Method and system for controlling carbon monoxide concentration in sintering flue gas based on dust ball

Abstract

This invention provides a method and system for controlling carbon monoxide concentration in sintering flue gas based on the preparation of dust-collecting ash pellets, relating to the field of metallurgical solid waste treatment technology. The method includes: S1, obtaining raw materials for dust-collecting ash pellet preparation and mixing the raw materials; S2, using an improved visual recognition model to perform image segmentation and particle size calculation of the dust-collecting ash pellets to obtain the dust-collecting ash pellets; S3, using a combined edge contour and Hough transform detection method to calculate the particle size of the dust-collecting ash pellets; S4, controlling the carbon monoxide concentration during the sintering process through the dust-collecting ash pellets. This invention, through the identification of dust-collecting ash pellet particle size and standardized control of the pelletizing process, ensures that the pelletizing particle size is concentrated within the target range; simultaneously, it establishes the relationship between pellet particle size and carbon monoxide concentration, clarifying the influence of particle size on carbon monoxide emissions; and through the synergistic regulation of the dust-collecting ash ratio and pellet particle size on carbon monoxide concentration, it solves the problem of dust-collecting ash accumulation pollution in the steel manufacturing process.

Description

Technical Field

[0001] This invention relates to the field of metallurgical solid waste treatment technology, specifically to a method and system for controlling carbon monoxide concentration in sintering flue gas based on dust removal ash pellets. Background Technology

[0002] Steel production generates a large amount of metallurgical dust, which, if not properly handled, can lead to various pollution problems affecting soil, air, and water. During steel processing, the dust is typically collected and treated, and then added to sintered ore for solid waste utilization. Currently, relying on manual or basic image detection methods results in large errors in particle size identification and dispersed pellet size, making it difficult to explore the core mechanism by which particle size affects carbon monoxide (CO) concentration.

[0003] CN116287691A discloses a method and granulation reinforcing agent for improving the permeability of thick material layer sintering. This granulation reinforcing agent uses silicon dioxide (SiO2), alumina (Al2O3), calcium oxide (CaO), and sodium oxide (Na2O) as core components, with magnesium oxide (MgO) and potassium oxide (K2O) added as needed at a ratio of 0.1%-1% of the total sintering raw materials. Through one or two mixing and granulation processes, it can improve the permeability of the sintering layer and also replace some quicklime to reduce greenhouse gas emissions, meeting the requirements of thick and ultra-thick material layer sintering processes. The shortcomings of this process are: although it improves the permeability of the sintering layer through a specific component granulation reinforcing agent, it does not optimize for the characteristics of specific dust-containing materials such as dust collector ash, which has fine particles and strong hydrophobicity, nor does it address the control of carbon monoxide (CO) concentration caused by the addition of dust collector ash. It only focuses on improving the overall performance of the sintering layer and lacks targeted solutions for the specific problems caused by dust collector ash.

[0004] CN112553462A discloses a sintered ore containing sintered dust removal ash pellets and its preparation method. The method involves grinding sintered dust removal ash through a high-pressure roller mill, adding a binder composed of starch, syrup, and quicklime to form dust removal ash pellets, and then mixing them with blended ore, recycled ore, fuel, and flux. Sintering is performed while controlling parameters such as material layer thickness, ignition time, and ignition negative pressure. This method utilizes the dust removal ash as a resource and simultaneously improves the average particle size and yield of the sintered mixture. However, this process has shortcomings. While it solves the problem of granulation and recovery of dust removal ash through direct batching, it fails to address the carbon monoxide (CO) concentration issue inherent in the dust removal ash itself. It does not optimize the chemical composition of the dust removal ash pellets, such as adjusting the ratio of calcium oxide (CaO) to additives. It only focuses on the utilization of dust removal ash and the basic properties of the sintered ore, lacking specific control design for the crucial indicator of CO concentration.

[0005] Currently, in the field of dust collector ash pellet preparation, traditional particle size identification relies on manual sieving or simple image detection, which suffers from problems such as large errors, poor real-time performance, and weak adaptability. A large amount of data on the compositional characteristics of dust collector ash, the relationship between the ratio of calcium oxide (CaO) and additives and the air permeability and carbon monoxide (CO) concentration of pellets has not been effectively integrated and utilized. This fails to provide data support and theoretical guidance for addressing the CO concentration problem caused by dust collector ash addition, thus hindering the green and efficient development of dust collector ash pellet preparation processes. Therefore, this invention, based on experimental data from multiple sets of dust collector ash pellet preparation, embeds the YOLOv11 model structure with a multi-scale enhanced upsampling module (MEUM) and a local weighted attention mechanism (LIA) to perform dust collector ash pellet image segmentation. Morphological opening and bilateral filtering are then performed on the segmented mask, and particle size is calculated using a joint edge contour and Hough transform detection method. Establishing a correlation between the ratio of dust collector ash, calcium oxide (CaO), and binder and the pellet size is of great significance for optimizing the dust collector ash pellet preparation process, reducing carbon monoxide (CO) emissions, and enhancing the resource utilization value of metallurgical solid waste. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention aims to provide a method and system for controlling carbon monoxide (CO) concentration in sintering flue gas based on the preparation of dust collector ash pellets. By standardizing the identification of dust collector ash pellet particle size and the pelletizing process, the method identifies particle size online and regulates pelletizing to concentrate the particle size within a target range, producing standardized pellets. Simultaneously, it establishes a quantitative correlation between pellet particle size and CO concentration, clarifying the specific impact of particle size on CO emissions. This dual-factor synergistic regulation of the dust collector ash ratio and pellet particle size on CO concentration targets and reduces CO concentration during dust collector ash recycling, solving the problems of dust collector ash storage pollution and resource waste in steel enterprises, while eliminating the potential CO emission hazards associated with its recycling.

[0007] Specifically, on the one hand, the present invention provides a method for controlling the carbon monoxide concentration in sintering flue gas prepared from dust-removed ash pellets, which includes the following steps: S1: Obtain the mixed raw materials for preparing dust removal ash pellets; S2: The mixed raw materials obtained in step S1 are processed into pellets. A local weighted attention module and a multi-scale enhanced upsampling module are introduced into the visual recognition model. The local weighted attention module improves the network's sensitivity to pellet features. A gating mechanism is used to refine the calculation of the attention map. By inputting the overall feature map of the dust removal ash pellet image, the accuracy of feature extraction is improved. The multi-scale enhanced upsampling module optimizes the network's recognition performance on the dust removal ash pellet image, reduces and activates the pellet ore image features, obtains multi-scale pooling features at the particle size level of the dust removal ash pellets, extracts and enhances the edge features at the particle size level of the dust removal ash pellets, and fuses the multi-scale features of the dust removal ash pellets. The improved visual recognition model is used to segment the dust removal ash pellet image and calculate the particle size, identify the pellet particle size, obtain the pellet particle size distribution, and obtain the dust removal ash pellets. S3: The particle size of the dust collector ash pellets identified in step S2 is calculated using a joint detection method of edge contour and Hough transform; the dust collector ash pellets are segmented and binarized using a mask, and the contour area of ​​the dust collector ash pellets is selected to determine whether it meets the requirements. The conditions are the selected dust collector ash pellets; the Hough transform equation of the dust collector ash pellets is established to determine the Hough candidate circles of the dust collector ash pellets, and the relative difference parameter of the radius of the dust collector ash pellets is determined. Parameters of distance between the center of dust collection pellets Determine the crossover ratio and dynamic crossover ratio threshold of dust collector ash pellets; obtain the true value of dust collector ash pellet particle size by determining the matching parameters and joint matching conditions of dust collector ash pellets; S4: The dust removal ash pellets obtained in step S3 are subjected to cold solidification treatment. A thermo-kinematic model for controlling the particle size of the dust removal ash pellets to reduce the carbon monoxide concentration is established. The carbon monoxide concentration during the sintering process is controlled by the dust removal ash pellets obtained in step S3.

