target evaluation prediction model, y p Evaluate the predicted value of the prediction model for the target, y i is the actual value of the i-th input data.

[0023] Optionally, the calculating the F1 score index value of the target evaluation prediction model includes:

[0024] According to the F1 score index value calculation formula, the F1 score index value of the target evaluation prediction model is calculated, wherein the F1 score index value calculation formula is:

[0025]

[0026] F1 is the F1 score indicator value, q is the prediction accuracy of the target evaluation prediction model, and r is the recall rate of the target evaluation prediction model.

[0027] Optionally, the method further comprises:

[0028] Based on the statistical results of the slag inclusion defect blocking rate after the molten steel in the crystallizer is poured into ingots under different liquid level fluctuation conditions, the periodic liquid level abnormal fluctuation range of the molten steel is determined. When it is predicted based on the liquid level periodic fluctuation prediction model that the liquid level difference of the target predicted molten steel in the crystallizer exceeds the periodic liquid level abnormal fluctuation range, the liquid level of the target predicted molten steel undergoes periodic abnormal fluctuations.

[0029] According to another aspect of the present application, a device for predicting periodic fluctuations of a liquid level is provided, the device comprising:

[0030] A key chemical composition determination module is used to determine the key chemical composition that affects the periodic abnormal fluctuation of the target molten steel level based on a micro-segregation model, wherein the fluctuation of the target molten steel level is abnormal and changes periodically;

[0031] A continuous casting machine parameter acquisition module, used to acquire continuous casting parameter data of various continuous casting machines, wherein the continuous casting parameter data include roller diameter, roller spacing and casting speed;

[0032] A multi-model training module, for generating a training data set based on the roller diameter, the roller spacing, the drawing speed and the element content of the key chemical components, and training a plurality of target evaluation prediction models based on the training data set, wherein the target evaluation prediction models include a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model;

[0033] The optimal prediction model determination module is used to calculate the comprehensive performance evaluation value of each target evaluation prediction model under multiple performance evaluation indicators, and determine the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value as the liquid level periodic fluctuation prediction model, wherein the performance evaluation indicators include the mean absolute error indicator, the mean square error indicator and the F1 score indicator.

[0034] Optionally, the key chemical component determination module is further used to:

[0035] A microsegregation model is established based on the solute redistribution of various alloying elements in the target steel liquid during solidification, wherein various alloying elements do not contain carbon and are composed of different chemical compositions;

[0036] The solidification process of target molten steel with different chemical compositions is simulated based on the microsegregation model, and the uneven distribution of each group of chemical compositions at the grain scale is obtained in various solidification processes. Based on the various uneven distributions obtained, the key chemical composition causing the periodic abnormal fluctuation of the target molten steel level is determined.

[0037] Optionally, the device further comprises: a performance evaluation module, configured to:

[0038] For any target evaluation prediction model, calculate the mean absolute error index value, mean square error index value and F1 score index value of the target evaluation prediction model;

[0039] Based on the mean absolute error index value, the mean square error index value, the F1 score index value and the preset evaluation weights of various performance evaluation indicators, a comprehensive performance evaluation value of the target evaluation prediction model is obtained.

[0040] Optionally, the performance evaluation module is further used to:

[0041] According to the mean absolute error index value calculation formula, the mean absolute error index value of the target evaluation prediction model is calculated, wherein the mean absolute error index value calculation formula is:

[0042]

[0043] MAE is the mean absolute error indicator value, n is the total number of data input into the target evaluation prediction model, and y p Evaluate the predicted value of the prediction model for the target, y i is the actual value of the i-th input data.

[0044] Optionally, the performance evaluation module is further used to:

[0045] According to the mean square error index value calculation formula, the mean square error index value of the target evaluation prediction model is calculated, wherein the mean square error index value calculation formula is:

[0046]

[0047] MSE is the mean square error index value, n is the total number of data input into the target evaluation prediction model, y p Evaluate the predicted value of the prediction model for the target, y i is the actual value of the i-th input data.

[0048] Optionally, the performance evaluation module is further used to:

[0049] According to the F1 score index value calculation formula, the F1 score index value of the target evaluation prediction model is calculated, wherein the F1 score index value calculation formula is:

[0050]

[0051] F1 is the F1 score indicator value, q is the prediction accuracy of the target evaluation prediction model, and r is the recall rate of the target evaluation prediction model.

