Blast furnace temperature anomaly prediction method based on digital twinning

Abstract

The invention provides a blast furnace temperature anomaly prediction method based on digital twinning, and relates to the technical field of digital twinning, an adaptive diffusion coefficient is determined according to a tuyere convolute area image acquired in real time, an error caused by an unfixed value of the adaptive diffusion coefficient in the prior art can be avoided, and then the temperature anomaly of a blast furnace can be predicted based on the determined adaptive diffusion coefficient. According to the method, an energy conservation equation can be determined, a velocity field is solved according to real-time observation data corresponding to the tuyere convolution area image, a mass conservation equation and a momentum conservation equation, the velocity of the velocity field is input into the energy conservation equation corresponding to the tuyere convolution area image, the equation is solved by adopting a finite difference method, and a temperature field is obtained. According to the method, the temperature field characteristics of the temperature field are extracted, the temperature field characteristics are processed through the abnormal condition prediction model, the temperature label of the temperature field is obtained, and experimental results show that blast furnace temperature anomaly prediction is improved, and the real-time performance is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital twins, and in particular to a method for predicting abnormal temperature of a blast furnace based on digital twins. Background Art

[0002] Ironmaking is a complex physicochemical process that aims to reduce a mixture of iron ore, coke, and flux into molten iron in a blast furnace. The ironmaking process has complex thermodynamic, fluid dynamic, and chemical reaction characteristics, requiring effective and precise control to stabilize production. The complexity of the ironmaking process is mainly reflected in temperature changes, gas flow distribution, reaction equilibrium, and equipment degradation. It is crucial to stabilize the ironmaking process to produce high-quality molten iron; in particular, operating conditions must be monitored and abnormal conditions detected.

[0003] Various monitoring methods have been used to detect abnormal operating conditions in blast furnaces. The existing methods are divided into model-based methods, knowledge-based methods, and data-driven methods. Although model-based or knowledge-based methods show limitations in dealing with the complexity of the blast furnace ironmaking process, various data-driven methods have been explored. For example, the deep weighted joint distributed adaptive network (DWJDAN) combines convolutional neural networks and transfer learning techniques to address the scarcity and inconsistency of real-time data. Adaptive dynamic interpretable analysis stationary subspace analysis (DiASSA) separates dynamic and static components through inference observation decomposition and iterative modeling strategies. A framework integrating the principal component analysis (PCA) method into the graph neural network (GNN) framework is proposed for diagnosing faults in blast furnaces. The multivariate statistical process control model (GTM-SPC) is improved, in which TOPSIS and grey models are used for optimization. The time series prediction model (SCPM) is based on spatiotemporal aggregation and further uses Granger causality analysis to detect faults in industrial processes. The unsupervised fault monitoring framework combines gated recurrent units (GRUs), Gaussian mixture models (GMMs), and autoencoders (AEs). Joint training methods improve performance by minimizing maximum entropy through joint optimization of feature extractors and classifiers. RSLADCGRU proposed by Li et al. uses graph neural networks to describe sparse relationships of process variables. The local-DBKSSA method uses dynamic generalized nonlinear feature extraction and local statistics to monitor nonlinear and dynamic characteristics. Regularized mutual kernel analysis stationary subspace analysis (RMK-ASSA) is combined with deep generalized stationary kernel network (DBSKNet) to handle nonlinear and nonstationary characteristics of data. Principal component analysis (PCA) and independent component analysis (ICA) are combined to monitor and diagnose anomalies in data sets with Gaussian or non-Gaussian distributions. Multiple local manifold learning improves fault detection by exploiting local geometric structures. A deep learning method called DSKL-SVM handles nonstationary data by analyzing stationary subspaces and deep kernel networks. A multi-channel dynamic graph convolutional network (MDGCN) is developed to characterize the dynamic characteristics of industrial processes by assigning dynamic weights and applying multi-level feature learning. Multiple cooperation preserving projections (MCPP) are used to extract spatiotemporal structures; multi-source heterogeneous data are used to reduce the impact of industrial noise.