[0008] Preferably, step S2 specifically includes: S21: Use a pelletizing machine to prepare dust removal ash pellets; S22: Introduce a local weighted attention module into the visual recognition model to improve the sensitivity to sphere features in small targets or complex backgrounds and improve the accuracy of feature extraction; S23: In the visual recognition model, a multi-scale enhanced upsampling module is used to optimize the network's recognition performance on dust removal ash pellet images, reduce the dimensionality and activate the pellet image features, obtain multi-scale pooling features at the particle size level of dust removal ash pellets, extract and enhance the edge features at the particle size level of dust removal ash pellets, and fuse the multi-scale features of dust removal ash pellets.

[0009] Preferably, step S22 specifically includes: ; ; in, For dust ash pellet image pixels Local weight parameters; The image pixels represent dust clumps. To remove dust ash pellets image pixels The neighborhood centered on; Weighting parameters used to adjust weighted measurements; For dust ash pellet image pixels The surrounding sub-region; natural constant of Power; For the pixel index of the dust ash pellet image; It belongs to the symbol; For the first Each dust ash pellet image pixel; This is a feature map of the overall characteristics of the dust ash pellet image; Index for sub-regions; This refers to the multi-scale hierarchical sequence number corresponding to the particle size of the pellets; For attention feature maps; For element-wise dot product; For bilinear interpolation used for rescaling; This is the first channel of the input feature map; This is an activation operation for the characteristics of pellet ore.

[0010] Preferably, the multi-scale pooling feature of the dust removal ash pellet particle size hierarchy obtained in step S23 specifically includes: ; ; in, For the first Pooling characteristics of dust collection ash pellets at different particle size levels; This is a 3×3 average pooling operation; For the first -1 intermediate feature map of the particle size hierarchy of dust collection ash pellets; This is a characteristic of intermediate pellet ore; It exhibits characteristics of pelletized mineralization; For the first A refined pooling characteristic diagram of the particle size hierarchy of dust removal ash pellets; Standardized batch operations for pellet ore characteristics; This is a 1×1 convolution operation on the feature map of pellet ore.

[0011] Preferably, the extraction of edge features at the particle size level of enhanced dust removal ash pellets in step S23 specifically involves: ; ; in, For the first Edge feature map of the particle size hierarchy of dust removal ash pellets; For the first Enhanced feature map of the particle size hierarchy of dust removal ash pellets.

[0012] Preferably, step S3 specifically includes: S31: Perform segmentation masking binarization on the dust removal ash pellets, and filter out the obtained dust removal ash pellets; determine the minimum circumcircle of the dust removal ash pellet outline, establish the Hough transform equation and Hough candidate circles for the dust removal ash pellets, determine the relative difference parameter of the radius of the dust removal ash pellets, and determine the center distance parameter of the dust removal ash pellets. S32: Determine the crossover ratio of dust removal ash pellets and the dynamic crossover ratio threshold of dust removal ash pellets. ,in, For dynamic intersection-union ratio threshold; Based on the intersection-union ratio threshold; This is the adjustment coefficient; S33: Determine the matching parameters and joint matching conditions of dust collector ash pellets, and use the Hough transform of dust collector ash pellets that meet the above conditions as matching candidates to obtain the true value of the particle size of dust collector ash pellets.

[0013] Preferably, the matching parameters for dust collection ash pellets in step S33 are as follows: ; , ; in, For matching parameters; This is the radius difference penalty coefficient; The intersection-union ratio of the two circular regions; The outline is a circular area; This is the region of Hough transform; It is a function of the area of ​​the region; The symbol for intersection; This is a merge symbol; For the first The outline circle and the first The relative difference parameter of the radius between the Hough transforms; The radius of the outline circle; The Hough transform radius of the dust collection pellets.

[0014] Preferably, the joint matching conditions for dust collection ash pellets in step S33 are as follows: ; in, The radius difference threshold; This is the control coefficient for the distance from the center of the circle; The distance parameter between the center of the dust collection ash pellet and the center of the Hough transform is the distance parameter between the outline circle and the center of the dust collection ash pellet. This is the dynamic crossover ratio threshold.

[0015] Preferably, step S4 specifically includes: S41: Cold solidification treatment of dust removal ash pellets: The mixed raw materials obtained in step S1 and the dust removal ash pellets of various particle sizes after cold solidification treatment are respectively subjected to high-temperature calcination. S42: During the sintering process of dust removal ash pellets, the pellet size is adjusted by changing the air permeability of the bed and the oxygen diffusion resistance inside the pellets, thereby regulating the ratio of complete combustion to incomplete combustion of carbon and controlling the carbon monoxide concentration.

[0016] On the other hand, the present invention provides a sintering flue gas carbon monoxide concentration control system based on a method for controlling carbon monoxide concentration in sintering flue gas prepared from dust-removing ash pellets, which includes: a dust-removing ash pellet preparation module, a dust-removing ash pellet image recognition module, a dust-removing ash pellet particle size recognition module, and a carbon monoxide concentration control module. The dust collector ash pellet preparation module is used to obtain raw materials for dust collector ash pellet preparation, process the raw materials for dust collector ash pellet preparation, and mix the raw materials for dust collector ash pellet preparation. The dust removal ash pellet image recognition module introduces a local weighted attention module and a multi-scale enhanced upsampling module into the visual recognition model; the improved visual recognition model is used to segment the dust removal ash pellet image and calculate the particle size, identify the pellet particle size, obtain the pellet particle size distribution, and obtain the dust removal ash pellet. The dust collector ash pellet size identification module uses a joint detection method of edge contour and Hough transform to calculate the particle size of the dust collector ash pellets from the particle size feature map; it performs segmentation masking binarization on the dust collector ash pellets, filters the contour area of ​​the dust collector ash pellets, and determines the dust collector ash pellets that meet the conditions; it establishes the Hough transform equation for the dust collector ash pellets, determines the Hough candidate circles for the dust collector ash pellets, and determines the relative difference parameters of the dust collector ash pellet radius and the distance parameters between the dust collector ash pellet circles; it determines the crossover ratio and dynamic crossover ratio threshold of the dust collector ash pellets; and it obtains the true value of the dust collector ash pellet particle size by using the dust collector ash pellet matching parameters and joint matching conditions. The carbon monoxide concentration control module performs cold solidification treatment on the dust removal ash pellets, establishes a thermo-kinematic model for adjusting the particle size of the dust removal ash pellets to reduce the carbon monoxide concentration, and controls the carbon monoxide concentration during the sintering process.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention achieves standardized control of dust removal ash pellet particle size identification and pelletizing process, online identification of particle size, and regulation of pelletizing to concentrate particle size in the target range, producing standardized pellets. This makes particle size a controllable variable in the calcination experiment, eliminates particle size dispersion interference, enables comparison of experimental data, establishes a quantitative correlation between pellet particle size and carbon monoxide (CO) concentration, and clarifies the specific influence of particle size on CO emissions.

