Blast furnace molten iron silicon content estimation method and system based on iron sample image

Abstract

The invention provides a blast furnace molten iron silicon content estimation method and system based on an iron sample image, and relates to the technical field of blast furnace ironmaking, and the method comprises the following steps: collecting a blast furnace field iron sample image, and calibrating according to the silicon content; constructing an iron sample image data set, and dividing the iron sample image data set into a training set, a verification set and a test set; processing the data set by adopting a data enhancement strategy; training a blast furnace molten iron silicon content estimation regression model by using the enhanced training set and verification set; designing an evaluation index, and verifying the performance of the model by using the enhanced test set; and inputting an on-site iron sample image into the trained model, and predicting the silicon content. According to the method, the computer vision technology is applied to estimation of the silicon content of the molten iron of the blast furnace, high-precision and high-real-time silicon content prediction is realized, meanwhile, the cost is reduced, the environmental adaptability is enhanced, and the problems that the number of images is limited and label distribution is unbalanced are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of blast furnace ironmaking, and in particular to a method and system for estimating silicon content in molten iron of a blast furnace based on an iron sample image. Background Art

[0002] Blast furnace ironmaking is the core link of the steel industry, and its production efficiency and product quality are directly related to the competitiveness of the entire industry. The thermal state of the blast furnace is a key parameter in the smelting process, which directly determines the quality of molten iron, fuel consumption and overall production efficiency. However, the extreme working environment inside the blast furnace, including high temperature, high pressure and strong corrosion, makes it extremely difficult to directly measure the thermal state inside the blast furnace.

[0003] In order to indirectly reflect the thermal state inside the blast furnace, the industry generally uses the silicon content of molten iron as an indicator parameter. The silicon content of molten iron is closely related to the furnace temperature. Too high or too low silicon content will have an adverse effect on the quality and fluidity of the molten iron, and thus affect the subsequent smelting and processing process. Therefore, accurately controlling the silicon content of molten iron has become the core task of regulating the thermal state of the blast furnace.

[0004] Currently, the methods for determining silicon content mainly include offline measurement method, online measurement method, mechanism model prediction method and data-driven model prediction method.

[0005] Offline measurement method: This method measures the silicon content by sampling and testing molten iron. Although it is highly accurate, it is time-consuming, usually taking 1 to 1.5 hours, resulting in time lags and making it difficult to adjust the blast furnace operating conditions in a timely manner.

[0006] Online measurement method: Based on the laser induced breakdown spectroscopy (LIBS) technology, the silicon content on the surface of molten iron is monitored in real time through a spectrometer installed on site. However, this method has high equipment costs and is easily affected by harsh working environments, such as molten iron level fluctuations, dust, and high temperature radiation. In addition, the processing of spectral signals and the quantitative analysis model still need to be further improved.

[0007] Mechanism model prediction method: According to the chemical reactions in blast furnace smelting, the mechanism model of silicon content is derived based on the principles of heat balance and conservation of matter. However, the blast furnace ironmaking process is complex and changeable, involving many random phenomena and disturbances, making it difficult for the mechanism model to accurately describe the actual process, and the simulation results are often distorted.

[0008] Data-driven model prediction method: Use historical process data captured by sensors to establish a prediction model, mainly using statistical methods and machine learning techniques. Although this method has great potential, it is highly dependent on the quality and quantity of data, and the continuity, dynamics, nonlinearity and multi-scale effects of the ironmaking process pose great challenges to the selection and construction of the model. In addition, the accuracy of some process variables is easily affected by the environment when the furnace conditions are unstable, which further increases the difficulty of model learning.

[0009] In view of the limitations of existing methods in terms of measurement accuracy, real-time performance, cost and environmental adaptability, there is an urgent need to develop a more accurate, efficient, economical and adaptable method for estimating the silicon content in molten iron to better guide the blast furnace ironmaking process and improve production efficiency and product quality. Summary of the invention

[0010] To this end, an embodiment of the present invention provides a method and system for estimating the silicon content in molten iron of a blast furnace based on an iron sample image, which is used to solve the problems of low accuracy, poor real-time performance, high cost and weak environmental adaptability in measuring the silicon content in molten iron in the prior art.

[0011] In order to solve the above problem, an embodiment of the present invention provides a method for estimating silicon content in molten iron of a blast furnace based on an iron sample image, the method comprising:

[0012] Step S1: collecting iron sample images at the blast furnace site, and calibrating the corresponding iron sample images according to the silicon content of each iron sample;

[0013] Step S2: constructing an iron sample image dataset based on the calibrated iron sample image, and dividing the iron sample image dataset into a training set, a validation set, and a test set according to a set ratio;

[0014] Step S3: using the set data enhancement strategy to perform data enhancement processing on the iron sample image dataset;

[0015] Step S4: using the image-enhanced training set and validation set to train the pre-built regression model for estimating silicon content in blast furnace iron based on iron sample images, to obtain a trained regression model for estimating silicon content in blast furnace iron based on iron sample images;

[0016] Step S5: designing evaluation indicators and using the image-enhanced test set to verify the performance of the trained blast furnace iron silicon content estimation regression model based on iron sample images;

[0017] Step S6: input the iron sample image collected on site into the trained blast furnace molten iron silicon content estimation regression model based on the iron sample image to obtain the silicon content of the iron sample corresponding to the iron sample image.