[0004] Although some progress has been made in developing various data-driven methods, existing methods face many technical obstacles: First, these methods heavily rely on the quality and adequacy of data; however, it is extremely difficult to directly obtain sufficient data related to various faults. Second, the results of data-driven methods are often not supported by physical explanations; they are not interpretable and transparent. Third, when the data is non-stationary, dynamic, and has complex characteristics, data-driven methods show limited robustness and generalization ability. Finally, the accuracy of existing data-driven methods in detecting new or rare faults is unsatisfactory. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to propose a blast furnace temperature anomaly prediction method based on digital twin, comprising:

[0006] Step 1: Real-time collection of production data during blast furnace smelting, wherein the production data includes multiple tuyere raceway images and real-time observation data, wherein the real-time observation data includes at least velocity data of the tuyere and temperature data of the tuyere, and each tuyere raceway image corresponds to the real-time observation data one by one;

[0007] Step 2: for each tuyere raceway image, determine the adaptive diffusion coefficient corresponding to the tuyere raceway image, and determine the energy conservation equation corresponding to the tuyere raceway image according to the adaptive diffusion coefficient;

[0008] Step 3: For each tuyere vortex zone image, according to the real-time observation data corresponding to the tuyere vortex zone image, as well as the mass conservation equation and the momentum conservation equation, the velocity field is solved, and the velocity of the velocity field is input into the energy conservation equation corresponding to the tuyere vortex zone image. The finite difference method is used to solve the equation to obtain the temperature field;

[0009] Step 4: Extract features from the temperature field to obtain temperature field features, and process the temperature field features through an abnormal situation prediction model based on a random forest regression algorithm to obtain temperature labels of the temperature field, where the temperature labels include abnormally low temperature, normal temperature, and abnormally high temperature. The abnormal situation prediction model is obtained by training the model based on multiple first training samples, where the first training sample includes a first input sample and a first output sample, where the first input sample is a historical temperature field feature, and the first output sample is a temperature label corresponding to the historical temperature field feature.

[0010] Optionally, in step 2, for each tuyere raceway image, determining the diffusion coefficient corresponding to the tuyere raceway image specifically includes:

[0011] For each air outlet vortex area image, the texture features of the air outlet vortex area image are extracted through the gray level co-occurrence matrix, and then the texture features corresponding to each air outlet vortex area image are obtained; the texture features, the speed data of the air outlet and the temperature data of the air outlet are input into the adaptive diffusion coefficient construction model to obtain the adaptive diffusion coefficient corresponding to the air outlet vortex area image, wherein the adaptive diffusion coefficient construction model is obtained by model training based on multiple second training samples, the second training sample includes a second input sample and a second output sample, the second input sample is the texture features of the historical air outlet vortex area image, the speed data of the air outlet and the temperature data of the air outlet, and the second output sample is the adaptive diffusion coefficient of the historical air outlet vortex area image.

[0012] Optionally, the texture features include contrast, correlation, energy and homogeneity.

[0013] Optionally, in step 3, the velocity field is obtained by solving the real-time observation data corresponding to the tuyere whirlpool image, as well as the mass conservation equation and the momentum conservation equation, which specifically includes:

[0014] The real-time observation data corresponding to the tuyere whirlpool image is input into the mass conservation equation and the momentum conservation equation to obtain the target equation, the finite volume method is used to discretize the target equation, the initial number of iterations is set, the initial velocity field under the current number of iterations is set, and the predicted velocity field is solved according to the initial velocity field and the discretized target equation, and the predicted velocity field is revised to obtain the revised velocity field, and it is judged whether the current number of iterations reaches the preset threshold. When the current number of iterations reaches the preset threshold, the revised velocity field is output, and when the current number of iterations does not reach the preset threshold, the current number of iterations is increased by one and used as the new current number of iterations, and the execution is returned: the initial velocity field under the current number of iterations is set until the current number of iterations reaches the preset threshold.

