Iron ore sintering fuel particle size identification control system and method

Abstract

The application discloses an iron ore sintering fuel particle size identification control system, which comprises an image acquisition unit arranged above the end of a belt at a position behind a crusher and used for acquiring image data of finished products after crushing; a server used for acquiring the image data of the image acquisition unit; and an industrial computer used for acquiring transmission data of the server, calculating a prediction result of roller spacing of the crusher, and issuing an instruction for adjusting the roller spacing of the crusher. The application further discloses an iron ore sintering fuel particle size identification control method. The application improves the automatic control level of fuel particle size, achieves the purposes of stabilizing sintering fuel particle size, reducing labor intensity, and improving the automatic and intelligent level, thereby supporting stable and smooth sintering production.

Description

Technical Field

[0001] This invention relates to industrial material particle size identification and control technology, and more specifically, to a fuel particle size identification control system and method for iron ore sintering. Background Technology

[0002] Iron ore sintering is the first high-temperature process in the long-process iron and steel production, with sinter accounting for more than 70% of the blast furnace burden. Coke powder is the main fuel in iron ore sintering, and its combustion behavior in the ore bed is affected by its particle size, directly influencing the thermal state of the sintering process and consequently the quality and yield of the sinter. Furthermore, coke powder is a major source of pollutants in the sintering flue gas. Controlling the particle size of coke powder, and thus its combustion behavior in the burden, is an effective means to improve sintering quality and reduce emissions.

[0003] The particle size distribution of sintering fuel has a significant impact on sinter yield, quality indicators, and flue gas pollutant emissions. The main function of sintering fuel is to provide heat, and coke powder or anthracite is usually used as the main fuel, with a moisture content between 7% and 15%. In order to meet the particle size requirements of the sintering process, the fuel usually needs to undergo a two-stage crushing process: coarse crushing with a double-roll crusher and fine crushing with a four-roll crusher.

[0004] To ensure the stability and quality of the sintering process, strict requirements are placed on fuel particle size. Typically, coke particles smaller than 3mm must account for 70-90%. Appropriate fuel particle size is crucial for sintering product quality indicators and fuel consumption. If the fuel particle size is too fine, the combustion rate is too fast, which is not conducive to heat accumulation, affecting the maximum achievable temperature of the material layer and reducing sintering quality. If the fuel particle size is too coarse, insufficient fuel dispersion leads to excessively high local temperatures and increased reducibility, easily causing problems such as a wider combustion zone, reduced permeability of the sintered material, and lower yield. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a fuel particle size identification and control system and method for iron ore sintering, thereby improving the level of automatic control of fuel particle size, stabilizing sintering fuel particle size, reducing labor intensity, and improving the level of automation and intelligence, thus supporting the stable and smooth operation of sintering production.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] The first aspect of this invention provides a fuel particle size identification and control system for iron ore sintering, comprising:

[0008] The image acquisition unit is located above the end of the belt at the rear of the crusher and is used to acquire image data of the crushed finished fuel.

[0009] A server is used to acquire image data from the image acquisition unit.

[0010] An industrial control computer is used to acquire the data transmitted from the server, calculate the predicted result of the roller spacing of the crusher, and issue an instruction to adjust the roller spacing of the crusher.

[0011] Preferably, the image acquisition unit is a camera or an industrial camera.

[0012] The second aspect of this invention provides a method for identifying and controlling the particle size of fuel in iron ore sintering. This method employs the same particle size identification and control system for iron ore sintering as described in the first aspect of this invention. Based on the image data detection results of the crushed finished fuel acquired by the image acquisition unit, and the working status of the equipment at the production site, the industrial control computer calculates the predicted result of the roller spacing of the crusher and issues an instruction to adjust the roller spacing of the crusher. This adjusts the roller spacing between the upper and lower rollers of the crusher during the production process, thereby achieving closed-loop particle size control during the sintering fuel preparation stage.

[0013] Preferably, the fuel particle size identification and control method includes the following steps:

[0014] S1, use the image acquisition unit to acquire image data of the crushed finished fuel;

[0015] S2, perform fuel particle size image data processing on the image data;

[0016] S3. A training dataset is constructed based on the MatchPattern method and applied to the PredictKNN algorithm. By averaging the target output data of the data points, the prediction result of the roller spacing of the crusher is obtained.