[0018] (2) This invention addresses the carbon monoxide (CO) emission problem caused by carbon oxidation of the dust itself by controlling the ratio of dust ash and pellet size through two-factor synergistic regulation. It establishes a quantitative correlation between the ratio and particle size. By combining the dust ash with calcium oxide (CaO) and binder, and combining the high-temperature calcination experiment with 1-8mm graded particles, the correspondence between the two factors and the CO concentration is clarified, thereby reducing the CO emission of the dust ash when it is reused.

[0019] (3) This invention utilizes metallurgical dust ash with a high proportion of resource utilization, and uses 93%~100% high proportion of dust ash as the core raw material to prepare pellets, thereby maximizing the reduction and recycling of metallurgical solid waste. This solves the problem of pollution and resource waste caused by the stockpiling of dust ash in steel enterprises, and at the same time eliminates the carbon monoxide (CO) emission hazards caused by its reuse, so that the high-value utilization of solid waste and the reduction of flue gas emissions can form a virtuous cycle. Attached Figure Description

[0020] Figure 1 This is a flowchart of the method for controlling carbon monoxide concentration in sintering flue gas based on dust removal ash pellets according to the present invention. Figure 2 This is a schematic diagram of the layered structure dust removal ash pellets of the present invention; Figure 3 This is a flowchart illustrating the effect of the dust removal ash pellet preparation process of this invention on carbon monoxide concentration; Figure 4 This is a structural diagram of the initial YOLOv11 model of this invention; Figure 5 The diagram shows the structure of the YOLOv11 model after introducing the local weighted attention module and the multi-scale enhanced upsampling module in this invention. Figure 6 The present invention uses a combined edge contour and Hough transform detection method to obtain a physical image of the dust collection ash pellets; Figure 7 This is a schematic diagram of the high-temperature hot furnace flue gas analysis reaction device in an embodiment of the present invention; Figure 8 This is a graph showing the change in carbon monoxide (CO) concentration under different mixing ratios in an embodiment of the present invention. Figure 9 This is a graph showing the change in carbon monoxide (CO) concentration of 1-3 mm pellets under different formulation conditions in an embodiment of the present invention. Figure 10 This is a graph showing the change in carbon monoxide (CO) concentration of 3-5mm pellets under different formulation conditions in an embodiment of the present invention. Figure 11 This is a graph showing the change in carbon monoxide (CO) concentration in 5-8mm pellets under different formulation conditions in an embodiment of the present invention. Detailed Implementation

[0021] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0022] This invention proposes a method for controlling the carbon monoxide concentration in sintering flue gas based on the preparation of dust-removed ash pellets, such as... Figure 1 As shown, the raw materials for preparing dust removal ash pellets are obtained and mixed; an improved visual recognition model is used for image segmentation and particle size calculation of the dust removal ash pellets to obtain the dust removal ash pellets; the particle size of the dust removal ash pellets is calculated using a joint detection method of edge contour and Hough transform; the carbon monoxide concentration during the sintering process is controlled by the dust removal ash pellets; the specific steps include: Step S1: Obtain the raw materials for preparing dust removal ash pellets, and perform raw material processing and mixing.

[0023] Step S11: Processing raw materials for preparing metallurgical dust pellets; The metallurgical dust is sieved, and the undersized dust is collected for later use. Specifically, a 100-mesh standard sieve is used to remove dust pellet particles larger than 100 mesh. The metallurgical dust is used as the processing target, and a 100-mesh standard sieve is used to sieve it to remove coarse particles larger than 100 mesh from the sieve. Only the dust with uniform particle size is collected and used for later use, ensuring a uniform particle size distribution in the raw materials for subsequent pellet preparation. The composition of the dust is shown in Table 1. Fe is the most abundant valuable metal element and has resource utilization value. Calcium (Ca) exists in calcium carbonate (CaCO3) and calcium sulfate (CaSO4) compounds.

[0024] Table 1. Mass fraction of chemical composition of metallurgical dust. The adhesive components are shown in Table 2.

[0025] Table 2. Mass fraction of chemical composition of adhesive Step S12: Mixing dust collector ash pellets to prepare raw materials; mixing dust collector ash, calcium oxide (CaO) and binder according to a preset ratio; in this embodiment of the invention, the preset ratio composition is: dust collector ash 93%~100%, calcium oxide (CaO) 0%~5%, binder 0%~5%.

[0026] Step S2: Obtain the mixed raw materials obtained in Step S1 for pellet preparation. Use an improved visual recognition model to segment the dust removal ash pellet images and calculate the particle size, identify the pellet particle size, obtain the pellet particle size distribution, and obtain the dust removal ash pellets. The visual recognition model in this embodiment of the invention is the YOLOv11 model.

[0027] Step S21: The preparation of dust collector ash pellets used in this embodiment of the invention includes three methods; Method 1: The raw material, after being thoroughly mixed with dust collector ash, calcium oxide (CaO), and binder in a preset ratio, is added to a disc pelletizer for pelletizing. Method 2: 50% of the dust collector ash in a preset ratio is first added to the pelletizer for pelletizing. After the initial pellet core is formed, the remaining 50% of the dust collector ash, all of the calcium oxide (CaO), and binder are mixed evenly in a preset ratio and then added to the pelletizer at a feeding rate to form layered pellets. Method 3: 50% of the dust collector ash in a preset ratio, all of the calcium oxide (CaO), and binder are mixed evenly in a preset ratio and then added to the pelletizer for pelletizing. After the initial pellet core is formed, the remaining 50% of the dust collector ash is added to the pelletizer at a feeding rate to form layered pellets. Figure 2 As shown, dust collector pellets are prepared using a disc pelletizer with a raw material addition rate of 150g every 30 seconds, followed by two additions of 50g every 30 seconds; Figure 3 The diagram shows the effect of the dust collector ash pellet preparation process on carbon monoxide concentration. The process uses dust collector ash as raw material, combined with calcium oxide (CaO) and binder to complete the batching. After particle size identification and control, pelletizing, cold solidification, and sieving, the pellets are calcined. After calcination, the CO concentration is recorded and the data is processed to verify the effect, thereby evaluating the overall operational effectiveness of the production process in a closed loop.