[0052] Optionally, the device further includes: an abnormal fluctuation range determination module, configured to:

[0053] Based on the statistical results of the slag inclusion defect blocking rate after the molten steel in the crystallizer is poured into ingots under different liquid level fluctuation conditions, the periodic liquid level abnormal fluctuation range of the molten steel is determined. When it is predicted based on the liquid level periodic fluctuation prediction model that the liquid level difference of the target predicted molten steel in the crystallizer exceeds the periodic liquid level abnormal fluctuation range, the liquid level of the target predicted molten steel undergoes periodic abnormal fluctuations.

[0054] According to another aspect of the present application, a storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned method for predicting periodic fluctuations of the liquid level is implemented.

[0055] According to another aspect of the present application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the above-mentioned method for predicting periodic fluctuations of liquid level when executing the program.

[0056] By means of the above technical scheme, the present application provides a method and device for predicting periodic fluctuations of liquid level, a storage medium, and a computer device. Based on the roller diameter, roller spacing, drawing speed, and element content of key chemical components, a training data set is generated, and multiple target evaluation prediction models are trained based on the training data set, wherein the target evaluation prediction models include a random forest model, a support vector machine model, and a genetic algorithm optimized back propagation neural network model; the comprehensive performance evaluation values ​​of each target evaluation prediction model under multiple performance evaluation indicators are calculated respectively, and the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value is determined as the liquid level periodic fluctuation prediction model, wherein the performance evaluation indicators include the mean absolute error indicator, the mean square error indicator, and the F1 score indicator. Combining the metallurgical mechanism with the big data training prediction model, the trained model can predict the periodic abnormal fluctuations of the molten steel level and improve the continuous casting efficiency.

[0057] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0059] Figure 1 A schematic diagram of a process of predicting a periodic fluctuation of a liquid level provided in an embodiment of the present application is shown;

[0060] Figure 2 A schematic diagram showing a flow chart of another method for predicting periodic fluctuations of liquid level provided in an embodiment of the present application is shown;

[0061] Figure 3 A schematic diagram showing the relationship between manganese content and conversion rate provided in an embodiment of the present application is shown;

[0062] Figure 4 A schematic diagram of a random forest model architecture provided in an embodiment of the present application is shown;

[0063] Figure 5 A schematic diagram of the performance evaluation index values ​​of each target evaluation model provided in the embodiment of the present application is shown;

[0064] Figure 6 A schematic diagram showing the relationship between a liquid level fluctuation and a surface longitudinal crack rate, a slag blocking rate and a downline manual cleaning rate provided in an embodiment of the present application is shown;

[0065] Figure 7 A schematic diagram of the structure of a liquid level periodic fluctuation prediction device provided in an embodiment of the present application is shown;

[0066] Figure 8 A schematic structural diagram of another liquid level periodic fluctuation prediction device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0067] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0068] In this embodiment, a method for predicting periodic fluctuations of liquid level is provided. Figure 1 As shown, it is applied to a continuous casting machine, the continuous casting machine includes a crystallizer and a roller. When the continuous casting machine is used to cast molten steel, the molten steel flows out of the crystallizer and is transported through the roller, and a plurality of rollers are installed on the roller. The method includes:

[0069] Step 101, based on a micro-segregation model, determine the key chemical components that affect the abnormal periodic fluctuation of the target molten steel level, wherein the abnormal fluctuation of the target molten steel level changes periodically.

[0070] Although there are many studies on periodic abnormal liquid level fluctuations, there are few studies on the prediction of periodic abnormal fluctuations based on continuous casting big data analysis and artificial intelligence models. Using deep learning models to combine metallurgical mechanisms with big data to predict periodic abnormal fluctuations has become an inevitable trend in the study of crystallizer liquid level fluctuations.

[0071] In the above embodiments of the present application, firstly, based on the microsegregation model, the key chemical components that affect the periodic abnormal fluctuation of the target molten steel liquid level are determined. Specifically, based on the microsegregation model, the solute microsegregation is calculated at different cooling rates, and the activity is determined according to the interaction coefficients of different components, and then the influence of different chemical components on the peritectic phase transformation is judged, so as to select the main chemical components that affect the liquid level fluctuation according to the above results.

[0072] Step 102, obtaining continuous casting parameter data of each of the plurality of continuous casting machines, wherein the continuous casting parameter data includes roller diameter, roller spacing and casting speed.