[0017] S4, based on the image acquisition unit monitoring the particle size of the finished fuel, the industrial control computer automatically adjusts the roller spacing of the crusher through a feedback mechanism to ensure the stability of the fuel particle size.

[0018] Preferably, step S1 specifically includes the following steps:

[0019] S11, The image acquisition unit is positioned above the end of the belt at the rear end of the crusher;

[0020] S12, the image acquisition unit's shooting angle covers the entire width of the belt, and the height of the image acquisition unit is adjusted according to the size of the fuel particles and the width of the belt.

[0021] Preferably, step S2, which uses a ResNet model, an XGBoost model, and a Bayesian optimization method to process fuel particle size image data, specifically includes the following steps:

[0022] S21, replace the output layer of the ResNet model with an output layer suitable for fuel particle size classification;

[0023] S22 utilizes the XGBoost model to improve the gradient boosting algorithm. When solving for the extreme value of the loss function, Newton's method is used, and the loss function is expanded to the second order using Taylor expansion. In addition, a regularization term is added to the loss function. During training, the objective function consists of two parts: the first part is the gradient boosting algorithm loss, and the second part is the regularization term.

[0024] S23. The fuel particle size identification model constructed using the ResNet and XGBoost models is tuned based on the Bayesian optimization method; a ResNet model is constructed for feature extraction and input into the XGBoost model for classification; hyperparameter tuning is performed using BayesSearchCV and the Bayesian optimization method is executed; after tuning based on the Bayesian optimization method, the optimal hyperparameter combination is found and the performance of the optimized model is evaluated using a test set.

[0025] Preferably, in step S22, the loss function is defined as follows:

[0026]

[0027] Where n is the number of training function samples; 1 represents the loss on a single sample, assumed to be a convex function y. i ' represents the model's prediction of the training samples, y' is the prediction value of the training samples. i The true label values ​​of the training samples;

[0028] The regularization term defines the complexity of the model:

[0029]

[0030] Where γ and λ are manually set parameters; ω is the vector formed by the values ​​of all leaf nodes in the decision tree; and T is the number of leaf nodes.

[0031] Preferably, step S3 specifically includes the following steps:

[0032] S31, based on the training dataset, calculates the Euclidean distance between the target data and the input data using the EuclideanDistance method, and stores the distance and its corresponding output data using the Neighbor class;

[0033] S32, sort the Euclidean distances between the target data and the input data, and find the three historical data points that are closest to the target data;

[0034] S33 uses intelligent averaging of the target output data from these three historical data points to obtain a predicted result for the roller spacing of the crusher.

[0035] This invention provides a fuel particle size identification control system and method for iron ore sintering, installed above the end of the belt conveyor after the crusher to ensure accurate and stable acquisition of image data of the crushed finished fuel. A pre-trained ResNet-50 model is used as the base model, with the output layer of the ResNet model replaced by an output layer suitable for particle size classification. XGBoost is used to improve the gradient boosting algorithm, and Newton's method is used to solve for the extreme value of the loss function. Feature extraction and the XGBoost model are combined, and BayesSearchCV is used for hyperparameter tuning. Bayesian optimization is performed to find the optimal hyperparameter combination, and the performance of the optimized model is evaluated using a test set. The MatchPattern method is used to intelligently extract status codes and working state patterns from historical datasets, accurately matching historical data samples with the same working state pattern as the current one, thereby constructing a high-quality training dataset. The Neighbor class is defined in the algorithm implementation, and the EuclideanDistance method is implemented, providing core support for the KNN prediction algorithm. By utilizing the EuclideanDistance method to calculate the Euclidean distance between the target data and the input data, and storing these distances and their corresponding output data through a Neighbor class, the Euclidean distances between the target data and the input data are sorted to identify the three historical data points closest to the target data. The target output data from these three data points is intelligently averaged to obtain the predicted result for the crusher roller spacing. This achieves the PredictKNN method's prediction and adjustment of the crusher roller spacing. Through the tight coupling of fuel particle size identification and roller spacing adjustment, the entire system enables efficient linkage from raw material detection to crushing control, not only improving the automation level of the production process but also playing a crucial role in ensuring product quality and energy efficiency. This innovative combination transforms fuel particle size control from traditional passive monitoring to proactive adjustment, achieving more refined production management goals. It improves prediction accuracy by adjusting the roller spacing of iron ore sintering crushers, providing data support and scientific basis for fuel particle size control decisions in the metallurgical field. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the structural framework of the fuel particle size identification and control system of the present invention;