[0028] Step S22: During the pellet preparation process, images of dust-laden pellets are captured using a camera device, and the pellet size is identified using an improved YOLOv11 model; for example... Figure 5 The diagram shows the structure of the YOLOv11 model after introducing the Local Weighted Attention Module (LIA) and the Multi-Scale Enhanced Upsampling Module (MEUM).

[0029] Images of dust-collecting ash pellets were captured using camera equipment, ensuring the data accurately reflects actual industrial production conditions. During data acquisition, varying lighting conditions and background complexity were considered to enhance data diversity and robustness. This avoids particle size identification errors caused by insufficient scene adaptation, preventing uncontrolled particle size distribution and mitigating the increased CO generation during sintering due to uneven reaction caused by particle size dispersion. Data augmentation methods such as geometric transformation, Gaussian noise addition, contrast adjustment, dynamic blurring, and brightness variation were employed to expand the original dataset and improve the model's scene adaptability. The dataset was divided into training, testing, and validation sets in a 6:2:2 ratio to maintain the rigor of training and evaluation. The dust-collecting ash pellet images were labeled, ensuring consistent labeling. This allows the YOLOv11 model to learn more comprehensive dust-collecting ash pellet features, avoiding misjudgments of target particle size ranges due to identification errors. This ensures that subsequent pellet particle sizes are concentrated within the optimal reaction range, providing data support for complete combustion and CO reduction. The initial YOLOv11 model is a one-stage instance segmentation model, and its structure is as follows: Figure 4 As shown, Conv represents the convolution module, used for local feature extraction, channel transformation, and nonlinear representation of dust ash clump images. C3k2 represents the feature aggregation module, used to enhance the network's representation of clump features at different scales and reduce computational complexity while maintaining feature extraction performance. SPPF represents the fast spatial pyramid pooling module, used to expand the network's receptive field, enabling the model to acquire richer multi-scale contextual information. C2PSA represents the partial self-attention feature enhancement module, used to improve the network's attention to key clump regions and target features in complex backgrounds. Upsample represents the upsampling operation, used to improve the spatial resolution of high-level feature maps, enabling them to be fused with shallow detail features. Concat represents the feature concatenation operation, used to connect features at different levels, fusing shallow edge detail information with deep semantic information. Segment represents the instance segmentation output head, which is used to output the segmentation result of the pellet target, that is, to obtain the segmentation mask corresponding to the pellet region. Overall, the initial YOLOv11 network first extracts multi-scale features through the backbone network, then performs multi-scale feature fusion through the neck structure, and finally outputs the pellet instance segmentation result by the segmentation head, providing a basis for subsequent dust removal ash pellet particle size calculation.

[0030] Traditional self-attention mechanisms based on global computation are prone to losing details in dust agglomerate image segmentation tasks. If small agglomerates are missed, they may over-aggregate and fill the gaps between agglomerates, causing oxygen to be difficult to penetrate, resulting in local hypoxia and the generation of carbon monoxide (CO). To address this problem, this invention introduces a Locally Weighted Attention (LIA) module to improve the sensitivity to agglomerate features in small targets or complex backgrounds, thereby improving the accuracy of feature extraction. Specifically: Image pixels of dust-removed ash pellets obtained during the sintering process In the neighborhood The weight parameters in the code are: ; in, For dust ash pellet image pixels Local weight parameters; The image pixels represent dust clumps. To remove dust ash pellets image pixels The neighborhood centered on; Weighting parameters used to adjust weighted measurements; For dust ash pellet image pixels The surrounding sub-region; natural constant of Power; For the pixel index of the dust ash pellet image; It belongs to the symbol; For the first Each dust ash pellet image pixel; This is a feature map of the overall characteristics of the dust ash pellet image; Index for sub-regions; This refers to the multi-scale hierarchical number corresponding to the particle size of the pellets.

[0031] Unlike traditional complex subnetworks, the Locally Weighted Attention Module (LIA) refines the computation through soft pooling and 3×3 convolution. It employs a gating mechanism to finely compute the attention map, and this gating is derived from the first channel of the overall feature map of the input dust clump image. Specifically: ; in, For attention feature maps; For element-wise dot product; For bilinear interpolation used for rescaling; The first channel of the input feature map is used as the gating signal; This is an activation operation for the characteristics of pellet ore.

[0032] Step S23: Traditional upsampling methods lead to the loss of details in the image of dust-collected ash pellets, especially at the edges of the pellets, making it difficult to identify subtle differences between mineral particles. This defect causes particle size measurement deviation, resulting in uneven pellet permeability, unstable reaction interfaces, and ultimately increased carbon monoxide (CO) emissions. Therefore, the multi-scale enhanced upsampling module MEUM is introduced into the YOLOv11 model to replace the original upsampling module Upsample, optimizing the network's recognition performance for dust-collected ash pellet images. Its core execution process is as follows: Dimensionality reduction and activation of pellet ore image features, specifically: ; in, This is an intermediate feature map of the pellet ore image; Standardized batch operations for pellet ore characteristics; This is a 1×1 convolution operation on the feature map of pellet ore; The image pixels represent dust clumps.

[0033] The multi-scale pooling characteristics obtained at the particle size levels of dust collection ash pellets are as follows: ; in, For the first Pooling characteristics of dust collection ash pellets at different particle size levels; This is a 3×3 average pooling operation; For the first -1 intermediate feature map of the particle size hierarchy of dust collection ash pellets; This is a characteristic of intermediate pellet ore; It exhibits characteristics of pelletized mineralization.

[0034] The multi-scale pooling characteristics of dust collector ash agglomerate particle size are refined as follows: ; in, For the first A refined pooling characteristic diagram of the particle size hierarchy of dust removal ash pellets; This is a 1×1 convolution operation on the feature map of pellet ore.

[0035] Extracting edge features of dust collection ash pellets at different particle size levels, specifically: ; in, For the first Edge feature map of the particle size hierarchy of dust removal ash pellets.

[0036] Edge feature enhancement is performed at the particle size level of dust collection ash pellets, specifically as follows: ; in, For the first Enhanced feature map of the particle size hierarchy of dust removal ash pellets.

[0037] The multi-scale characteristics of dust agglomerate particle size are as follows: ; in, The final output is a particle size characteristic map of dust collection ash pellets; The activation function is ReLU.

[0038] By performing multi-scale feature fusion and edge detail preservation on the dust removal ash pellet images as described above, this optimization can capture the edge features of the pellets and avoid the dispersion of the particle size distribution of the dust removal ash pellets due to the particle size recognition deviation, thus making the pellets have uniform air permeability and stable reaction interface.