[0015] Optionally, the temperature field characteristics include the average temperature, average temperature gradient, maximum temperature gradient, maximum temperature difference, standard deviation of the temperature gradient and median temperature gradient of the tuyere convolution zone.

[0016] Optionally, after the velocity field is obtained in step 3, the following steps are further included:

[0017] According to the size and shape of the blast furnace, a volume fitting coordinate grid is generated based on the elliptic partial differential equation. The velocity field is combined with the volume fitting coordinate grid to obtain a visual representation of the velocity field.

[0018] Optionally, after obtaining the temperature field in step 3, the method further includes:

[0019] According to the size and shape of the blast furnace, a volume fitting coordinate grid is generated based on the elliptic partial differential equation. The temperature field is combined with the volume fitting coordinate grid to obtain a visual representation of the velocity field.

[0020] The beneficial effects of adopting the above technical solution are:

[0021] The present invention determines the adaptive diffusion coefficient based on the real-time collected tuyere whirlpool area image, which can avoid the error caused by the unfixed value of the adaptive diffusion coefficient in the prior art, and then based on the determined adaptive diffusion coefficient, the energy conservation equation can be determined, and then the velocity field is solved according to the real-time observation data corresponding to the tuyere whirlpool area image, as well as the mass conservation equation and the momentum conservation equation, and the velocity field is input into the energy conservation equation corresponding to the tuyere whirlpool area image, and the finite difference method is used to solve the equation to obtain the temperature field, and then the temperature field characteristics of the temperature field are extracted, and then the temperature field characteristics are processed by the abnormal situation prediction model to obtain the temperature label of the temperature field. According to experimental results, the present invention improves the prediction of blast furnace temperature anomalies and significantly improves the real-time performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a flow chart of a method for predicting abnormal temperature of a blast furnace based on digital twin in an embodiment of the present invention;

[0023] Figure 2 A schematic diagram of a process of constructing a model through an adaptive diffusion coefficient in an embodiment of the present invention;

[0024] Figure 3 A schematic diagram of a flow chart of a method for solving a mechanism model in an embodiment of the present invention;

[0025] Figure 4 1 are images of the tuyere raceway under three temperature conditions in an embodiment of the present invention, wherein (a) is an image of the tuyere raceway at abnormally low temperature, (b) is an image of the tuyere raceway at normal temperature, and (c) is an image of the tuyere raceway at abnormally high temperature;

[0026] Figure 5 1 are temperature field images under three temperature conditions in an embodiment of the present invention, wherein Figure (a) is a temperature field image of abnormally low temperature, Figure (b) is a temperature field image of normal temperature, and Figure (c) is a temperature field image of abnormally high temperature;

[0027] Figure 6 A schematic diagram of a process of processing by an abnormal situation prediction model in an embodiment of the present invention;

[0028] Figure 7 It is a schematic diagram of the comparison results of the random forest model on the measured and predicted diffusion coefficients in an embodiment of the present invention;

[0029] Figure 8 Schematic diagram of comparison results of top edge average temperature under different diffusion coefficients in an embodiment of the present invention;

[0030] Fig. 9 Schematic diagram of comparison results of learning curves of four algorithms in an embodiment of the present invention;

[0031] Fig.10 Schematic diagram of a confusion matrix in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0033] In view of the problems existing in the prior art, two innovations are used to improve the existing data-driven methods to monitor the ironmaking process of blast furnaces: (1) A digital twin (DT) framework based on hybrid modeling is proposed, which combines mechanism modeling and data-driven modeling to detect abnormal conditions in blast furnaces. (2) An adaptive diffusion coefficient is introduced to improve the accuracy of the temperature determined by simplifying the energy conservation equation. Specifically, the present invention provides a blast furnace temperature anomaly prediction method based on digital twins, combined with Figure 1 , which may include the following steps:

[0034] Step 1: Real-time collection of production data in a blast furnace production process, wherein the production data includes a plurality of tuyere raceway images and real-time observation data, wherein the real-time observation data includes at least velocity data of the tuyere and temperature data of the tuyere, and each tuyere raceway image corresponds to the real-time observation data one by one;

[0035] Step 2: for each tuyere raceway image, determine the adaptive diffusion coefficient corresponding to the tuyere raceway image, and determine the energy conservation equation corresponding to the tuyere raceway image according to the adaptive diffusion coefficient;

[0036] It should be noted that traditionally, the diffusion coefficient α in the energy conservation equation is fixed, as shown in the following formula:

[0037]

[0038] Where k is the thermal conductivity, ρ is the density of the material, and c v Represents heat capacity at constant volume.