[0018] Preferably, the data enhancement strategy includes an automatic contrast and color adjustment transformation strategy, a geometric transformation, and a Cutout transformation.

[0019] Preferably, the method of using the image-enhanced training set and the validation set to train the pre-constructed blast furnace molten iron silicon content estimation regression model based on the iron sample image to obtain the trained blast furnace molten iron silicon content estimation regression model based on the iron sample image comprises:

[0020] Firstly, the parameters of the pre-constructed regression model for estimating silicon content in blast furnace iron based on iron sample images are initialized; then, the image-enhanced training set is input into the model, forward propagation calculation is performed, and the prediction result is obtained; then, the constructed weighted penalty loss function is used to calculate the error between the model prediction result and the actual silicon content, and the network parameters are updated through back propagation; during the training process, the image-enhanced validation set is regularly used to evaluate the performance of the model, and the hyperparameters of the model are adjusted according to the evaluation results; after multiple iterations of training, the network parameters are continuously optimized until the preset number of iterations is reached, at which time the model training is terminated, and finally a trained regression model for estimating silicon content in blast furnace iron based on iron sample images is obtained.

[0021] Preferably, the network structure of the pre-constructed blast furnace iron molten silicon content estimation regression model based on iron sample images is:

[0022] The first layer is the scaling layer, which first scales the image to (1,224,224) to meet the structural operation of the feature extraction network;

[0023] The second layer is the convolution and pooling layer. First, a convolution operation with a kernel size of 7×7, a step size of 2, and a padding of 3 is performed, and through batch normalization and ReLU activation function, the size of the output feature map is (64, 112, 112), where 64 represents the number of channels of the feature map, and 112×112 is the size of each feature map; secondly, a maximum pooling with a pooling window of 3×3, a step size of 2, and a padding of 1 is performed, and the size of the output feature map is (64, 56, 56), where 64 represents the number of channels of the feature map, and 56×56 is the size of each feature map;

[0024] The third to tenth layers are residual layers, which are composed of convolution, multi-scale attention module and residual connection. In one branch, a 3×3 convolution operation is first performed with a step size of 1. The fifth, seventh and ninth layers have a step size of 2 and a padding of 1. Then, batch normalization and ReLU activation function are performed, and then a convolution with a convolution kernel size of 3×3, a step size of 1 and a padding of 1 and batch normalization are input into the multi-scale attention module. The multi-scale attention module divides the input features into three branches, two of which are spliced ​​after global pooling in the X and Y directions, and then channel fusion is performed through 1×1 convolution to obtain global features. The three branches capture local features through a 3×3 convolutional layer; after obtaining the feature representations of two different scales, the corresponding weights are generated respectively, and matrix multiplication is performed with the features of the other scale, that is, the weights generated by the features of different scales are used to adjust the own features, and finally the global spatial attention representation at the current scale is obtained; finally, the spatial attention of the two different scales is added, and the weights are generated by the sigmoid function to adjust the input and obtain the final output; in the second branch of the residual layer, the input features are kept unchanged directly; after adding the features of the two branches, the features are finally output through a ReLU activation function;

[0025] The eleventh layer performs a global average pooling operation, and the output feature map size is (512, 1, 1);

[0026] The twelfth layer performs feature calibration, keeping the size of the feature map unchanged, and finally obtains a feature vector of size 512 for representing the silicon content of the iron sample;

[0027] The thirteenth layer uses a fully connected layer to map the feature map of (512, 1, 1) to the silicon content value of (1, 1, 1), and finally realizes the task of estimating the silicon content.

[0028] Preferably, a feature calibration module is introduced in the twelfth layer to smooth the feature distribution between samples with similar silicon content, thereby adjusting the extracted feature map.

[0029] Preferably, the feature calibration module is used to smooth the feature distribution between samples with similar silicon content, specifically comprising:

[0030] After receiving the features of the previous layer, first, the samples are sorted from small to large according to the silicon content label, and are evenly divided into multiple intervals; for each silicon content interval, represented by Table b, the feature mean μ of the interval is calculated by the following formula, which is arranged in ascending order according to the silicon content. b and standard deviation σ b :

[0031]

[0032]

[0033] Where N b represents the number of samples in the bth silicon content interval, z i is the feature vector of the i-th sample in the b-th silicon content interval, ⊙ represents the element-by-element multiplication operation;

[0034] Then, a symmetric kernel function is used to smooth the feature information of adjacent intervals to obtain a smoother feature distribution. The smoothed feature mean is calculated by the following formula: and standard deviation

[0035]

[0036]

[0037] Where k(b,b′) is the symmetric kernel function that measures the similarity between the bth and b′th intervals, B represents the index set of all intervals within the silicon content range; μ b′ represents the characteristic mean of the b′th silicon content interval, σ b′ Characteristic standard deviation of the b′th silicon content interval;

[0038] Finally, feature calibration is performed through normalization and recoloring;

[0039] z i ′=(z i -μ b ) / σ b ;

[0040]

[0041] Among them, z i ′ represents the standardized feature vector of the i-th sample in the b-th silicon content interval, represents the calibration feature vector of the i-th sample in the b-th silicon content interval, i=1,...,N b , b∈B.