[0039] The evaluation metrics used are precision, recall, mean precision, mAP50-95, mean precision with IoU from 0.5 to 0.95, and mean precision with mAP50 and IoU of 0.5. The Locally Weighted Attention Module (LIA) can identify small-sized pellets, preventing small particle aggregation from affecting breathability, while ensuring that the characteristics of large-sized pellets are not ignored. This ensures that the pellet size distribution meets the requirements of sufficient breathability and complete reaction, providing a guarantee for reducing carbon monoxide (CO) emissions. This model structure is improved by the Multi-Scale Enhanced Upsampling Module (MEUM) and the Locally Weighted Attention Module (LIA).

[0040] Step S3: The particle size of the dust collector ash pellets identified in Step S2 is calculated using a combined edge contour and Hough transform detection method. The particle size calculation of the dust collector ash pellets directly affects the effectiveness of dust collector ash pellet size control. Only by mastering the particle size of the dust collector ash pellets can they be concentrated within the target range, avoiding uneven oxygen diffusion and combustion rate fluctuations caused by particle size deviations, and reducing the generation of carbon monoxide (CO), a product of incomplete combustion. Figure 6 The diagram illustrates the method for detecting dust clumps using a joint edge contour and Hough transform detection approach. This method first binarizes and filters the instance segmentation mask, then uses the minimum circumcircle of the contour to obtain a preliminary estimate of the dust clump size. Subsequently, multi-scale Hough transform detection is performed on the edge image, and the contour circle is matched with the Hough transform using radius difference, center distance, and intersection-union ratio (IU / UGAR) to ultimately determine the detection circle for the dust clumps. The specific steps include: Step S31: Perform segmentation masking binarization on the dust collection ash pellets, specifically as follows: ; in, To segment the mask image at pixels The grayscale value at that location; It is a binary image; This is the binarization threshold; To segment the mask image by the x-coordinate of each pixel; The vertical coordinates of the pixels in the segmentation mask image.

[0041] Filter the outline area of ​​dust collection ash pellets to determine if they meet the requirements. The condition is the dust collection ash pellets obtained through screening; among which, For the first A silhouette The area; This is the minimum area threshold used to eliminate noise or interfering profiles. The minimum circumcircle of the dust collection ash pellet profile is determined as follows: ,in, For the first A circular outline; The coordinates of the center of the outline circle; The radius of the contour circle.

[0042] Establish the Hough transform equation for dust collection ash pellets: ; in, The x-coordinate of the Hough transform center of the dust collection pellets; The ordinate of the Hough transform center of the dust collection pellets; Let be the radius of the circle.

[0043] The candidate circle for the Hough of dust collection pellets was determined as follows. ;in, For the first A candidate circle of dust removal ash pellets; The coordinates of the Hough transform center of the dust collection pellets; The Hough transform radius of the dust collection pellets.

[0044] Determine the relative difference parameters of the radius of dust collection ash pellets: ; in, For the first The outline circle and the first The relative difference parameter of the radius between the Hough transforms.

[0045] Determine the center-to-center distance parameter for dust collection pellets: ; in, The distance parameter between the center of the dust collection ash pellet and the center of the Hough transform is the distance parameter between the outline circle and the center of the dust collection ash pellet. The x-coordinate of the center of the outline circle; The x-coordinate of the Hough transform center of the dust collection pellets; The ordinate of the center of the outline circle; The ordinate of the Hough transform center of the dust collection pellets is given.

[0046] Step S32: Determine the crossover ratio of dust collection ash pellets as follows: ; in, The intersection-union ratio of the two circular regions; The outline is a circular area; This is the region of Hough transform; It is a function of the area of ​​the region; The symbol for intersection; This is a merge symbol.

[0047] The dynamic crossover ratio threshold for dust collection ash pellets is: ; in, For dynamic intersection-union ratio threshold; Based on the intersection-union ratio threshold; This is the adjustment coefficient.

[0048] Step S33: The greater the difference in the radius of the dust collection ash pellets, the higher the required crossover ratio threshold for matching the dust collection ash pellets; Dust collection ash pellet matching parameters: ; in, For matching parameters; This is the radius difference penalty coefficient.

[0049] Matching parameters The higher the value, the better the match between the contour circle and the Hough transform.

[0050] Dust collector ash pellets joint matching conditions: ; in, The radius difference threshold; This is the control coefficient for the distance from the center of the circle; Using the Hough transform of the dust agglomerates that meet the above conditions as matching candidates, the final detection result is as follows: ; in, For the first The final measured circle length of each pellet; The Hough transform length that best matches the contour circle; It is the length of the minimum circumcircle of the profile.

[0051] Finally, the true value of the dust collection ash pellet size was obtained. ,in This is a pre-set scale.

[0052] In summary, when a Hough transform simultaneously satisfies the radius difference, center distance, and dynamic intersection-union ratio constraints, it and the corresponding contour circle are considered as the same sphere. If there are multiple matching candidates for the same contour circle, the Hough transform with the highest score is selected as the final result. If there is no matching Hough transform, the minimum circumcircle of the contour is retained.

[0053] The above model was applied to the preparation of dust collector ash pellets. An alarm was triggered and the pelletizing machine was stopped when the proportion of pellets in the 1-3mm, 3-5mm, and 5-8mm sizes reached its highest values ​​during the three pelletizing stages, completing the pelletizing process. This ensured that the particle size of each batch of pellets was concentrated within a single target range, avoiding fluctuations in the sintering process caused by particle size dispersion. Particle size became a controllable variable, providing reliable data for establishing the correlation between pellet particle size and carbon monoxide (CO) concentration. Furthermore, data analysis was used to determine the optimal particle size range, allowing for targeted reduction of CO concentration.

[0054] Step S4: The dust removal ash pellets obtained in step S3 are subjected to cold solidification treatment. A thermo-kinematic model for controlling the particle size of the dust removal ash pellets to reduce the concentration of carbon monoxide (CO) is established. The concentration of CO is controlled by the dust removal ash pellets obtained in step S3.

[0055] Step S41: Cold solidification treatment is performed at room temperature for 3 days to allow natural evaporation of moisture and solidification reaction of additives between dust collector ash pellets, forming cold-solidified pellets with uniform structure and consistent reactivity. This improves the air permeability and reaction interface stability of pellets within the same particle size range, physically avoiding localized oxygen deficiency caused by structural inhomogeneity. The mixed raw materials obtained in Step S1, as well as the dust collector ash pellets of various particle sizes after cold solidification treatment, are then subjected to high-temperature calcination. The high-temperature calcination conditions are as follows: 1.5g of sample is added to a high-temperature furnace, and the temperature is raised to 1300℃. The dust collector ash pellets are 1-3mm, 3-5mm, and 5-8mm in size. The high-temperature furnace is heated steadily at a rate of 10℃ / min, and a blower connected to the mixing box blows air into the high-temperature furnace at a flow rate of 1.5L / min. The carbon monoxide (CO) concentration is recorded every 20 seconds using a flue gas analyzer.