[0073] Next, the continuous casting parameter data of various continuous casting machines are obtained. The continuous casting parameter data include roller diameter, roller spacing and drawing speed. Since the roller arrangement of each continuous casting machine is fixed, when obtaining parameter data, it can be collected from different steel plants to ensure the richness of data and the applicability of the model. In particular, the aforementioned continuous casting parameter data can be directly read from the PLC controller in the continuous casting production line. For example, in Steel Plant A, the continuous casting parameter data of each sample continuous casting machine are shown in Table 1, and the continuous casting parameter data of Steel Plant B are shown in Table 2. In addition, the data of different drawing speeds of every 5 furnaces of low carbon steel, peritectic steel and subperitectic steel can be collected respectively. The chemical composition of the ingots cast by sample continuous casting machines 1 to 6 based on the molten steel is shown in Table 3.

[0074] Table 1

[0075]

[0076] Table 2

[0077]

[0078] Table 3

[0079]

[0080] Step 103, based on the roller diameter, the roller spacing, the drawing speed and the element content of the key chemical components, generate a training data set, and train multiple target evaluation prediction models based on the training data set, wherein the target evaluation prediction model includes a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model.

[0081] Next, a training data set is generated based on the roller diameter, roller spacing, pulling speed and element content of key chemical components. Multiple target evaluation prediction models are trained based on the training data set. The target evaluation prediction models include a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model, so as to determine the model with the best prediction performance.

[0082] Step 104, respectively calculate the comprehensive performance evaluation value of each target evaluation prediction model under multiple performance evaluation indicators, and determine the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value as the liquid level periodic fluctuation prediction model, wherein the performance evaluation indicators include the mean absolute error indicator, the mean square error indicator and the F1 score indicator.

[0083] Finally, the comprehensive performance evaluation values ​​of each target evaluation prediction model under multiple performance evaluation indicators are calculated respectively, and the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value is determined as the liquid level periodic fluctuation prediction model. The performance evaluation indicators include the mean absolute error index, the mean square error index and the F1 score index. In order to evaluate the prediction model according to multiple levels. To this end, metallurgical data of the continuous casting production site are collected in real time, including liquid level fluctuation, drawing speed, billet width, steel chemical composition, roller spacing, roller diameter and other data. The influence of chemical composition on the periodic abnormal fluctuation of the liquid level is analyzed by thermodynamic principles, and the key chemical composition is found. It is used as input data with the roller diameter, roller spacing, and drawing speed to train the deep learning model. The trained deep learning model can be used to predict whether the liquid level fluctuation is normal. If the prediction result is abnormal fluctuation, the corresponding process is adjusted to ensure stable fluctuation of the liquid level, which can improve the continuous casting efficiency.

[0084] By applying the technical solution of this embodiment, a training data set is generated based on roller diameter, roller spacing, drawing speed and element content of key chemical components, and multiple target evaluation prediction models are trained based on the training data set, wherein the target evaluation prediction model includes a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model; the comprehensive performance evaluation value of each target evaluation prediction model under multiple performance evaluation indicators is calculated respectively, and the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value is determined as the liquid level periodic fluctuation prediction model, wherein the performance evaluation indicators include the mean absolute error indicator, the mean square error indicator and the F1 score indicator. Combining the metallurgical mechanism with the big data training prediction model, the trained model can predict the periodic abnormal fluctuation of the molten steel level, which can improve the continuous casting efficiency.

[0085] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, another method for predicting periodic fluctuations of liquid level is provided, such as Figure 2 As shown, it is applied to a continuous casting machine, the continuous casting machine includes a crystallizer and a roller. When the continuous casting machine is used to cast molten steel, the molten steel flows out of the crystallizer and is transported through the roller, and a plurality of rollers are installed on the roller. The method includes:

[0086] Step 201, based on the solute redistribution of various alloy elements in the target molten steel during solidification, a micro segregation model is established, wherein various alloy elements do not contain carbon elements and various alloy elements are composed of different chemical components.

[0087] Step 202, based on the micro-segregation model, simulate the solidification process of the target molten steel with different chemical components, and obtain the uneven distribution of each group of chemical components at the grain scale in various solidification processes, and determine the key chemical components that cause the target molten steel to have periodic abnormal fluctuations in the liquid level based on the various obtained uneven distributions, wherein the fluctuation of the target molten steel level is periodic.