[0042] Preferably, the weighted penalty loss function is:

[0043]

[0044] In the formula, Loss represents the weighted penalty loss function, N represents the total number of samples, and y i represents the true silicon content value of the i-th sample, It represents the estimated silicon content of the sample, and α and β are penalty factors.

[0045] The evaluation indicators include root mean square error RMSE, mean absolute error MAE and cumulative score CS;

[0046] The calculation formula of the root mean square error RMSE is:

[0047]

[0048] The calculation formula of the mean absolute error MAE is:

[0049]

[0050] The calculation formula of the cumulative score CS is:

[0051]

[0052]

[0053] In the formula, represents the actual silicon content of sample t, represents the predicted silicon content output by the model, N Si represents the total number of samples, H 0.1 (t) and H 0.05 (t) is the Heaviside step function.

[0054] Preferably, the Heaviside step function is defined as follows:

[0055]

[0056] In the formula, H x (t) is the Heaviside step function, and x is the independent variable.

[0057] The embodiment of the present invention further provides a system for estimating silicon content in molten iron of a blast furnace based on an iron sample image, which is used to implement the above-mentioned method for estimating silicon content in molten iron of a blast furnace based on an iron sample image, and specifically includes:

[0058] The iron sample image collection and calibration module is used to collect iron sample images at the blast furnace site and calibrate the corresponding iron sample images according to the silicon content of each iron sample;

[0059] A data set division module is used to construct an iron sample image data set based on the calibrated iron sample images, and divide the iron sample image data set into a training set, a validation set and a test set according to a set ratio;

[0060] A data enhancement module, used to perform data enhancement processing on the iron sample image dataset using a set data enhancement strategy;

[0061] A model training module is used to train a pre-built regression model for estimating silicon content in molten iron in a blast furnace based on an iron sample image using a training set and a validation set after image enhancement, so as to obtain a trained regression model for estimating silicon content in molten iron in a blast furnace based on an iron sample image;

[0062] The model performance verification module is used to design evaluation indicators and use the image-enhanced test set to verify the performance of the trained blast furnace iron silicon content estimation regression model based on iron sample images;

[0063] The silicon content prediction module is used to input the iron sample image collected on site into the trained blast furnace molten iron silicon content estimation regression model based on the iron sample image to obtain the silicon content of the iron sample corresponding to the iron sample image.

[0064] It can be seen from the above technical solutions that the present invention has the following beneficial effects:

[0065] (1) The present invention introduces computer vision technology into the field of silicon content estimation in the blast furnace ironmaking process for the first time. By converting this practical problem into a deep regression task, the present invention can automatically extract features from a large number of iron sample images and learn the target value of silicon content, thereby achieving accurate prediction of silicon content.

[0066] (2) Compared with traditional methods, the present invention relies entirely on iron sample images to estimate silicon content, without the need for human intervention or the support of complex sensor equipment. This innovation not only reduces dependence on external resources, but also significantly improves the estimation speed, providing more timely and effective operation guidance for on-site operators.

[0067] (3) In view of the relatively limited number of iron sample images, the present invention designs a set of efficient data enhancement methods to generate diversified training data, thereby enhancing the robustness and prediction accuracy of the model. In addition, in view of the actual situation of unbalanced label distribution in the data set, the present invention optimizes the network structure and loss function, effectively alleviating the negative impact of uneven label distribution, and further improving the overall performance of the model in the silicon content estimation task. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the implementation cases of the present invention or the technical solutions in the prior art, the following is a brief description of the drawings required for use in the embodiments. By referring to the drawings, the features and advantages of the present invention will be more clearly understood. The drawings are schematic and should not be understood as limiting the present invention in any way. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0069] Figure 1 A flow chart of a method for estimating silicon content in molten iron in a blast furnace based on an iron sample image provided in an embodiment;

[0070] Figure 2 It is a schematic diagram of on-site sampling in the embodiment;

[0071] Figure 3 It is a structural schematic diagram of a feature calibration module in an embodiment;

[0072] Figure 4 A scatter plot of estimated silicon content and actual silicon content in the examples;

[0073] Figure 5 A block diagram of a system for estimating silicon content in molten iron in a blast furnace based on an iron sample image provided in an embodiment. DETAILED DESCRIPTION

[0074] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] Embodiment 1

[0076] In order to solve the problems of low accuracy, poor real-time performance, high cost and weak environmental adaptability in measuring silicon content in molten iron in the existing technology. Figure 1 As shown, an embodiment of the present invention provides a method for estimating silicon content in molten iron of a blast furnace based on an iron sample image, the method comprising:

[0077] Step S1: collecting iron sample images at the blast furnace site, and calibrating the corresponding iron sample images according to the silicon content of each iron sample;