[0037] Figure 2 This is a schematic diagram of the image processing flow in an embodiment of the fuel particle size identification and control method of the present invention;

[0038] Figure 3These are comparative schematic diagrams of the fuel particle size identification and control method embodiment of the present invention after algorithm processing, (a) is the original image, (b) is the algorithm-identified image, and (c) is the mask image;

[0039] Figure 4 This is a schematic diagram of the process for adjusting the roller spacing of a crusher in an embodiment of the fuel particle size identification and control method of the present invention. Detailed Implementation

[0040] To better understand the above-mentioned technical solutions of the present invention, the technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0041] Combination Figure 1 As shown, the present invention provides a fuel particle size identification and control system for iron ore sintering, comprising:

[0042] Image acquisition unit 1 is located above the end of belt 3 at the rear position of crusher 2, and is used to acquire image data of the crushed finished fuel.

[0043] Server 4 is used to acquire image data from image acquisition unit 1.

[0044] The industrial control computer 5 is used to obtain the data transmitted from the server 4, calculate the predicted result of the roller spacing of the crusher 2, and issue an instruction to adjust the roller spacing of the crusher 2.

[0045] The image acquisition unit 1 uses a camera or industrial camera. The installation position should cover the entire width of the belt 3. The height of the image acquisition unit 1 should be adjusted according to the particle size of the finished fuel and the width of the belt 3 to ensure that the image data of the crushed finished fuel can be clearly captured.

[0046] This invention also provides a method for identifying and controlling the particle size of fuel in iron ore sintering. The method uses the fuel particle size identification and control system of this invention. Based on the image data detection results of the crushed finished fuel obtained by the image acquisition unit 1 and the working status of the equipment at the production site, the industrial control computer 5 calculates the predicted result of the roller spacing of the crusher 2 and issues an instruction to adjust the roller spacing of the crusher 2. This adjusts the roller spacing between the upper and lower rollers of the crusher 2 during the production process, thereby achieving closed-loop control of particle size in the sintering fuel preparation stage.

[0047] The fuel particle size identification and control method of the present invention includes the following steps:

[0048] S1, Image data of the crushed finished fuel is acquired using image acquisition unit 1;

[0049] S2, the image data is processed using ResNet model, XGBoost model and Bayesian optimization method to complete the fuel particle size image data processing;

[0050] S3, based on the MatchPattern method, a training dataset is constructed and applied to the PredictKNN algorithm. By intelligently averaging the target output data of the data points, the prediction result of the roller spacing of the crusher is obtained.

[0051] S4. Based on the image acquisition unit 1 monitoring the particle size of the finished fuel, the industrial control computer automatically adjusts the roller spacing of the crusher through a feedback mechanism, and adaptively adjusts the fuel particle size identification and control system when necessary, forming a closed-loop control system to ensure the stability of the fuel particle size.

[0052] In step S1 above, image data of the crushed finished fuel is acquired using image acquisition unit 1, specifically including the following steps:

[0053] S11, The image acquisition unit is positioned above the end of the belt at the rear of the crusher to ensure that image data of the crushed finished fuel can be acquired accurately and stably.

[0054] S12, the image acquisition unit's shooting angle covers the entire width of the belt. On-site personnel adjust the height of the image acquisition unit according to the size of the fuel particles and the width of the belt to ensure that the particle images can be clearly captured.

[0055] In step S2 above, the image data is processed using ResNet, XGBoost, and Bayesian optimization methods to complete the fuel particle size image data processing. Specifically, this includes the following steps:

[0056] S21, replace the output layer of the ResNet model with an output layer suitable for fuel particle size classification;

[0057] S22 utilizes the XGBoost model to improve the gradient boosting algorithm. When solving for the extreme value of the loss function, Newton's method is used, and the loss function is expanded to the second order using Taylor expansion. In addition, a regularization term is added to the loss function. During training, the objective function consists of two parts: the first part is the gradient boosting algorithm loss, and the second part is the regularization term.

[0058] The loss function is defined as follows:

[0059]

[0060] Where n is the number of training function samples; 1 represents the loss on a single sample, assumed to be a convex function y. i ' represents the model's prediction of the training samples, y' is the prediction value of the training samples. i The true label values ​​of the training samples;

[0061] It is a regularization term used to control model complexity, prevent overfitting, and improve the model's generalization ability. k The form changes as the form of the regularization term changes.