[0088] In the above embodiments of the present application, the microscopic segregation model can be used to analyze the influence of chemical composition on abnormal fluctuation of liquid level, the δ transformation rate under different Mn (manganese) contents, for example Figure 3 As shown in Figure 2, with the increase of Mn (manganese) content, the δ transformation rate also increases accordingly. Figure 3 It can be seen that the increase of Mn (manganese) content will aggravate the degree of shrinkage of peritectic phase transformation, and the greater the degree of shrinkage of peritectic phase transformation, the more serious the solidification inhomogeneity. In addition, due to the narrow range of δ to γ ​​(γ represents the microstructural state or phase state of the material, and the transition from δ to γ ​​may mean that a certain structure or phase change has occurred inside the material), and L (L represents another state or condition of the material, such as liquid, a certain microstructure or phase state, etc., and the transition from L to δ may indicate that the material has changed from one state to another) to δ, a larger δ transformation rate is more likely to lead to abnormal PMLF (Periodic Mold Level Fluctuation, periodic fluctuation of mold shell liquid level). To this end, in addition to the C (carbon) content, the Mn content is a factor that affects the abnormal fluctuation of the abnormal liquid level. That is, a micro-segregation model is established based on the solute redistribution of various alloy elements in the target molten steel during solidification. The solidification process of molten steel with different chemical compositions is simulated based on the micro-segregation model, and the uneven distribution of each group of chemical components at the grain scale is obtained in various solidification processes. Based on the various uneven distributions obtained, the key chemical components that cause periodic abnormal fluctuations in the target molten steel level are determined. In the above embodiment of the present application, the last key chemical component may be manganese.

[0089] Step 203, obtaining continuous casting parameter data of each of the plurality of continuous casting machines, wherein the continuous casting parameter data includes roller diameter, roller spacing and casting speed.

[0090] Step 204, based on the roller diameter, the roller spacing, the drawing speed and the element content of the key chemical components, generate a training data set, and train multiple target evaluation prediction models based on the training data set, wherein the target evaluation prediction model includes a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model.

[0091] Next, the continuous casting parameter data of various continuous casting machines are obtained, and a training data set is generated based on the roller diameter, roller spacing, casting speed and element content of key chemical components. Based on the training data set, multiple target evaluation prediction models are trained respectively. The target evaluation prediction models include a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model. For example, when training the random forest model, the casting speed, roller diameter, roller spacing and Mn content are used as inputs, and the random forest model is used to predict the quality of liquid level fluctuations. In particular, when the liquid level fluctuation range exceeds ±5mm, it can be considered as an abnormal fluctuation. The random forest model architecture is as follows: Figure 4 shown.

[0092] Step 205, for any target evaluation prediction model, calculate the mean absolute error index value of the target evaluation prediction model according to the mean absolute error index value calculation formula, wherein the mean absolute error index value calculation formula is:

[0093]

[0094] MAE is the mean absolute error indicator value, n is the total number of data input into the target evaluation prediction model, and y p Evaluate the predicted value of the prediction model for the target, y i is the actual value of the i-th input data.

[0095] Step 206, according to the mean square error index value calculation formula, calculate the mean square error index value of the target evaluation prediction model, wherein the mean square error index value calculation formula is:

[0096]

[0097] MSE is the mean square error index value, n is the total number of data input into the target evaluation prediction model, y p Evaluate the predicted value of the prediction model for the target, y i is the actual value of the i-th input data.

[0098] Step 207, according to the F1 score index value calculation formula, calculate the F1 score index value of the target evaluation prediction model, wherein the F1 score index value calculation formula is:

[0099]

[0100] F1 is the F1 score indicator value, q is the prediction accuracy of the target evaluation prediction model, and r is the recall rate of the target evaluation prediction model.

[0101] Step 208, based on the mean absolute error index value, the mean square error index value, the F1 score index value and the preset evaluation weights of various performance evaluation indicators, obtain the comprehensive performance evaluation value of the target evaluation prediction model.

[0102] Step 209: determine the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value as the liquid level periodic fluctuation prediction model.

[0103] Next, the mean absolute error (MAE), mean square error (MSE) and F1 score (F1score) are used to evaluate the performance of the model. In particular, the prediction accuracy of each model can be calculated for joint judgment. The performance of the RF (random forest) model, the support vector machine model (SVM) and the GA-BP (genetic algorithm optimized back propagation neural network) model is shown in the figure. Figure 5 As shown. Figure 4 It can be seen that the prediction accuracy of the GA-BP model is 76.41%, and the prediction accuracy of the RF model is 92.76%, which is 21.39% higher than the GA-BP model. The MAE of the RF model is 0.09, which is 67.86% smaller than the GA-BP model, the MSE of the RF model is 0.02, which is 77.78% smaller than the GA-BP model, and the F1score of the RF model is 0.92, which is 21.05% higher than the GA-BP model. For this reason, in the end, the RF model has the best ability to predict the quality of liquid level fluctuations, so the random forest model can be selected as the liquid level periodic fluctuation prediction model.