[0078] Step S2: constructing an iron sample image dataset based on the calibrated iron sample image, and dividing the iron sample image dataset into a training set, a validation set, and a test set according to a set ratio;

[0079] Step S3: using the set data enhancement strategy to perform data enhancement processing on the iron sample image dataset;

[0080] Step S4: using the image-enhanced training set and validation set to train the pre-built regression model for estimating silicon content in blast furnace iron based on iron sample images, to obtain a trained regression model for estimating silicon content in blast furnace iron based on iron sample images;

[0081] Step S5: designing evaluation indicators and using the image-enhanced test set to verify the performance of the trained blast furnace iron silicon content estimation regression model based on iron sample images;

[0082] Step S6: input the iron sample image collected on site into the trained blast furnace molten iron silicon content estimation regression model based on the iron sample image to obtain the silicon content of the iron sample corresponding to the iron sample image.

[0083] It can be seen from the above technical solution that the present invention proposes a method for estimating the silicon content of molten iron in a blast furnace based on iron sample images. By collecting and calibrating the iron sample images of the blast furnace site, constructing a data set and applying a data enhancement strategy, a deep regression model is trained using a training set and a validation set to automatically extract image features and learn the target value of the silicon content. This method applies computer vision technology to the estimation of silicon content in blast furnace ironmaking for the first time, overcoming the problems of traditional methods that rely on manual or complex equipment, low accuracy, and poor real-time performance, and has significant advantages such as fast speed, low cost, and strong environmental adaptability. At the same time, in response to the challenges of a limited number of images and an unbalanced label distribution, the present invention designs a special data enhancement method and an optimized network structure to improve the robustness and accuracy of the model.

[0084] In step S1, during the blast furnace tapping operation, the on-site operator will use a sampling spoon to collect molten iron samples at the molten iron ditch, such as Figure 2 As shown. When the molten iron contacts the cooler sampling spoon, it quickly solidifies into thin flakes to form an iron sample. In order to obtain the image of the iron sample, a professional shooting box is used to shoot with a camera with fixed parameters under constant color temperature and brightness conditions. This shooting method can minimize the impact of external factors on the color of the iron sample image and ensure image quality. Subsequently, these iron samples will be sent to the laboratory for testing to determine their accurate silicon content. Based on the silicon content result of each iron sample, the corresponding iron sample image will be calibrated. Throughout the process, all iron sample images are taken at the blast furnace site using a controlled light box to ensure that the lighting conditions of all images are consistent.

[0085] In step S2, an iron sample image dataset is constructed based on the calibrated iron sample images, and the iron sample image dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:1.

[0086] In step S3, in order to ensure that the training model has strong robustness, a large number of iron sample images with known silicon content are required as a data set. However, the actual situation is that only one iron sample can be provided for each molten iron, and these iron samples need to be photographed on site, which undoubtedly increases the difficulty of obtaining these images. In order to overcome this challenge, this embodiment designs a new data enhancement technology to generate more diverse training data.

[0087] Furthermore, the present embodiment introduces an automatic contrast color adjustment strategy. This strategy can simulate various lighting conditions that may be encountered during on-site shooting, and helps the model better recognize features under different lighting conditions by enhancing or weakening the light intensity in the image. In addition to color adjustment, the present embodiment also uses geometric transformation technology, including transformation methods such as ShearX, ShearY, TranslateX and TranslateY, to cope with changes in camera angles and possible slight misalignments during shooting. These geometric transformations can simulate natural changes, allowing the model to recognize image patterns in different spatial directions. In addition, the present embodiment also randomly introduces occluded areas in the image through the Cutout transformation to enhance the detection ability of the model when features are incomplete.

[0088] In step S4, considering that the overall color depth of the iron sample and the local golden spots on it have important visual indications for predicting the silicon content, it is particularly important to extract multi-scale features in the process of improving the prediction accuracy of the model. The model network structure proposed in this embodiment is designed based on this requirement. The network structure of the designed blast furnace iron silicon content estimation regression model based on the iron sample image is:

[0089] The first layer is the scaling layer, which first scales the image to (1,224,224) to meet the structural operation of the feature extraction network.

[0090] The second layer is the convolution and pooling layer. First, a convolution operation with a kernel size of 7×7, a stride of 2, and a padding of 3 is performed, and through batch normalization and ReLU activation function, the size of the output feature map is (64, 112, 112), where 64 represents the number of channels of the feature map, and 112×112 is the size of each feature map. Secondly, a maximum pooling with a pooling window of 3×3, a stride of 2, and a padding of 1 is performed, and the size of the output feature map is (64, 56, 56), where 64 represents the number of channels of the feature map, and 56×56 is the size of each feature map.