[0062] For example, when using L2 regularization (Ridge), in Ridge regression, f k It is the weight associated with the k-th feature.

[0063] The regularization term defines the complexity of the model:

[0064]

[0065] Where γ and λ are manually set parameters; ω is the vector formed by the values ​​of all leaf nodes in the decision tree; and T is the number of leaf nodes.

[0066] S23. The fuel particle size identification model constructed using the ResNet and XGBoost models is tuned based on the Bayesian optimization method; a ResNet model is constructed for feature extraction and input into the XGBoost model for classification; hyperparameter tuning is performed using BayesSearchCV and the Bayesian optimization method is executed; after tuning based on the Bayesian optimization method, the optimal hyperparameter combination is found and the performance of the optimized model is evaluated using a test set.

[0067] In step S3 above, a training dataset is constructed based on the MatchPattern method and applied to the PredictKNN algorithm. By intelligently averaging the target output data of the data points, the predicted result of the crusher's roller spacing is obtained. Specifically, this includes the following steps:

[0068] S31, based on the training dataset, calculates the Euclidean distance between the target data and the input data using the EuclideanDistance method, and stores the distance and its corresponding output data using the Neighbor class;

[0069] S32, sort the Euclidean distances between the target data and the input data, and find the three historical data points that are closest to the target data;

[0070] S33 uses intelligent averaging of the target output data from these three historical data points to obtain a predicted result for the roller spacing of the crusher.

[0071] Example

[0072] See again Figure 1As shown, this embodiment provides a fuel particle size recognition and control system for iron ore sintering. The image acquisition unit 1 is installed above the end of the belt 3 at the downstream position of the four-roll crusher 2 to ensure accurate and stable acquisition of image data of the crushed finished fuel. The installation position of the image acquisition unit 1 should cover the entire width of the belt 3, and the height of the camera or industrial camera should be adjusted according to the size of the fuel particles and the width of the belt 3 to ensure clear capture of fuel particle images.

[0073] When image acquisition begins, the image acquisition unit will automatically start if the fuel crusher system is powered on. Images are acquired via a camera or industrial camera. The images are then preprocessed, segmented, processed for different regions, and feature extracted. Fuel particle size data is then stored in a database (e.g.,...). Figure 2 The image processing flow shown can be used to obtain more suitable particle size image data.

[0074] In step 2, which replaces the output layer of the ResNet model with an output layer suitable for particle size classification, XGBoost is an improvement on the gradient boosting algorithm. It uses Newton's method to solve the extreme value of the loss function, expands the loss function to the second order using Taylor expansion, and adds a regularization term to the loss function.

[0075] In step 2 of this embodiment, the objective function during training consists of two parts: the first part is the gradient boosting algorithm loss, and the second part is the regularization term. The loss function is defined as:

[0076]

[0077] Where n is the number of training function samples; 1 represents the loss on a single sample, assumed to be a convex function y. i ' represents the model's prediction of the training samples, y' is the prediction value of the training samples. i The true label values ​​of the training samples;

[0078] The regularization term defines the complexity of the model:

[0079]

[0080] Where γ and λ are manually set parameters; ω is the vector formed by the values ​​of all leaf nodes in the decision tree; and T is the number of leaf nodes.

[0081] In addition, ensure that the fuel particle size image data is ready and divided into training, validation, and test sets. The data needs preprocessing to meet the input requirements of ResNet and XGBoost. Construct a ResNet model for feature extraction and input the extracted features into an XGBoost model for classification. Determine the hyperparameters to be optimized and define their search space. For ResNet, this includes the learning rate and batch size; for XGBoost, it includes the learning rate, tree depth, and subsampling rate. Create a pipeline to combine feature extraction and the XGBoost model, using BayesSearchCV for hyperparameter tuning, performing Bayesian optimization to find the optimal hyperparameter combination. Evaluate the performance of the optimized model using the test set, comparing the particle size identification algorithm to... Figure 3 As shown in Table 1, the performance data of the algorithm for particle size identification is as follows:

[0082] Table 1. Fuel Particle Size Identification Data

[0083] <0.5mm 0.5~3mm 3-5mm >5mm total 94 83 65 23 265

[0084] The above is an exemplary description of a process for acquiring and processing fuel particle size images in iron ore sintering. The following will further combine... Figure 1 and Figure 4 In this embodiment, a high-quality training dataset is constructed using the MatchPattern method based on fuel particle size image data processing. This high-quality training dataset is then applied to the PredictKNN algorithm. By intelligently averaging the target output data of the data points, the predicted result of the crusher roller spacing is obtained and the roller spacing is adjusted.