[0039] In traditional methods, the diffusion coefficient α is treated as a constant, which simplifies the modeling of the heat transfer process. However, this assumption may not accurately reflect the complexity of the real system. For example, in blast furnace operation, heat transfer is affected by dynamic factors such as changes in material composition, furnace operating conditions, and temperature gradients. The fixed diffusion coefficient assumes uniform material properties, but in reality, due to phase changes, chemical reactions, and non-uniform material distribution, thermal conductivity k and specific heat capacity c are often different. v will change significantly. These changes may lead to deviations in temperature predictions. Specifically, the texture features of the flame image (contrast, correlation, brightness, and uniformity) are closely related to the diffusion coefficient. Higher contrast indicates more significant differences, and intensity changes in the flame generally increase gas mixing and the diffusion coefficient. High brightness is associated with strong combustion reactions, resulting in increased diffusion coefficients. Higher flame correlation indicates more stable gas flow, thereby reducing the diffusion coefficient, while lower uniformity reflects unstable gas flow, which increases the diffusion coefficient. In addition, the velocity and temperature at the nozzle directly affect the diffusion coefficient. Higher nozzle velocities and temperatures enhance gas energy and mixing, thereby increasing the diffusion coefficient.

[0040] To address these issues, an adaptive diffusion coefficient model can be used. This model adjusts the diffusion coefficient based on real-time data to more accurately represent the heat transfer dynamics. By accounting for variability and adapting to changing conditions, the adaptive model can significantly improve the accuracy of temperature assessments and the performance of the entire system.

[0041] In step 2, for each tuyere vortex area image, the diffusion coefficient corresponding to the tuyere vortex area image is determined, and combined with Figure 2 Specifically include:

[0042] For each tuyere vortex area image, the texture features of the tuyere vortex area image are extracted through the gray level co-occurrence matrix, that is, Figure 2 The texture features extracted by the texture analysis and feature extraction module of the air outlet image include contrast, correlation, energy and homogeneity; and then the texture features corresponding to each air outlet raceway image are obtained; the texture features, the speed data of the air outlet and the temperature data of the air outlet are input into the adaptive diffusion coefficient construction model to obtain the adaptive diffusion coefficient corresponding to the air outlet raceway image, wherein the adaptive diffusion coefficient construction model is obtained by model training based on multiple second training samples, the second training sample includes a second input sample and a second output sample, the second input sample is the texture features of the historical air outlet raceway image, the speed data of the air outlet and the temperature data of the air outlet, and the second output sample is the adaptive diffusion coefficient of the historical air outlet raceway image;

[0043] Among them, the outlet temperature data is obtained based on vision, and the outlet speed data is collected based on sensors.

[0044] Based on this, the diffusion coefficient is determined, and then the diffusion coefficient is substituted into the energy conservation equation to obtain the final energy conservation equation.

[0045] Step 3: For each tuyere vortex zone image, according to the real-time observation data corresponding to the tuyere vortex zone image, as well as the mass conservation equation and the momentum conservation equation, the velocity field is solved, and the velocity of the velocity field is input into the energy conservation equation corresponding to the tuyere vortex zone image. The finite difference method is used to solve the equation to obtain the temperature field;

[0046] Combination Figure 3 , which is a flow chart of the entire step 3, wherein the mechanism model includes the energy conservation equation, the mass conservation equation and the momentum conservation equation. In the specific implementation, the energy conservation equation needs to be simplified to obtain the simplified energy conservation equation, that is, Figure 3 The gas temperature model in can be specifically expressed by the following formula:

[0047]

[0048] The mass conservation equation is expressed by the following formula:

[0049]

[0050] The momentum conservation equation is expressed by the following formula:

[0051]

[0052] Combination Figure 3 In step 3, the velocity field is obtained by solving the real-time observation data corresponding to the tuyere whirlpool image, as well as the mass conservation equation and momentum conservation equation, including:

[0053] The real-time observation data corresponding to the tuyere whirlpool image is input into the mass conservation equation and the momentum conservation equation to obtain the target equation, the finite volume method is used to discretize the target equation, the initial number of iterations is set, the initial velocity field under the current number of iterations is set, and the predicted velocity field is solved based on the initial velocity field and the discretized target equation, and the predicted velocity field is revised to obtain the revised velocity field, and it is judged whether the current number of iterations reaches the preset threshold. When the current number of iterations reaches the preset threshold, the revised velocity field is output, and when the current number of iterations does not reach the preset threshold, the current number of iterations is increased by one and used as the new current number of iterations, and the execution is returned: the initial velocity field under the current number of iterations is set until the current number of iterations reaches the preset threshold, that is, Figure 3 Method for solving the flowing gas model in .

[0054] After the velocity field is obtained in step 3, the following steps are also included:

[0055] According to the size and shape of the blast furnace, based on the elliptic partial differential equation, a volume fitting coordinate grid is generated, and the velocity field is combined with the volume fitting coordinate grid to obtain a visual representation of the velocity field. Similarly, according to the size and shape of the blast furnace, based on the elliptic partial differential equation, a volume fitting coordinate grid is generated, and the temperature field is combined with the volume fitting coordinate grid to obtain a visual representation of the velocity field, that is, Figure 3 In the mesh generation part, is the elliptic partial differential equation, 29.8 m is the height of the blast furnace, and 273 horizontal lines and 61 vertical lines are the dimensions of the grid.

[0056] Combination Figure 3 , step 3 can specifically include the following parts:

[0057] A. Mesh Generation and Boundary Conditions

[0058] The blast furnace consists of five main parts: throat, body, waist, belly, and furnace bottom. A body-fitted coordinate (BFC) mesh was created using elliptic partial differential equations. To reduce computational complexity, the solid model was simplified using the symmetric structure of the blast nozzle and the furnace, resulting in a 2D model. The modeling region extends from the bottom of the blast nozzle to the throat of the furnace, where lateral temperature measurements were taken. The following boundary conditions (BCs) were defined to simulate the temperature distribution in the blast furnace:

[0059] (1) Inlet: The nozzle is modeled as an inlet boundary where the values ​​of the field variables are set based on field measurements.

[0060] (2) Outlet: The top of the furnace is modeled as an outlet, and its pressure is specified by measurement.

[0061] (3) No-slip surface: The inner wall of the furnace is set as a no-slip surface for the molten metal to flow; therefore, the wall is impermeable and the relative velocity is zero.

[0062] (4) Symmetry plane: The amplitude and gradient of the velocity perpendicular to the symmetry plane are set to 0.

[0063] B. Gas Flow Model

[0064] Solving the Navier-Stokes equations also requires modeling the coupling of pressure and velocity in gas flows. In the flow model, the source term F usually represents the drag effect evaluated by the Ergen equation. In addition, a lift source term F: is introduced to account for the buoyancy effect caused by the temperature gradient. The velocity field can be obtained by solving the mass conservation and momentum conservation equations simultaneously.

[0065] C. Gas Temperature Model

[0066] The unsimplified energy conservation equation includes various energy transfer mechanisms, such as convection transfer, heat conduction, thermal radiation, work transfer, chemical energy change, heat source, and latent heat of phase change. In the gas temperature model, the temperature distribution reflects the relationship between gas convection transfer, heat conduction, heat exchange between gas and solid phase, and chemical reaction, and its temperature distribution is characterized by time and space changes. In order to consider the influence of chemical reactions on temperature distribution, the source term Q of chemical reactions is specified.