[0104] Step 210, based on the statistical results of the slag inclusion defect blocking rate obtained after the molten steel in the crystallizer is cast into ingots under different liquid level fluctuation conditions, determine the periodic liquid level abnormal fluctuation range of the molten steel; when it is predicted based on the liquid level periodic fluctuation prediction model that the liquid level difference of the target predicted molten steel in the crystallizer exceeds the periodic liquid level abnormal fluctuation range, the liquid level of the target predicted molten steel undergoes periodic abnormal fluctuations.

[0105] Next, the abnormal fluctuation range can be determined, such as Figure 6 As shown, Figure 6 The statistical results of the blocking rate of slag inclusion defects in castings under different liquid level fluctuation conditions are given in Figure 6In the crystallizer, when the steel liquid level fluctuation in the crystallizer is ≤±3mm, the slag inclusion blocking rate of the ingot is 0.02%; when ±3mm <crystallizer liquid level fluctuation range ≤±5mm, the slag inclusion blocking rate of the ingot is 0.03%, and when the crystallizer liquid level fluctuation is >±5mm, the slag inclusion blocking rate of the ingot is as high as 0.44%. It can be seen from this that the slag inclusion blocking rate on the surface of the ingot when the steel liquid level fluctuation in the crystallizer exceeds ±5mm is 14 times the slag inclusion blocking rate on the surface of the ingot when the crystallizer liquid level fluctuation does not exceed ±5mm. Therefore, when the steel liquid level fluctuation in the crystallizer exceeds ±5mm, the protective slag slag rolling phenomenon will occur. The tiny protective slag droplets rolled into the crystallizer are captured by the primary billet shell to form slag inclusion defects on the surface of the ingot, that is, when the mold liquid level difference exceeds ±5mm during the fluctuation, it is considered that an abnormal liquid level fluctuation has occurred. To this end, based on the on-site liquid level data, a large number of surface defect generation rates of ingots are analyzed, and the relationship between the two is established and compared to determine the range of abnormal fluctuations.

[0106] By applying the technical solution of the present embodiment, a micro-segregation model is established based on the solute redistribution of various alloy elements in the target molten steel during solidification. The solidification process of molten steel with different chemical compositions is simulated based on the micro-segregation model. In various solidification processes, the uneven distribution of each group of chemical components at the grain scale is obtained. Based on the various uneven distributions obtained, the key chemical components that cause the periodic abnormal fluctuation of the target molten steel level are determined. The continuous casting parameter data of various continuous casting machines are obtained. Based on the roller diameter, roller spacing, pulling speed and element content of the key chemical components, a training set is generated. Data set, based on the training data set, multiple target evaluation prediction models are trained respectively. For any target evaluation prediction model, the average absolute error index value of the target evaluation prediction model is calculated according to the average absolute error index value calculation formula, the mean square error index value of the target evaluation prediction model is calculated according to the mean square error index value calculation formula, and the F1 score index value of the target evaluation prediction model is calculated according to the F1 score index value calculation formula. Based on the average absolute error index value, the mean square error index value, the F1 score index value and the preset evaluation weights of various performance evaluation indicators, the comprehensive performance evaluation value of the target evaluation prediction model is obtained. The target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value is determined as the liquid level periodic fluctuation prediction model. Based on the statistical results of the slag inclusion defect blocking rate obtained after the molten steel in the crystallizer is cast into a billet under different liquid level fluctuation conditions, the periodic liquid level abnormal fluctuation range of the molten steel is determined. When the liquid level difference of the target predicted molten steel in the crystallizer exceeds the periodic liquid level abnormal fluctuation range based on the liquid level periodic fluctuation prediction model, the liquid level of the target predicted molten steel undergoes periodic abnormal fluctuations. By combining metallurgical mechanisms with big data training prediction models, the trained model can predict the periodic abnormal fluctuations of the molten steel level and improve the continuous casting efficiency.