[0085] In step 3 of this embodiment, a high-quality training dataset is constructed based on the MatchPattern method and applied to the PredictKNN algorithm. By intelligently averaging the target output data of the data points, the predicted result of the crusher roller spacing is obtained and the roller spacing is adjusted.

[0086] Furthermore, in practical applications, the system typically involves five pelletizing crushers, but not all crushers operate in real time; usually only one or two are running, with a maximum of five operating simultaneously. Based on this, this invention innovatively extracts machine operating status codes through meticulous preprocessing of the latest data entries, defining unique operating status patterns accordingly. This provides strong support for subsequent data filtering and pattern matching. The scheme aims to deeply mine and utilize particle size data from historical databases to reveal the complex relationship between the proportion of different coke powder particle sizes and the roller spacing of the four-roll crusher. Through a carefully designed mapper layer, this invention achieves efficient data extraction, laying a solid foundation for subsequent analysis.

[0087] First, the EuclideanDistance method is used to calculate the Euclidean distance between the target data and the input data, and these distances and their corresponding output data are stored in a Neighbor class. Next, the Euclidean distances between the target data and the input data are sorted, and the three historical data points closest to the target data are identified. By intelligently averaging the target output data from these three data points, the predicted crusher roller spacing is obtained. This method fully leverages the capabilities of the KNN algorithm, achieving accurate prediction and analysis of the target data (crusher roller spacing) based on the complex relationship between particle size distribution and crusher spacing in historical data.

[0088] In step 4 of this embodiment, the system monitors the fuel particle size in real time, automatically adjusts the roller spacing through a feedback mechanism, and adaptively adjusts the system when necessary, forming a closed-loop control system to ensure the stability of the fuel particle size. Specifically, this includes adjusting the crusher roller spacing based on the target particle size distribution and the predicted results obtained after processing the particle size image. Figure 4 As shown.

[0089] By constructing an integrated closed-loop control system, fuel particle size can be monitored in real time, and the identification results can be directly fed back to the roller spacing adjustment module. Through continuous data flow and feedback mechanisms, this system achieves automatic adjustment of the roller spacing, ensuring that the fuel particle size fluctuates stably within the set target range.

[0090] As the core sensing layer of the control system, the fuel particle size identification system uses advanced image processing and machine learning algorithms to capture and analyze the distribution of fuel particles after passing through the four-roll crusher in real time.

[0091] The high-precision particle size data provided by the identification system not only serves as an information source for process monitoring, but also as an important reference for adjusting the roller gap.

[0092] Furthermore, through closed-loop feedback, the system can automatically compare and analyze the identified particle size distribution with the preset ideal particle size distribution. When a particle size deviation is detected, it immediately triggers an adjustment command for the roller spacing. This closed-loop system has an adaptive adjustment function, which can dynamically optimize control parameters based on actual production conditions and historical data. This adaptive capability ensures that the system can maintain efficient operation under different working conditions, avoiding particle size instability caused by raw material fluctuations or changes in equipment status, thereby significantly improving the accuracy and consistency of fuel particle size control.

[0093] This invention predicts the crusher roller spacing based on fuel particle size data identified by an algorithm. The particle size control interface performs routine management of the particle size data, obtaining a data map of particle size distribution and the actual operating status of the crusher rollers. The parameter setting interface allows for manual or automatic control of fuel particle size parameters. The target particle size distribution setting interface allows for step-by-step setting of the target fuel particle size required in actual production. The crusher parameter setting interface adjusts different crushers according to the crusher roller spacing required for actual production, thereby achieving the purpose of controlling fuel particle size.

[0094] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0095] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely software embodiment or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD ROM, optical storage, etc.) containing computer-usable program code.

[0096] Those skilled in the art should recognize that the above embodiments are merely illustrative of the present invention and are not intended to limit the present invention. Any variations or modifications to the above embodiments that are within the spirit and essence of the present invention will fall within the scope of the claims of the present invention.