[0067] In the iron-making process, the temperature distribution in the blast furnace is an important indicator for measuring the operating status. Under normal circumstances, the temperature changes steadily and the corresponding smelting process is stable. Since the iron-making process in reality is affected by many factors, when an abnormal situation occurs, the temperature change becomes unstable. The present invention studies two typical abnormal situations, namely, the abnormal decrease and increase of the blast furnace temperature. The decrease in the blast furnace temperature will affect the efficiency and quality of the smelting process, and then cause the cooling system to overload. The increase in the blast furnace temperature will accelerate the loss of the refractory material of the furnace wall, affect the furnace lining, and thus cause the blast furnace operation to be interrupted, so it is necessary to predict the temperature abnormality, which is specifically achieved through step 4.

[0068] like Figure 4 Figure (a) shows the image of the wind outlet vortex area at abnormally low temperature, Figure (b) shows the image of the wind outlet vortex area at normal temperature, and Figure (c) shows the image of the wind outlet vortex area at abnormally high temperature. Figure 5 Figure (a) shows the temperature field image of abnormally low temperature, Figure (b) shows the temperature field image of normal temperature, and Figure (c) shows the temperature field image of abnormally high temperature. Figure 4 and Figure 5 From the results, it can be seen that there are clear differences between the measured and predicted temperature fields. These differences indicate the changes in the heat transfer conditions in the blast furnace and the need to extract and characterize the temperature distribution and proceed to step 4.

[0069] Step 4: Combine Figure 6 , extract the characteristics of the temperature field to obtain the temperature field characteristics, wherein the temperature field characteristics include the average temperature, average temperature gradient, maximum temperature gradient, maximum temperature difference, standard deviation of temperature gradient and median temperature gradient of the tuyere whirlpool area, wherein, Figure 1 and Figure 6 The furnace temperature field in is the temperature field in step 4. Figure 6 The region of interest in is the tuyere swirl region.

[0070] The temperature field characteristics are processed by an abnormal situation prediction model based on a random forest regression algorithm to obtain temperature labels of the temperature field, wherein the temperature labels include abnormally low temperature, normal temperature, and abnormally high temperature. The abnormal situation prediction model is obtained by model training based on multiple first training samples, wherein the first training sample includes a first input sample and a first output sample, wherein the first input sample is a historical temperature field feature, and the first output sample is a temperature label corresponding to the historical temperature field feature.

[0071] Among them, during model training, an initial model is first constructed based on the random forest regression algorithm, and then the initial model is trained. The initial model can refer to the model in the document "Random Forests".

[0072] Among them, the abnormally low temperature, normal temperature and abnormally high temperature included in the temperature label are artificially marked. The abnormally low temperature indicates that the temperature of the tuyere swirl zone is lower than the first preset threshold, which may be 1710°C. The normal temperature indicates that the temperature of the tuyere swirl zone is in a threshold interval, which may be 1726°C to 2000°C. The abnormally high temperature indicates that the temperature of the tuyere swirl zone is higher than the second preset threshold, which may be 2234°C to 2482°C.

[0073] The above contents are verified experimentally, including:

[0074] A. Evaluation of Adaptive Diffusion Coefficient

[0075] In order to evaluate the suitability of machine learning regression models for specific regression tasks, a comparative study was conducted using support vector machine (SVM), K nearest neighbor (KNN), random forest (RF), linear regression (LR), ridge regression (RR), and gradient boosting regression (GBR) on 1000 constructed samples. 5% of the samples were used as samples in the test dataset. Root mean square error (RMSE) and mean absolute percentage error (MAPE) were used as evaluation criteria. The two criteria were calculated as follows:

[0076]

[0077] Among them, y i and are measured values ​​and predicted values ​​respectively; n is the number of samples.

[0078] The results of the six regression models are compared from the perspective of the two evaluation criteria in Table 1. The comparison results show that the RF regression model predicts the diffusion coefficient best. Figure 7 The measured values ​​were further compared with the diffusion coefficients predicted by the RF regression model.