[0107] Further, as Figure 1The specific implementation of the method, the embodiment of the present application provides a liquid level periodic fluctuation prediction device, such as Figure 7 As shown, the device comprises:

[0108] The key chemical component determination module 301 is used to determine the key chemical components that affect the periodic abnormal fluctuation of the target molten steel level based on the micro-segregation model, wherein the fluctuation of the target molten steel level is abnormal and changes periodically;

[0109] A continuous casting machine parameter acquisition module 302 is used to acquire continuous casting parameter data of various continuous casting machines, wherein the continuous casting parameter data includes roller diameter, roller spacing and casting speed;

[0110] A multi-model training module 303 is used to generate a training data set based on the roller diameter, the roller spacing, the casting speed and the element content of the key chemical components, and train multiple target evaluation prediction models based on the training data set, wherein the target evaluation prediction models include a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model;

[0111] The optimal prediction model determination module 304 is used to calculate the comprehensive performance evaluation values ​​of each target evaluation prediction model under multiple performance evaluation indicators, and determine the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value as the liquid level periodic fluctuation prediction model, wherein the performance evaluation indicators include the mean absolute error indicator, the mean square error indicator and the F1 score indicator.

[0112] Optionally, the key chemical component determination module 301 is further used to:

[0113] A microsegregation model is established based on the solute redistribution of various alloying elements in the target steel liquid during solidification, wherein various alloying elements do not contain carbon and are composed of different chemical compositions;

[0114] The solidification process of target molten steel with different chemical compositions is simulated based on the microsegregation model, and the uneven distribution of each group of chemical compositions at the grain scale is obtained in various solidification processes. Based on the various uneven distributions obtained, the key chemical composition causing the periodic abnormal fluctuation of the target molten steel level is determined.

[0115] Furthermore, the present application provides another liquid level periodic fluctuation prediction device, such as Figure 8 As shown, the device comprises:

[0116] The key chemical composition determination module 401 is used to determine the key chemical composition that affects the periodic abnormal fluctuation of the target molten steel level based on the micro-segregation model, wherein the fluctuation of the target molten steel level is abnormal and changes periodically;

[0117] The continuous casting machine parameter acquisition module 402 is used to acquire continuous casting parameter data of various continuous casting machines, wherein the continuous casting parameter data includes roller diameter, roller spacing and casting speed;

[0118] A multi-model training module 403 is used to generate a training data set based on the roller diameter, the roller spacing, the casting speed and the element content of the key chemical components, and train multiple target evaluation prediction models based on the training data set, wherein the target evaluation prediction models include a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model;

[0119] The optimal prediction model determination module 404 is used to respectively calculate the comprehensive performance evaluation values ​​of each target evaluation prediction model under multiple performance evaluation indicators, and determine the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value as the liquid level periodic fluctuation prediction model, wherein the performance evaluation indicators include the mean absolute error indicator, the mean square error indicator and the F1 score indicator;

[0120] The performance evaluation module 405 is used to calculate the mean absolute error index value, mean square error index value and F1 score index value of any target evaluation prediction model; based on the mean absolute error index value, the mean square error index value, the F1 score index value and the preset evaluation weights of various performance evaluation indicators, obtain the comprehensive performance evaluation value of the target evaluation prediction model;

[0121] The abnormal fluctuation range determination module 406 is used to determine the periodic liquid level abnormal fluctuation range of the molten steel based on the statistical results of the slag inclusion defect blocking rate obtained after the molten steel in the crystallizer is poured into ingots under different liquid level fluctuation conditions. When it is predicted based on the liquid level periodic fluctuation prediction model that the liquid level difference of the target predicted molten steel in the crystallizer exceeds the periodic liquid level abnormal fluctuation range, the liquid level of the target predicted molten steel undergoes periodic abnormal fluctuations.

[0122] Optionally, the key chemical component determination module 401 is further used to:

[0123] A microsegregation model is established based on the solute redistribution of various alloying elements in the target steel liquid during solidification, wherein various alloying elements do not contain carbon and are composed of different chemical compositions;

[0124] The solidification process of target molten steel with different chemical compositions is simulated based on the microsegregation model, and the uneven distribution of each group of chemical compositions at the grain scale is obtained in various solidification processes. Based on the various uneven distributions obtained, the key chemical composition causing the periodic abnormal fluctuation of the target molten steel level is determined.

[0125] Optionally, the performance evaluation module 405 is further used to:

[0126] According to the mean absolute error index value calculation formula, the mean absolute error index value of the target evaluation prediction model is calculated, wherein the mean absolute error index value calculation formula is:

[0127]

[0128] MAE is the mean absolute error indicator value, n is the total number of data input into the target evaluation prediction model, and y p Evaluate the predicted value of the prediction model for the target, y i is the actual value of the i-th input data.