[0091] The third to tenth layers are residual layers, which are composed of convolution, multi-scale attention module and residual connection. In one branch, a 3×3 convolution operation is first performed with a step size of 1. The fifth, seventh and ninth layers have a step size of 2 and a padding of 1, and then pass through batch normalization and ReLU activation function. After that, it passes through a convolution with a kernel size of 3×3, a step size of 1, and a padding of 1 and batch normalization before entering the multi-scale attention module. The multi-scale attention module divides the input features into three branches, two of which are spliced ​​after global pooling in the X and Y directions, and then channel fusion is performed through 1×1 convolution to obtain global features. The third branch captures local features through a 3×3 convolution layer. After obtaining the feature representations of two different scales, the corresponding weights are generated respectively, and matrix multiplication is performed with the features of the other scale, that is, the weights generated by the features of different scales are used to adjust the features of the current scale, and finally the global spatial attention representation at the current scale is obtained. Finally, the spatial attention of the two different scales is added, and the weights are generated through the sigmoid function to adjust the input and obtain the final output. The feature dimension does not change after passing through the attention module. In the second branch of the residual layer, the input feature is kept unchanged; the features of the two branches are added together, and finally the features are output through a ReLU activation function.

[0092] The eleventh layer performs a global average pooling operation, and the output feature map size is (512, 1, 1).

[0093] The twelfth layer performs feature calibration, keeping the size of the feature map unchanged, and finally obtains a feature vector of size 512 for representing the silicon content of the iron sample.

[0094] The thirteenth layer uses a fully connected layer to map the feature map of (512, 1, 1) to the silicon content value of (1, 1, 1), and finally realizes the task of estimating the silicon content.

[0095] Since the number of samples with high silicon content and low silicon content is relatively small, the problem of data imbalance is caused, which makes the feature statistics more susceptible to noise interference, resulting in large differences between the feature statistics of adjacent silicon content samples. In order to meet this challenge, this embodiment innovatively introduces a feature calibration module in the twelfth layer of the network structure. The main function of this module is to smooth the feature distribution between samples with similar silicon content, so as to achieve fine adjustment of the extracted feature map.

[0096] Furthermore, if Figure 3 As shown in Table b, after receiving the features of the previous layer, first, the samples are sorted from small to large according to the silicon content label, and are evenly divided into multiple intervals; for each silicon content interval, as shown in Table b, the feature mean μ of the interval is calculated by the following formula.b and standard deviation σ b :

[0097]

[0098]

[0099] Where N b represents the number of samples in the bth silicon content interval, z i is the characteristic vector of the i-th sample in the b-th silicon content interval, and ⊙ represents the element-by-element multiplication operation.

[0100] Then, a symmetric kernel function is used to smooth the feature information of adjacent intervals to obtain a smoother feature distribution. The smoothed feature mean is calculated by the following formula: and standard deviation

[0101]

[0102]

[0103] Where k(b,b′) is the symmetric kernel function that measures the similarity between the bth and b′th intervals, B represents the index set of all intervals within the silicon content range; μ b′ represents the characteristic mean of the b′th silicon content interval, σ b′ Characteristic standard deviation of the b′th silicon content interval.

[0104] Finally, feature calibration is performed through normalization and recoloring;

[0105] z i ′=(z i -μ b ) / σ b ;

[0106]

[0107] Among them, z i ′ represents the standardized feature vector of the i-th sample in the b-th silicon content interval, represents the calibration feature vector of the i-th sample in the b-th silicon content interval, i=1,...,N b , b∈B.

[0108] By adopting the above method, this embodiment successfully achieves smoothing in the feature space of the iron sample image, making samples with similar silicon content closer in feature representation. This improvement not only reduces the interference of intrinsic noise by transferring the features of adjacent intervals, but also enables the model to learn more reliable and consistent feature representations, especially when dealing with underrepresented iron sample features, significantly improving the performance of the model on unbalanced data sets.

[0109] Next, the loss function design of this embodiment is introduced in detail. Aiming at the specific goal of the silicon content estimation task, the loss function designed in this embodiment is composed of multiple components. First, the mean square error (MSE) term is introduced to quantify the square difference between the true silicon content and the model estimated value. In view of the fact that it is difficult to achieve absolutely accurate estimates in practical applications, the hit rate with an absolute error within 0.1% or 0.05% is usually used as the evaluation criterion. Therefore, this embodiment adds two additional penalty terms to the loss function, corresponding to error thresholds of 0.1% and 0.05%, respectively. Among them, the threshold of 0.1% is set as the maximum acceptable error of silicon content estimation, while the threshold of 0.05% is intended to encourage the model to pursue higher prediction accuracy. For estimates exceeding these thresholds, corresponding penalties will be imposed to guide the model to continuously improve the accuracy of predictions.

[0110] Combining the effective label density and the above penalty threshold, this embodiment constructs a complete weighted penalty loss function, which is specifically in the following form:

[0111]

[0112] In the formula, Loss represents the weighted penalty loss function, N represents the total number of samples, and y i represents the true silicon content value of the i-th sample, It represents the estimated silicon content of the sample, and α and β are penalty factors.