[0079] Table 1 Comparison of different regression models

[0080] Regression Model <![CDATA[RMSE(10 -3 m 2 / s)]]> MAPE(%) SVM 5.497 4.413 KNM 2.102 1.480 LR 2.373 1.639 RR 2.354 1.626 GBR 2.156 1.460 RF 2.040 1.438

[0081] B. Temperature distribution

[0082] To verify the effectiveness of using adaptive diffusion coefficients to improve the accuracy of predicted temperature fields, the results using adaptive diffusion coefficients were compared with those using fixed diffusion coefficients.

[0083] In addition, if Figure 8 As shown, the temperature at the top edge of the furnace was measured and averaged. Clearly, when the diffusion coefficient is adaptive, the average error is 3.17%, while, by comparison, the average error is 6.90% when the diffusion coefficient is fixed.

[0084] C. Evaluation of abnormal condition prediction models

[0085] To evaluate the accuracy of the proposed RF spectrum (RF) method, it is compared with three other classification models, including K-nearest neighbor (KNN), naive Bayes classifier (NB), and support vector machine (SVM). 30% of the samples are used as test samples, and Table 2 compares the results of these four classifiers from the perspective of classification accuracy.

[0086] Table 2 Comparison of different classification methods

[0087] method Accuracy (%) NB 83.77 SVM 87.35 KNM 95.58 RF 97.49

[0088] In order to evaluate the generalization ability of the proposed model, a 5-fold cross validation was performed as follows. The training dataset was randomly divided into 5 subsets. One of the subsets was used as the validation set, and the remaining 4 subsets were used as the training dataset. This cycle was repeated 5 times, and the results of 5 groups were obtained.

[0089] Finally, the training and validation results were obtained. The performance evaluation results of a classification method are:

[0090]

[0091] Among them, N correct Indicates the number of samples classified correctly; N total is the total number of data samples.

[0092] The average of these 5 groups is taken as the final training score and validation score of the model. Fig. 9 The learning curves for each model are shown.

[0093] Table 2 and Fig. 9 It shows that Random Forest (RF) performs best. The accuracy, precision and recall are used to evaluate the comprehensive performance of the constructed RF spectrum classification model. Table III shows the specific results. Fig.10The confusion matrix is ​​shown to visually demonstrate how the model classifies different categories. The model achieves an accuracy of 97.49% on the test set. The proposed prediction method performs well in terms of accuracy and has a significant improvement in prediction speed. To monitor and detect anomalies, all these methods are implemented in Python 3.8 and executed on a local computer. The computer is equipped with an Intel Core i5-13600KF CPU (5.1 GHz) and 64GB of memory. The prediction time is always within 5 minutes, which means that the requirements of real-time digital twin applications are met.

[0094] Monitoring and detecting anomalies are crucial to ensure safety and improve production efficiency. We propose a digital twin (DT) framework based on hybrid modeling to detect anomalies in production. First, a simplified mechanistic model is established, and an adaptive diffusion coefficient is introduced to improve the accuracy and determine the temperature field. Then, the temperature field is described by several key attributes, which are used as inputs to the classification model to train and identify anomalies. Experimental results show that the proposed DT framework effectively integrates mechanistic modeling and data-driven modeling. This integration significantly improves the real-time performance.

[0095] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) to form a technical solution.