[0129] Optionally, the performance evaluation module 405 is further used to:

[0130] According to the mean square error index value calculation formula, the mean square error index value of the target evaluation prediction model is calculated, wherein the mean square error index value calculation formula is:

[0131]

[0132] MSE is the mean square error index value, n is the total number of data input into the target evaluation prediction model, y p Evaluate the predicted value of the prediction model for the target, y i is the actual value of the i-th input data.

[0133] Optionally, the performance evaluation module 405 is further used to:

[0134] According to the F1 score index value calculation formula, the F1 score index value of the target evaluation prediction model is calculated, wherein the F1 score index value calculation formula is:

[0135]

[0136] F1 is the F1 score indicator value, q is the prediction accuracy of the target evaluation prediction model, and r is the recall rate of the target evaluation prediction model.

[0137] It should be noted that for other corresponding descriptions of the functional units involved in the liquid level periodic fluctuation prediction device provided in the embodiment of the present application, reference can be made to Figure 1 to Figure 2The corresponding description in the method will not be repeated here.

[0138] Based on the above Figure 1 to Figure 2 The method shown in the embodiment of the present application is accordingly provided with a storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned Figure 1 to Figure 2 The method for predicting periodic fluctuations of liquid level is shown.

[0139] Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0140] Based on the above Figure 1 to Figure 2 The method shown, and Figure 7 , Figure 8 In order to achieve the above-mentioned purpose, the embodiment of the present application further provides a computer device, which can be a personal computer, a server, a network device, etc. The computer device includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to achieve the above-mentioned Figure 1 to Figure 2 The method for predicting periodic fluctuations of liquid level is shown.

[0141] Optionally, the computer device may further include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, etc. The user interface may include a display, an input unit such as a keyboard, etc., and the optional user interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a Bluetooth interface, a WI-FI interface), etc.

[0142] Those skilled in the art will appreciate that the computer device structure provided in this embodiment does not limit the computer device, and may include more or fewer components, or a combination of certain components, or different component arrangements.

[0143] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages and saves the hardware and software resources of the computer device, and supports the operation of information processing programs and other software and / or programs. The network communication module is used to realize communication between the components inside the storage medium, and communication with other hardware and software in the physical device.

[0144] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware platforms, or by hardware based on roller diameter, roller spacing, drawing speed and element content of key chemical components, to generate a training data set, and train multiple target evaluation prediction models based on the training data set, wherein the target evaluation prediction model includes a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model; calculate the comprehensive performance evaluation value of each target evaluation prediction model under a variety of performance evaluation indicators, and determine the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value as a liquid level periodic fluctuation prediction model, wherein the performance evaluation indicators include the mean absolute error indicator, the mean square error indicator and the F1 score indicator. Combining the metallurgical mechanism with the big data training prediction model, the trained model can predict the periodic abnormal fluctuation of the molten steel level, which can improve the continuous casting efficiency.

[0145] Those skilled in the art will appreciate that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily necessary for implementing the present application. Those skilled in the art will appreciate that the modules in the devices in the implementation scenario can be distributed in the devices of the implementation scenario according to the description of the implementation scenario, or can be changed accordingly and located in one or more devices different from the present implementation scenario. The modules of the above-mentioned implementation scenario can be combined into one module, or can be further split into multiple submodules.

[0146] The above serial numbers of this application are only for description and do not represent the advantages and disadvantages of the implementation scenarios. The above disclosure is only a few specific implementation scenarios of this application, but this application is not limited to them, and any changes that can be thought of by technicians in this field should fall within the scope of protection of this application.

Claims

1. A method for predicting periodic fluctuations of liquid level, characterized in that: Applied to a continuous casting machine, the continuous casting machine includes a crystallizer and a roller table. When the continuous casting machine is used to cast molten steel, the molten steel flows out of the crystallizer and is transported through the roller table, on which a plurality of rollers are installed. The method includes: Based on the micro-segregation model, determining the key chemical components that affect the abnormal periodic fluctuation of the target molten steel level, wherein the abnormal fluctuation of the target molten steel level changes periodically; Acquiring continuous casting parameter data of each of the plurality of continuous casting machines, wherein the continuous casting parameter data includes roller diameter, roller spacing and casting speed; Generate a training data set based on the roller diameter, the roller spacing, the drawing speed and the element content of the key chemical components, and train multiple target evaluation prediction models based on the training data set, wherein the target evaluation prediction models include a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model; The comprehensive performance evaluation values ​​of each target evaluation prediction model under multiple performance evaluation indicators are calculated respectively, and the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value is determined as the liquid level periodic fluctuation prediction model, wherein the performance evaluation indicators include the mean absolute error indicator, the mean square error indicator and the F1 score indicator.