[0056] Step S42: During the sintering of dust collector ash pellets, the formation of carbon monoxide (CO) is essentially an incomplete combustion reaction of carbon, and its concentration is mainly determined by the ratio of complete combustion to incomplete combustion. Pellet size affects the ratio of these two combustion reactions by altering bed permeability and oxygen diffusion resistance within the pellets, ultimately controlling the CO concentration. The oxidation of carbon in dust collector ash involves two parallel thermodynamic spontaneous reactions, the direction and extent of which are primarily regulated by oxygen concentration: The equation for the complete combustion of carbon is: .

[0057] The equation for the incomplete combustion of carbon is: .

[0058] Both complete and incomplete combustion of carbon are thermodynamically spontaneous reactions. Complete combustion of carbon is the more thermodynamically stable final state reaction and will dominate when oxygen is plentiful. When oxygen supply is insufficient, mass transfer limitations lead to an increase in the proportion of incomplete combustion reactions.

[0059] Pellet size and additive ratio jointly influence carbon monoxide (CO) formation behavior. Specifically, pellet size primarily affects the rate and concentration of oxygen reaching the carbon reaction interface through both overall bed mass transfer and internal pellet mass transfer. Calcium oxide (CaO) and binders further influence oxygen diffusion conditions and the degree of complete carbon oxidation by improving pellet structural stability and pore distribution. Together, they regulate the relative proportion of complete and incomplete oxidation reactions, thereby reducing the CO concentration. (1) Particle size effect on overall bed permeability. Bed permeability directly determines the overall diffusion rate of oxygen from the gas phase to the interior of the bed, while bed porosity is the core parameter affecting permeability. Within a reasonable particle size range, the pellet size and bed porosity are positively correlated: when there is no pelletizing dust, the particles are extremely fine, the bed porosity is low, and the permeability resistance is high. Oxygen is difficult to penetrate evenly into the interior of the bed, easily forming large-area local oxygen-deficient zones, carbon mainly undergoes incomplete combustion, and the amount of carbon monoxide (CO) generated increases; after fine dust particles agglomerate into larger dust collection ash pellets, the bed porosity increases, the permeability resistance decreases, oxygen can be more evenly distributed throughout the bed, oxygen supply conditions improve, the proportion of complete carbon combustion reaction increases, and the amount of carbon monoxide (CO) generated decreases. Increasing the pellet size within a suitable range can further optimize the bed porosity and permeability. However, when the particle size exceeds the reasonable critical range, the oxygen mass transfer resistance inside a single pellet will replace the bed permeability and become the main limiting factor in the reaction process.

[0060] (2) Particle size effect on mass transfer within the pellets. For a single pellet, the larger the particle size of the dust collector pellets, the longer it takes for oxygen to diffuse to the core area of ​​the pellet, the greater the internal oxygen diffusion resistance, the higher the risk of oxygen deficiency, and the easier it is for the proportion of incomplete oxidation reaction to increase. When the pellet particle size is small, the internal pore mass transfer resistance is small, oxygen can fully penetrate to the pellet core, the oxygen concentration inside the entire pellet is uniform and sufficient, and the amount of carbon monoxide (CO) generated is low; when the pellet particle size increases, oxygen can only penetrate to a certain depth inside the pellet, the oxygen concentration in the core area begins to decrease, the incomplete combustion reaction increases, and the amount of carbon monoxide (CO) generated increases; if the particle size continues to increase, oxygen will have difficulty penetrating to the pellet core, an oxygen-deficient environment will form in the core area, the incomplete combustion reaction will intensify, and the amount of carbon monoxide (CO) generated will further increase.

[0061] (3) Synergistic regulation mechanism of additives. Calcium oxide (CaO) and binder can form a continuous binder phase through high-temperature interaction, which improves the thermal stability of the pellets and avoids calcination breakage; at the same time, it reacts with silicon and aluminum components in the dust to generate a low-viscosity mineral phase, which homogenizes the pore distribution. Calcium oxide (CaO) can also react with adsorbed water to generate calcium hydroxide (Ca(OH)2), which releases water vapor through decomposition to form micropores and improve air permeability.

[0062] In this embodiment of the invention, the pelletizing machine was stopped when the proportions of 1-3mm, 3-5mm, and 5-8mm reached their highest values, respectively, thus completing the pelletizing process. The experimental analysis was carried out sequentially according to the proportions of the five raw materials in Table 3.

[0063] Table 3. Raw material ratio table for dust collector pellets Experiments 1, 2, 3, 4, and 5 were conducted using 100% dust collector ash as the raw material. The prepared dust collector ash pellets were cold-cured at room temperature for 3 days. 1.5g of raw material powder, 1-3mm, 3-5mm, and 5-8mm dust collector ash pellets were selected respectively, placed in crucibles, and then placed in a high-temperature furnace. The high-temperature furnace was connected to a mixing chamber and a flue gas analyzer. The mixing chamber was connected to a blower. A schematic diagram of the reaction apparatus is shown below. Figure 7 As shown, the high-temperature furnace parameters were set to increase by 10°C per minute, with a maximum temperature of 1300°C. The flow rate of the mixing chamber was set to 1.5 L / min. The flue gas analyzer was set to record data every 20 seconds. Then, the high-temperature furnace, flue gas analyzer, mixing chamber, and blower were turned on. The experiment was conducted, and the flue gas analyzer data was exported after the experiment.

[0064] All the obtained data were plotted with temperature on the horizontal axis and carbon monoxide concentration on the vertical axis.

[0065] like Figure 8The figure shows the carbon monoxide (CO) concentration variation curves of the unballed powder sample under different ratio conditions. It can be seen that the overall CO concentration of the unballed powder sample is relatively high, indicating that the bed permeability is poor when the dust removal ash powder is directly roasted, and oxygen is difficult to enter the interior of the powder accumulation layer evenly, which easily forms local oxygen-deficient areas, promotes incomplete combustion of carbon and produces a high CO concentration.

[0066] like Figure 9 The figure shows the carbon monoxide (CO) concentration variation curves of 1-3 mm pellet samples under different formulation conditions. Figure 8 In comparison, the carbon monoxide (CO) concentration in the 1-3 mm pellet samples was generally lower, indicating that after pelletizing, the powder particles changed from disordered stacking to a dust collector sludge pellet structure with a stable particle size. This resulted in more continuous airflow channels between the pellets, improving the uniformity of oxygen supply. Simultaneously, the shorter diffusion distance within the 1-3 mm pellets facilitated oxygen entry into the pellet interior and promoted complete carbon oxidation.

[0067] like Figure 10 The figure shows the carbon monoxide (CO) concentration variation curves of 3-5 mm pellets under different formulation conditions. Compared with unpelleted powder, 3-5 mm pellets still exhibit a lower CO concentration, indicating that this particle size range can improve bed permeability. However, compared with 1-3 mm pellets, as the particle size of dust collector pellets increases, the internal oxygen diffusion path lengthens, and some core areas of the pellets may experience insufficient oxygen supply, thus limiting the reduction in CO.