[0113] Furthermore, the method of training the pre-built regression model for estimating silicon content in molten iron in blast furnace based on iron sample images using the training set and the validation set after image enhancement to obtain the trained regression model for estimating silicon content in molten iron in blast furnace based on iron sample images includes:

[0114] Firstly, the parameters of the pre-constructed regression model for estimating silicon content in blast furnace iron based on iron sample images are initialized; then, the image-enhanced training set is input into the model, forward propagation calculation is performed, and the prediction result is obtained; then, the constructed weighted penalty loss function is used to calculate the error between the model prediction result and the actual silicon content, and the network parameters are updated through back propagation; during the training process, the image-enhanced validation set is regularly used to evaluate the performance of the model, and the hyperparameters of the model are adjusted according to the evaluation results; after multiple iterations of training, the network parameters are continuously optimized until the preset number of iterations is reached, at which time the model training is terminated, and finally a trained regression model for estimating silicon content in blast furnace iron based on iron sample images is obtained.

[0115] In step S5, in order to comprehensively and accurately evaluate the performance of the model proposed in this embodiment, this embodiment uses the root mean square error (RMSE), mean absolute error (MAE) and cumulative score (CS) as core evaluation indicators. In actual production applications, when the absolute error between the silicon content predicted by the model and the actual silicon content remains within the range of 0.1% or 0.05%, such predictions are generally considered to be accurate and reliable. Therefore, this embodiment specifically introduces the CS indicator to intuitively reflect the prediction ability of the model within this critical error range. The following are the specific calculation formulas for RMSE, MAE and CS:

[0116]

[0117]

[0118]

[0119]

[0120] In the formula, represents the actual silicon content of sample t, represents the predicted silicon content output by the model, N Si represents the total number of samples, H 0.1 (t) and H 0.05 (t) is the Heaviside step function, defined as follows:

[0121]

[0122] In the formula, H x (t) is the Heaviside step function, and x is the independent variable.

[0123] In step S6, the iron sample image collected on site is input into a trained regression model for estimating silicon content in molten iron of a blast furnace based on the iron sample image to obtain the silicon content of the iron sample corresponding to the iron sample image.

[0124] In order to further illustrate the advantages of the present invention, it is described below in conjunction with specific experiments.

[0125] 1. Data preparation and experimental platform

[0126] At the blast furnace operation site, we took strict measures to collect iron samples and take photos. All images were taken on site using a controlled light box to ensure that the lighting conditions of all photos were highly consistent. At the same time, a camera with fixed parameters was used to minimize external factors that could affect the color of the iron sample images. A total of 2,040 iron sample images were successfully collected, covering a wide range of silicon content from 0.06% to 0.77%.

[0127] In order to effectively evaluate and optimize our silicon content estimation regression model, we divided the dataset into training set, validation set and test set in a ratio of 8:1:1. The model was constructed based on the PyTorch deep learning framework, with an initial learning rate of 10^-3 (i.e. 0.001) set for it, and the AdamW optimizer was selected to dynamically adjust the learning rate in order to obtain better training results.

[0128] During training, we used a batch size of 32 and performed 100 rounds of training. All of these computational tasks were completed on a high-performance computer equipped with a 13th-generation Intel Core i7-13700K processor, 32GB of RAM, and an RTX 3070 GPU (with 8GB of video memory). Such hardware configuration provides strong support for our model training.

[0129] 2. Verify model performance

[0130] After the model training is completed, the performance of the model is verified on the test set. RMSE, MAE, CS-0.1, and CS-0.05 are used as evaluation criteria. The experimental results show that the RMSE of the model reached 0.0583, MAE reached 0.0441, CS-0.05 reached 66.02%, and CS-0.1 reached 92.71%. The silicon content error range within 0.1% is effective for blast furnace operation, so it can be seen that our model has achieved a good effect.

[0131] To visually demonstrate the relationship between the estimated silicon content and the actual silicon content, we plotted a scatter plot, where the x-axis represents the actual silicon content and the y-axis represents the estimated silicon content. Figure 4As shown in Figure 2, the scatter points are concentrated near the y=x line, and most of the errors are in the range of [-0.1, 0.1]. Even in extreme cases of high or low silicon content (where estimation is more difficult), the model can still maintain high accuracy. This consistency across different silicon content levels highlights the robustness and reliability of our method in real-time silicon content estimation.

[0132] Embodiment 2

[0133] like Figure 5 As shown, the present invention provides a system for estimating silicon content in molten iron of a blast furnace based on an iron sample image, which is used to implement the method for estimating silicon content in molten iron of a blast furnace based on an iron sample image of the above-mentioned embodiment 1, and specifically includes:

[0134] The iron sample image collection and calibration module 100 is used to collect iron sample images at the blast furnace site and calibrate the iron sample images according to the silicon content of each iron sample;

[0135] The data set division module 200 is used to construct an iron sample image data set based on the calibrated iron sample image, and divide the iron sample image data set into a training set, a validation set and a test set according to a set ratio;

[0136] The data enhancement module 300 is used to perform data enhancement processing on the iron sample image data set using a set data enhancement strategy;

[0137] The model training module 400 is used to train the pre-built regression model for estimating silicon content in molten iron in blast furnace based on iron sample images using the training set and the validation set after image enhancement, so as to obtain the trained regression model for estimating silicon content in molten iron in blast furnace based on iron sample images;

[0138] The model performance verification module 500 is used to design evaluation indicators and verify the performance of the trained blast furnace iron silicon content estimation regression model based on iron sample images using the image-enhanced test set;

[0139] The silicon content prediction module 600 is used to input the iron sample image collected on site into the trained blast furnace molten iron silicon content estimation regression model based on the iron sample image to obtain the silicon content of the iron sample corresponding to the iron sample image.