2. The method according to claim 1, characterized in that The key chemical components that affect the periodic abnormal fluctuation of the target molten steel level are determined based on the micro-segregation model, including: A microsegregation model is established based on the solute redistribution of various alloying elements in the target steel liquid during solidification, wherein various alloying elements do not contain carbon and are composed of different chemical compositions; The solidification process of target molten steel with different chemical compositions is simulated based on the microsegregation model, and the uneven distribution of each group of chemical compositions at the grain scale is obtained in various solidification processes. Based on the various uneven distributions obtained, the key chemical composition causing the periodic abnormal fluctuation of the target molten steel level is determined.

3. The method according to claim 1, characterized in that The comprehensive performance evaluation values ​​of each target evaluation prediction model under multiple performance evaluation indicators are calculated respectively, including: For any target evaluation prediction model, calculate the mean absolute error index value, mean square error index value and F1 score index value of the target evaluation prediction model; Based on the mean absolute error index value, the mean square error index value, the F1 score index value and the preset evaluation weights of various performance evaluation indicators, a comprehensive performance evaluation value of the target evaluation prediction model is obtained.

4. The method according to claim 3, characterized in that The calculating the mean absolute error index value of the target evaluation prediction model includes: According to the mean absolute error index value calculation formula, the mean absolute error index value of the target evaluation prediction model is calculated, wherein the mean absolute error index value calculation formula is: MAE is the mean absolute error indicator value, n is the total number of data input into the target evaluation prediction model, and y p Evaluate the predicted value of the prediction model for the target, y i is the actual value of the i-th input data.

5. The method according to claim 3, characterized in that: The calculating the mean square error index value of the target evaluation prediction model includes: According to the mean square error index value calculation formula, the mean square error index value of the target evaluation prediction model is calculated, wherein the mean square error index value calculation formula is: MSE is the mean square error index value, n is the total number of data input into the target evaluation prediction model, y p Evaluate the predicted value of the prediction model for the target, y i is the actual value of the i-th input data.

6. The method according to claim 3, characterized in that The calculating the F1 score index value of the target evaluation prediction model includes: According to the F1 score index value calculation formula, the F1 score index value of the target evaluation prediction model is calculated, wherein the F1 score index value calculation formula is: F1 is the F1 score indicator value, q is the prediction accuracy of the target evaluation prediction model, and r is the recall rate of the target evaluation prediction model.

7. The method according to claim 1, characterized in that The method further comprises: Based on the statistical results of the slag inclusion defect blocking rate after the molten steel in the crystallizer is poured into ingots under different liquid level fluctuation conditions, the periodic liquid level abnormal fluctuation range of the molten steel is determined. When it is predicted based on the liquid level periodic fluctuation prediction model that the liquid level difference of the target predicted molten steel in the crystallizer exceeds the periodic liquid level abnormal fluctuation range, the liquid level of the target predicted molten steel undergoes periodic abnormal fluctuations.

8. A device for predicting periodic fluctuations of liquid level, characterized in that: The device comprises: A key chemical composition determination module is used to determine the key chemical composition that affects the periodic abnormal fluctuation of the target molten steel level based on a micro-segregation model, wherein the fluctuation of the target molten steel level is abnormal and changes periodically; A continuous casting machine parameter acquisition module, used to acquire continuous casting parameter data of various continuous casting machines, wherein the continuous casting parameter data include roller diameter, roller spacing and casting speed; A multi-model training module, for generating a training data set based on the roller diameter, the roller spacing, the drawing speed and the element content of the key chemical components, and training a plurality of target evaluation prediction models based on the training data set, wherein the target evaluation prediction models include a random forest model, a support vector machine model and a genetic algorithm optimized back propagation neural network model; The optimal prediction model determination module is used to calculate the comprehensive performance evaluation value of each target evaluation prediction model under multiple performance evaluation indicators, and determine the target evaluation prediction model corresponding to the maximum comprehensive performance evaluation value as the liquid level periodic fluctuation prediction model, wherein the performance evaluation indicators include the mean absolute error indicator, the mean square error indicator and the F1 score indicator.

9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting periodic fluctuations of the liquid level as claimed in any one of claims 1 to 7 is implemented.

10. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the method for predicting periodic fluctuations of the liquid level as described in any one of claims 1 to 7 is implemented.

Images