[0068] like Figure 11 The figure shows the carbon monoxide (CO) concentration variation curves of 5-8 mm pellets under different formulation conditions. Within this particle size range, the inter-pellet porosity increases further, which facilitates gas passage through the bed. However, the oxygen diffusion resistance within individual pellets also increases further. Therefore, the CO concentration variation reflects the proportional relationship between improved bed permeability and restricted diffusion within the pellets. This indicates that larger pellet size is not always better; rather, a suitable range needs to be selected between bed permeability and internal mass transfer resistance.

[0069] Comparison of different formulation curves shows that while changing the ratio of calcium oxide (CaO) or binder can affect the peak concentration and release process of carbon monoxide (CO) to some extent, the poor permeability of the powder bed remains the main factor limiting oxygen supply in the un-pelletized state. Therefore, the effect of additives on reducing CO concentration is limited by the powder packing structure. Further changes in CO concentration after adding calcium oxide (CaO) or binder to prepare pellets indicate that calcium oxide (CaO) and binder can further reduce the proportion of incomplete oxidation reactions by improving the stability of the pellet structure, pore distribution, and oxygen diffusion conditions.

[0070] The second aspect of this invention proposes a carbon monoxide concentration control system for sintering flue gas based on a method for controlling carbon monoxide concentration in sintering flue gas prepared from dust-removed ash pellets. The system includes: a dust-removed ash pellet preparation module, a dust-removed ash pellet image recognition module, a dust-removed ash pellet particle size recognition module, and a carbon monoxide concentration control module.

[0071] The dust collector ash pellet preparation module is used to obtain, process, and mix the raw materials for dust collector ash pellet preparation.

[0072] The dust removal ash pellet image recognition module introduces a local weighted attention module and a multi-scale enhanced upsampling module into the YOLOv11 model; the improved YOLOv11 model is used to segment the dust removal ash pellet image and calculate the particle size, identify the pellet particle size, obtain the pellet particle size distribution, and obtain the dust removal ash pellets.

[0073] The dust collector ash pellet size identification module uses a joint detection method of edge contour and Hough transform to calculate the particle size of the dust collector ash pellets from the particle size feature map; it performs segmentation masking binarization on the dust collector ash pellets, filters the contour area of ​​the dust collector ash pellets, and determines the selected dust collector ash pellets that meet the conditions; it establishes the dust collector ash pellet Hough transform equation, determines the dust collector ash pellet Hough candidate circle, and determines the relative difference parameter of the dust collector ash pellet radius and the distance parameter between the dust collector ash pellet center; it determines the dust collector ash pellet crossover ratio and dynamic crossover ratio threshold; and it obtains the true value of the dust collector ash pellet particle size by using the dust collector ash pellet matching parameters and joint matching conditions.

[0074] The carbon monoxide concentration control module performs cold solidification treatment on the dust removal ash pellets, establishes a thermo-kinematic model for adjusting the particle size of the dust removal ash pellets to reduce the carbon monoxide concentration, and controls the carbon monoxide concentration during the sintering process.

[0075] The beneficial effects of this invention are as follows: This invention proposes a method and system for controlling carbon monoxide concentration in sintering flue gas based on the preparation of dust removal ash pellets. Through standardized control of the particle size identification and pelletizing process of the dust removal ash pellets, particle size is identified online, producing standardized pellets. This makes particle size a controllable variable in the calcination experiment, eliminating particle size dispersion interference and clarifying the specific influence of particle size on carbon monoxide (CO) emissions. The dual-factor synergistic regulation of the dust removal ash ratio and pellet particle size on CO concentration addresses the CO emission problem caused by carbon oxidation of the dust removal ash itself, establishing a quantitative relationship between ratio and particle size. By combining dust collector ash with calcium oxide (CaO) and a binder, and through high-temperature calcination experiments with 1-8mm graded particle sizes, the correlation between the two factors and carbon monoxide (CO) concentration was clarified, and CO emissions were reduced in a targeted manner when dust collector ash was recycled. The high-proportion resource utilization of metallurgical dust collector ash, using 93%~100% high-proportion dust collector ash as the core raw material to prepare pellets, maximizes the reduction and recycling of metallurgical solid waste, solves the pollution and resource waste problems caused by dust collector ash stockpiling in steel enterprises, and eliminates the potential CO emission hazards caused by its recycling, so that the high-value utilization of solid waste and flue gas emission reduction can form a virtuous cycle.

[0076] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for controlling carbon monoxide concentration in sintering flue gas prepared from dust-removed ash pellets, characterized in that: It includes: S1: Obtain the mixed raw materials for preparing dust removal ash pellets; S2: The mixed raw materials obtained in step S1 are processed into pellets. A local weighted attention module and a multi-scale enhanced upsampling module are introduced into the visual recognition model. The local weighted attention module improves the network's sensitivity to pellet features. A gating mechanism is used to refine the calculation of the attention map. By inputting the overall feature map of the dust removal ash pellet image, the accuracy of feature extraction is improved. The multi-scale enhanced upsampling module optimizes the network's recognition performance on the dust removal ash pellet image, reduces and activates the pellet ore image features, obtains multi-scale pooling features at the particle size level of the dust removal ash pellets, extracts and enhances the edge features at the particle size level of the dust removal ash pellets, and fuses the multi-scale features of the dust removal ash pellets. An improved visual recognition model was used to segment and measure the particle size of dust collection ash pellets, identify the pellet size, obtain the pellet size distribution, and obtain the dust collection ash pellets. S3: The particle size of the dust collector ash pellets identified in step S2 is calculated using a joint detection method of edge contour and Hough transform; the dust collector ash pellets are segmented and binarized using a mask, and the contour area of ​​the dust collector ash pellets is selected to determine whether it meets the requirements. The conditions are the selected dust collector ash pellets; the Hough transform equation of the dust collector ash pellets is established to determine the Hough candidate circles of the dust collector ash pellets, and the relative difference parameter of the radius of the dust collector ash pellets is determined. Parameters of distance between the center of dust collection pellets Determine the crossover ratio and dynamic crossover ratio threshold of dust collector ash pellets; obtain the true value of dust collector ash pellet particle size by determining the matching parameters and joint matching conditions of dust collector ash pellets; S4: The dust removal ash pellets obtained in step S3 are subjected to cold solidification treatment. A thermo-kinematic model for controlling the particle size of the dust removal ash pellets to reduce the carbon monoxide concentration is established. The carbon monoxide concentration during the sintering process is controlled by the dust removal ash pellets obtained in step S3.