[0140] A blast furnace molten iron silicon content estimation system based on iron sample images in this embodiment is used to implement the aforementioned blast furnace molten iron silicon content estimation method based on iron sample images. Therefore, the specific implementation method of the blast furnace molten iron silicon content estimation system based on iron sample images can be seen in the embodiment part of the blast furnace molten iron silicon content estimation method based on iron sample images in the previous text. For example, the iron sample image collection and calibration module 100, the data set division module 200, the data enhancement module 300, the model training module 400, the model performance verification module 500, and the silicon content prediction module 600 are respectively used to implement steps S1, S2, S3, S4, S5, and S6 in the above-mentioned blast furnace molten iron silicon content estimation method based on iron sample images. Therefore, its specific implementation method can refer to the description of the corresponding each part of the embodiment. In order to avoid redundancy, it will not be repeated here.

[0141] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0142] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0143] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0144] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from them are still within the protection scope of the invention.

Claims

1. A method for estimating silicon content in molten iron of a blast furnace based on an iron sample image, characterized in that: include: Step S1: collecting iron sample images at the blast furnace site, and calibrating the corresponding iron sample images according to the silicon content of each iron sample; Step S2: constructing an iron sample image dataset based on the calibrated iron sample image, and dividing the iron sample image dataset into a training set, a validation set, and a test set according to a set ratio; Step S3: using the set data enhancement strategy to perform data enhancement processing on the iron sample image dataset; Step S4: using the image-enhanced training set and validation set to train the pre-built regression model for estimating silicon content in blast furnace iron based on iron sample images, to obtain a trained regression model for estimating silicon content in blast furnace iron based on iron sample images; Step S5: designing evaluation indicators and using the image-enhanced test set to verify the performance of the trained blast furnace iron silicon content estimation regression model based on iron sample images; Step S6: input the iron sample image collected on site into the trained blast furnace molten iron silicon content estimation regression model based on the iron sample image to obtain the silicon content of the iron sample corresponding to the iron sample image.

2. The method for estimating silicon content in molten iron of a blast furnace based on an iron sample image according to claim 1, characterized in that: The data enhancement strategy includes an automatic contrast and color adjustment transformation strategy, a geometric transformation, and a Cutout transformation.

3. The method for estimating silicon content in molten iron of a blast furnace based on an iron sample image according to claim 1, characterized in that: The method of using the image-enhanced training set and the validation set to train the pre-built regression model for estimating silicon content in molten iron in blast furnace based on iron sample images to obtain the trained regression model for estimating silicon content in molten iron in blast furnace based on iron sample images comprises: Firstly, the parameters of the pre-constructed regression model for estimating silicon content in blast furnace iron based on iron sample images are initialized; then, the image-enhanced training set is input into the model, forward propagation calculation is performed, and the prediction result is obtained; then, the constructed weighted penalty loss function is used to calculate the error between the model prediction result and the actual silicon content, and the network parameters are updated through back propagation; during the training process, the image-enhanced validation set is regularly used to evaluate the performance of the model, and the hyperparameters of the model are adjusted according to the evaluation results; after multiple iterations of training, the network parameters are continuously optimized until the preset number of iterations is reached, at which time the model training is terminated, and finally a trained regression model for estimating silicon content in blast furnace iron based on iron sample images is obtained.

4. The method for estimating silicon content in molten iron of a blast furnace based on an iron sample image according to claim 1 or 3, characterized in that: The network structure of the pre-built blast furnace iron silicon content estimation regression model based on iron sample images is: The first layer is the scaling layer, which first scales the image to (1,224,224) to meet the structural operation of the feature extraction network; The second layer is the convolution and pooling layer. First, a convolution operation with a kernel size of 7×7, a step size of 2, and a padding of 3 is performed, and through batch normalization and ReLU activation function, the size of the output feature map is (64, 112, 112), where 64 represents the number of channels of the feature map, and 112×112 is the size of each feature map; secondly, a maximum pooling with a pooling window of 3×3, a step size of 2, and a padding of 1 is performed, and the size of the output feature map is (64, 56, 56), where 64 represents the number of channels of the feature map, and 56×56 is the size of each feature map; The third to tenth layers are residual layers, which are composed of convolution, multi-scale attention module and residual connection. In one branch, a 3×3 convolution operation is first performed with a step size of 1. The fifth, seventh and ninth layers have a step size of 2 and a padding of 1. Then, batch normalization and ReLU activation function are performed, and then a convolution with a convolution kernel size of 3×3, a step size of 1 and a padding of 1 and batch normalization are input into the multi-scale attention module. The multi-scale attention module divides the input features into three branches, two of which are concatenated after global pooling in the X and Y directions, and then channel fusion is performed through 1×1 convolution to obtain global features; The third branch captures local features through a 3×3 convolutional layer; after obtaining the feature representations of two different scales, the corresponding weights are generated respectively, and matrix multiplication is performed with the features of the other scale, that is, the weights generated by the features of different scales are used to adjust the own features, and finally the global spatial attention representation at the current scale is obtained; finally, the spatial attention of the two different scales is added, and the weights are generated by the sigmoid function to adjust the input and obtain the final output; in the second branch of the residual layer, the input features are kept unchanged directly; after adding the features of the two branches, the features are finally output through a ReLU activation function; The eleventh layer performs a global average pooling operation, and the output feature map size is (512, 1, 1); The twelfth layer performs feature calibration, keeping the size of the feature map unchanged, and finally obtains a feature vector of size 512 for representing the silicon content of the iron sample; The thirteenth layer uses a fully connected layer to map the feature map of (512, 1, 1) to the silicon content value of (1, 1, 1), and finally realizes the task of estimating the silicon content.