Claims

1. A fuel particle size identification and control system for iron ore sintering, characterized in that, include: The image acquisition unit is located above the end of the belt at the rear of the crusher and is used to acquire image data of the crushed finished fuel. A server is used to acquire image data from the image acquisition unit. An industrial control computer is used to acquire the data transmitted from the server, calculate the predicted result of the roller spacing of the crusher, and issue an instruction to adjust the roller spacing of the crusher.

2. The fuel particle size identification and control system for iron ore sintering according to claim 1, characterized in that: The image acquisition unit is a camera or an industrial camera.

3. A method for identifying and controlling fuel particle size in iron ore sintering, characterized in that: The fuel particle size identification and control system for iron ore sintering as described in claim 1 or 2 uses the image data detection results of the crushed finished fuel obtained by the image acquisition unit and the predicted result of the roller spacing of the crusher based on the working status of the equipment at the production site. The system then issues an instruction to adjust the roller spacing of the crusher, thereby adjusting the roller spacing between the upper and lower rollers of the crusher during the production process and realizing closed-loop control of particle size in the sintering fuel preparation stage.

4. The method for identifying and controlling fuel particle size in iron ore sintering according to claim 3, characterized in that, The fuel particle size identification and control method Includes the following steps: S1, use the image acquisition unit to acquire image data of the crushed finished fuel; S2, perform fuel particle size image data processing on the image data; S3. A training dataset is constructed based on the MatchPattern method and applied to the PredictKNN algorithm. By averaging the target output data of the data points, the prediction result of the roller spacing of the crusher is obtained. S4, based on the image acquisition unit monitoring the particle size of the finished fuel, the industrial control computer automatically adjusts the roller spacing of the crusher through a feedback mechanism to ensure the stability of the fuel particle size.

5. The method for identifying and controlling fuel particle size in iron ore sintering according to claim 4, characterized in that, Step S1 specifically includes the following steps: S11, The image acquisition unit is positioned above the end of the belt at the rear end of the crusher; S12, the image acquisition unit's shooting angle covers the entire width of the belt, and the height of the image acquisition unit is adjusted according to the size of the fuel particles and the width of the belt.

6. The method for identifying and controlling fuel particle size in iron ore sintering according to claim 4, characterized in that, In step S2, the processing of fuel particle size image data using the ResNet model, XGBoost model, and Bayesian optimization method specifically includes the following steps: S21, replace the output layer of the ResNet model with an output layer suitable for fuel particle size classification; S22 utilizes the XGBoost model to improve the gradient boosting algorithm. When solving for the extreme value of the loss function, Newton's method is used, and the loss function is expanded to the second order using Taylor expansion. In addition, a regularization term is added to the loss function. During training, the objective function consists of two parts: the first part is the gradient boosting algorithm loss, and the second part is the regularization term. S23. The fuel particle size identification model constructed using the ResNet and XGBoost models is tuned based on the Bayesian optimization method; a ResNet model is constructed for feature extraction and input into the XGBoost model for classification; hyperparameter tuning is performed using BayesSearchCV and the Bayesian optimization method is executed; after tuning based on the Bayesian optimization method, the optimal hyperparameter combination is found and the performance of the optimized model is evaluated using a test set.

7. The method for fuel particle size identification and control in iron ore sintering according to claim 6, characterized in that, In step S22, the loss function is defined as follows: Where n is the number of training function samples; 1 represents the loss for a single sample, assumed to be a convex function y. i ' represents the model's prediction of the training samples, y' is the prediction value of the training samples. i The true label values ​​of the training samples; The regularization term defines the complexity of the model: Where γ and λ are manually set parameters; ω is the vector formed by the values ​​of all leaf nodes in the decision tree; and T is the number of leaf nodes.

8. The method for identifying and controlling fuel particle size in iron ore sintering according to claim 4, characterized in that, Step S3 specifically includes the following steps: S31, based on the training dataset, calculates the Euclidean distance between the target data and the input data using the EuclideanDistance method, and stores the distance and its corresponding output data using the Neighbor class; S32, sort the Euclidean distances between the target data and the input data, and find the three historical data points that are closest to the target data; S33, by averaging the target output data of these three historical data points, the predicted result of the roller spacing of the crusher is obtained.

Images