Claims

1. A method for predicting abnormal temperature of a blast furnace based on digital twin, characterized in that: include: Step 1: Real-time collection of production data in a blast furnace production process, wherein the production data includes a plurality of tuyere raceway images and real-time observation data, wherein the real-time observation data includes at least velocity data of the tuyere and temperature data of the tuyere, and each tuyere raceway image corresponds to the real-time observation data one by one; Step 2: for each tuyere raceway image, determine the adaptive diffusion coefficient corresponding to the tuyere raceway image, and determine the energy conservation equation corresponding to the tuyere raceway image according to the adaptive diffusion coefficient; Step 3: For each tuyere vortex zone image, according to the real-time observation data corresponding to the tuyere vortex zone image, as well as the mass conservation equation and the momentum conservation equation, the velocity field is solved, and the velocity of the velocity field is input into the energy conservation equation corresponding to the tuyere vortex zone image. The finite difference method is used to solve the equation to obtain the temperature field; Step 4: Extract features from the temperature field to obtain temperature field features, and process the temperature field features through an abnormal situation prediction model based on a random forest regression algorithm to obtain temperature labels of the temperature field, where the temperature labels include abnormally low temperature, normal temperature, and abnormally high temperature. The abnormal situation prediction model is obtained by training the model based on multiple first training samples, where the first training sample includes a first input sample and a first output sample, where the first input sample is a historical temperature field feature, and the first output sample is a temperature label corresponding to the historical temperature field feature.

2. The method for predicting abnormal blast furnace temperature based on digital twin according to claim 1 is characterized in that: In step 2, for each tuyere raceway image, the diffusion coefficient corresponding to the tuyere raceway image is determined, specifically including: For each air outlet vortex area image, the texture features of the air outlet vortex area image are extracted through the gray level co-occurrence matrix, and then the texture features corresponding to each air outlet vortex area image are obtained; the texture features, the speed data of the air outlet and the temperature data of the air outlet are input into the adaptive diffusion coefficient construction model to obtain the adaptive diffusion coefficient corresponding to the air outlet vortex area image, wherein the adaptive diffusion coefficient construction model is obtained by model training based on multiple second training samples, the second training sample includes a second input sample and a second output sample, the second input sample is the texture features of the historical air outlet vortex area image, the speed data of the air outlet and the temperature data of the air outlet, and the second output sample is the adaptive diffusion coefficient of the historical air outlet vortex area image.

3. The method for predicting abnormal temperature of a blast furnace based on digital twin according to claim 2 is characterized in that: The texture features include contrast, correlation, energy and homogeneity.

4. The method for predicting abnormal temperature of a blast furnace based on digital twin according to claim 1 is characterized in that: In step 3, the velocity field is obtained by solving the real-time observation data corresponding to the tuyere whirlpool image, as well as the mass conservation equation and the momentum conservation equation, including: The real-time observation data corresponding to the tuyere whirlpool image is input into the mass conservation equation and the momentum conservation equation to obtain the target equation, the finite volume method is used to discretize the target equation, the initial number of iterations is set, the initial velocity field under the current number of iterations is set, and the predicted velocity field is solved according to the initial velocity field and the discretized target equation, and the predicted velocity field is revised to obtain the revised velocity field, and it is judged whether the current number of iterations reaches the preset threshold. When the current number of iterations reaches the preset threshold, the revised velocity field is output, and when the current number of iterations does not reach the preset threshold, the current number of iterations is increased by one and used as the new current number of iterations, and the execution is returned: the initial velocity field under the current number of iterations is set until the current number of iterations reaches the preset threshold.

5. The method for predicting abnormal temperature of a blast furnace based on digital twin according to claim 1, characterized in that: The temperature field characteristics include the average temperature, average temperature gradient, maximum temperature gradient, maximum temperature difference, standard deviation of temperature gradient and median temperature gradient in the tuyere rotation zone.

6. The method for predicting abnormal temperature of a blast furnace based on digital twin according to claim 1, characterized in that: After the velocity field is obtained in step 3, the following steps are also included: According to the size and shape of the blast furnace, a volume fitting coordinate grid is generated based on the elliptic partial differential equation. The velocity field is combined with the volume fitting coordinate grid to obtain a visual representation of the velocity field.

7. The method for predicting abnormal temperature of a blast furnace based on digital twin according to claim 1, characterized in that: After obtaining the temperature field in step 3, the following steps are also included: According to the size and shape of the blast furnace, a volume fitting coordinate grid is generated based on the elliptic partial differential equation. The temperature field is combined with the volume fitting coordinate grid to obtain a visual representation of the velocity field.

Images