2. The method for controlling carbon monoxide concentration in sintering flue gas based on dust-removed ash pellets according to claim 1, characterized in that: Step S2 is as follows: S21: Use a pelletizing machine to prepare dust removal ash pellets; S22: Introduce a local weighted attention module into the visual recognition model to improve the sensitivity to sphere features in small targets or complex backgrounds and improve the accuracy of feature extraction; S23: In the visual recognition model, a multi-scale enhanced upsampling module is used to optimize the network's recognition performance on dust removal ash pellet images, reduce the dimensionality and activate the pellet image features, obtain multi-scale pooling features at the particle size level of dust removal ash pellets, extract and enhance the edge features at the particle size level of dust removal ash pellets, and fuse the multi-scale features of dust removal ash pellets.

3. The method for controlling carbon monoxide concentration in sintering flue gas based on dust-removed ash pellets according to claim 2, characterized in that: Step S22 is as follows: ； ； in, For dust ash pellet image pixels Local weight parameters; The image pixels represent dust clumps. To remove dust ash pellets image pixels The neighborhood centered on; Weighting parameters used to adjust weighted measurements; For dust ash pellet image pixels The surrounding sub-region; natural constant of Power; For the pixel index of the dust ash pellet image; It belongs to the symbol; For the first Each dust ash pellet image pixel; This is a feature map of the overall characteristics of the dust ash pellet image; Index for sub-regions; This refers to the multi-scale hierarchical sequence number corresponding to the particle size of the pellets; For attention feature maps; For element-wise dot product; For bilinear interpolation used for rescaling; This is the first channel of the input feature map; This is an activation operation for the characteristics of pellet ore.

4. The method for controlling carbon monoxide concentration in sintering flue gas based on dust-removed ash pellets according to claim 2, characterized in that: The multi-scale pooling features obtained in step S23, based on the particle size hierarchy of dust agglomerates, are as follows: ； ； in, For the first Pooling characteristics of dust collection ash pellets at different particle size levels; This is a 3×3 average pooling operation; For the first -1 intermediate feature map of the particle size hierarchy of dust collection ash pellets; This is a characteristic of intermediate pellet ore; It exhibits characteristics of pelletized mineralization; For the first A refined pooling characteristic diagram of the particle size hierarchy of dust removal ash pellets; Standardized batch operations for pellet ore characteristics; This is a 1×1 convolution operation on the feature map of pellet ore.

5. The method for controlling carbon monoxide concentration in sintering flue gas based on dust-removed ash pellets according to claim 2, characterized in that: Step S23 involves extracting edge features at the particle size level of enhanced dust collection ash pellets, specifically as follows: ； ； in, For the first Edge feature map of the particle size hierarchy of dust removal ash pellets; For the first Enhanced feature map of the particle size hierarchy of dust removal ash pellets.

6. The method for controlling carbon monoxide concentration in sintering flue gas based on dust-removed ash pellets according to claim 1, characterized in that: Step S3 is as follows: S31: Perform segmentation masking binarization on the dust removal ash pellets, and filter out the obtained dust removal ash pellets; determine the minimum circumcircle of the dust removal ash pellet outline, establish the Hough transform equation and Hough candidate circles for the dust removal ash pellets, determine the relative difference parameter of the radius of the dust removal ash pellets, and determine the center distance parameter of the dust removal ash pellets. S32: Determine the crossover ratio of dust removal ash pellets and the dynamic crossover ratio threshold of dust removal ash pellets. ,in, For dynamic intersection-union ratio threshold; Based on the intersection-union ratio threshold; This is the adjustment coefficient; S33: Determine the matching parameters and joint matching conditions of dust collector ash pellets, and use the Hough transform of dust collector ash pellets that meet the above conditions as matching candidates to obtain the true value of the particle size of dust collector ash pellets.

7. The method for controlling carbon monoxide concentration in sintering flue gas based on dust-removed ash pellets according to claim 6, characterized in that: The specific parameters for matching dust ash pellets in step S33 are as follows: ； ， ； in, For matching parameters; This is the radius difference penalty coefficient; The intersection-union ratio of the two circular regions; The outline is a circular area; This is the region of Hough transform; It is a function of the area of ​​the region; The symbol for intersection; This is a merge symbol; For the first The outline circle and the first The relative difference parameter of the radius between the Hough transforms; The radius of the outline circle; The Hough transform radius of the dust collection pellets.

8. The method for controlling carbon monoxide concentration in sintering flue gas based on dust-removed ash pellets according to claim 6, characterized in that: The specific matching conditions for dust collection ash pellets in step S33 are as follows: ； in, The radius difference threshold; This is the control coefficient for the distance from the center of the circle; The distance parameter between the center of the dust collection ash pellet and the center of the Hough transform is the distance parameter between the outline circle and the center of the dust collection ash pellet. This is the dynamic crossover ratio threshold.

9. The method for controlling carbon monoxide concentration in sintering flue gas based on dust-removed ash pellets according to claim 1, characterized in that: Step S4 is as follows: S41: Cold solidification treatment of dust removal ash pellets: The mixed raw materials obtained in step S1 and the dust removal ash pellets of various particle sizes after cold solidification treatment are respectively subjected to high-temperature calcination. S42: During the sintering process of dust removal ash pellets, the pellet size is adjusted by changing the air permeability of the bed and the oxygen diffusion resistance inside the pellets, thereby regulating the ratio of complete combustion to incomplete combustion of carbon and controlling the carbon monoxide concentration.

10. A control system for the method of controlling carbon monoxide concentration in sintering flue gas based on the preparation of dust-removed ash pellets as described in any one of claims 1 to 9, characterized in that: It includes a dust removal ash pellet preparation module, a dust removal ash pellet image recognition module, a dust removal ash pellet particle size recognition module, and a carbon monoxide concentration control module; The dust collector ash pellet preparation module is used to obtain raw materials for dust collector ash pellet preparation, process the raw materials for dust collector ash pellet preparation, and mix the raw materials for dust collector ash pellet preparation. The dust removal ash pellet image recognition module introduces a local weighted attention module and a multi-scale enhanced upsampling module into the visual recognition model; the improved visual recognition model is used to segment the dust removal ash pellet image and calculate the particle size, identify the pellet particle size, obtain the pellet particle size distribution, and obtain the dust removal ash pellet. The dust collector ash pellet size identification module uses a joint detection method of edge contour and Hough transform to calculate the particle size of the dust collector ash pellet size feature map; the dust collector ash pellets are segmented and masked for binarization, the contour area of ​​the dust collector ash pellets is screened, and the dust collector ash pellets that meet the conditions are the screened dust collector ash pellets. The Hough transform equation for dust collector ash pellets is established to determine the candidate Hough circles for dust collector ash pellets, the relative difference parameters of the ash pellet radius and the distance parameters between the ash pellet circles' centers are determined, the crossover ratio and dynamic crossover ratio threshold of dust collector ash pellets are determined, and the matching parameters and joint matching conditions of dust collector ash pellets are determined to obtain the true value of the particle size of the dust collector ash pellets. The carbon monoxide concentration control module performs cold solidification treatment on the dust removal ash pellets, establishes a thermo-kinematic model for adjusting the particle size of the dust removal ash pellets to reduce the carbon monoxide concentration, and controls the carbon monoxide concentration during the sintering process.

Images