5. The method for estimating silicon content in molten iron of a blast furnace based on an iron sample image according to claim 4, characterized in that: A feature calibration module is introduced at the twelfth layer to smooth the feature distribution between samples with similar silicon content, thereby adjusting the extracted feature maps.

6. The method for estimating silicon content in molten iron of a blast furnace based on an iron sample image according to claim 5, characterized in that: The feature calibration module is used to smooth the feature distribution between samples with similar silicon content, and specifically includes: After receiving the features of the previous layer, first, the samples are sorted from small to large according to the silicon content label, and are evenly divided into multiple intervals; for each silicon content interval, represented by Table b, the feature mean μ of the interval is calculated by the following formula, which is arranged in ascending order according to the silicon content. b and standard deviation σ b : Where N b represents the number of samples in the bth silicon content interval, z i is the feature vector of the i-th sample in the b-th silicon content interval, ⊙ represents the element-by-element multiplication operation; Then, a symmetric kernel function is used to smooth the feature information of adjacent intervals to obtain a smoother feature distribution. The smoothed feature mean is calculated by the following formula: and standard deviation Where k(b,b′) is the symmetric kernel function that measures the similarity between the bth and b′th intervals, B represents the index set of all intervals within the silicon content range; μ b′ represents the characteristic mean of the b′th silicon content interval, σ b′ Characteristic standard deviation of the b′th silicon content interval; Finally, feature calibration is performed through normalization and recoloring; With i ′=(from i -μ b ) / σ b ; Among them, z i ′ represents the standardized feature vector of the i-th sample in the b-th silicon content interval, represents the calibration feature vector of the i-th sample in the b-th silicon content interval, i=1,...,N b , b∈B.

7. The method for estimating silicon content in molten iron of a blast furnace based on an iron sample image according to claim 3, characterized in that: The weighted penalty loss function is: In the formula, Loss represents the weighted penalty loss function, where b i is the silicon content interval to which the ith sample belongs, N represents the total number of samples, and y i represents the true silicon content value of the i-th sample, It represents the estimated silicon content of the sample, and α and β are penalty factors.

8. The method for estimating silicon content in molten iron of a blast furnace based on an iron sample image according to claim 1, characterized in that: The evaluation indicators include root mean square error RMSE, mean absolute error MAE and cumulative score CS; The calculation formula of the root mean square error RMSE is: The calculation formula of the mean absolute error MAE is: The calculation formula of the cumulative score CS is: In the formula, represents the actual silicon content of sample t, represents the predicted silicon content output by the model, N Si represents the total number of samples, H 0.1 (t) and H 0.05 (t) is the Heaviside step function.

9. The method for estimating silicon content in molten iron of a blast furnace based on an iron sample image according to claim 8, characterized in that: The Heaviside step function is defined as follows: In the formula, H x (t) is the Heaviside step function, and x is the independent variable.

10. A blast furnace iron molten silicon content estimation system based on iron sample images, characterized in that: The system is used to implement the method for estimating silicon content in molten iron in a blast furnace based on an iron sample image according to any one of claims 1 to 9, and specifically comprises: The iron sample image collection and calibration module is used to collect iron sample images at the blast furnace site and calibrate the corresponding iron sample images according to the silicon content of each iron sample; A data set division module is used to construct an iron sample image data set based on the calibrated iron sample images, and divide the iron sample image data set into a training set, a validation set and a test set according to a set ratio; A data enhancement module, used to perform data enhancement processing on the iron sample image dataset using a set data enhancement strategy; A model training module is used to train a pre-built regression model for estimating silicon content in molten iron in a blast furnace based on an iron sample image using a training set and a validation set after image enhancement, so as to obtain a trained regression model for estimating silicon content in molten iron in a blast furnace based on an iron sample image; The model performance verification module is used to design evaluation indicators and use the image-enhanced test set to verify the performance of the trained blast furnace iron silicon content estimation regression model based on iron sample images; The silicon content prediction module is used to input the iron sample image collected on site into the trained blast furnace molten iron silicon content estimation regression model based on the iron sample image to obtain the silicon content of the iron sample corresponding to the iron sample image.

Images