Self-adaptive partition combustion-supporting control method for smelting furnace

Abstract

The invention discloses a smelting furnace self-adaptive partition combustion-supporting control method, and relates to the technical field of smelting furnace temperature control. A thermodynamic image containing pixel point coordinates and temperature is obtained through an acquisition device, and partition templates are preset and are in one-to-one correspondence with combustion fans; based on temperature distribution focusing degree and uniformity balance and temperature gradient directivity and smoothness, combined with dynamic weight, calculating a comprehensive adaptation degree; when the degree of adaptation is lower than a threshold value, dynamically adjusting partitions according to expert experience, targeting and a global optimal re-partitioning strategy; and then the deflection angle of the combustion fan is controlled, so that the flame accurately covers the adjusted subarea. According to the method, partition self-adaptive optimization and precise combustion supporting are achieved, the heating efficiency and temperature uniformity are improved, and energy consumption is reduced.

Description

Technical Field

[0001] This invention relates to the field of smelting furnace temperature control technology, specifically to an adaptive zone combustion control method for smelting furnaces. Background Technology

[0002] In the field of temperature control technology for recycled aluminum smelting furnaces, traditional temperature control methods often employ fixed partition templates or overall unified temperature control, which have many technical limitations and application pain points: During the melting process of recycled aluminum, the location of cold material addition is random and the melting progress changes dynamically, causing the distribution of the cold material zone, molten zone, and overheated zone on the surface of the molten pool to continuously change. Fixed partition templates cannot adjust the size and shape of the partitions according to the real-time temperature field, often resulting in a disconnect between the partitions and the actual temperature area. This leads to the cold material zone not being accurately covered and heated, resulting in low melting efficiency, while the molten zone or overheated zone continues to be heated, causing energy waste and even localized burn-out of the furnace body. Furthermore, traditional control methods lack precise control over the directionality and smoothness of the temperature gradient. The current technology lacks a scientific and quantitative adaptation assessment system, relying solely on experience to judge heating effects. This makes it difficult to effectively identify abnormalities such as sudden temperature changes and gradient disturbances, leading to uneven temperature distribution in the molten pool, poor melting consistency of recycled aluminum, and a tendency to produce material defects. Furthermore, the angle of the combustion fan is often fixed or its adjustment is lagging, making it difficult to adapt the flame coverage to dynamically changing zones. This further exacerbates the problems of inaccurate heating and high energy consumption. Moreover, the existing technology fails to dynamically optimize control strategies based on the core requirements of each stage of recycled aluminum melting, making it impossible to achieve a multi-objective balance between heating efficiency, melting quality, energy consumption control, and equipment protection. Consequently, it is difficult to meet the actual needs of efficient, energy-saving, and precise temperature control in industrial production. Summary of the Invention

[0003] The purpose of this invention is to provide an adaptive zoned combustion control method for smelting furnaces, thereby solving the problems mentioned in the background art.

[0004] This invention is achieved through the following technical solution: An adaptive zone combustion control method for a smelting furnace includes the following steps: S1. Acquire a thermal image of the smelting furnace through a data acquisition device. Each pixel in the thermal image has a unique coordinate and a corresponding temperature value. Each pixel is represented as ((x, y), T), where (x, y) is the two-dimensional plane coordinate of the pixel and T is the real-time temperature value at that coordinate. S2. The thermal image is divided into partitions using a preset initial partition template, each partition is numbered, and the partition number corresponds one-to-one with the combustion fan. The combustion fan corresponding to the number is responsible for combustion in the corresponding partition. S3. As the temperature changes during the melting process of recycled aluminum, based on the temperature data in the thermal image, obtain the cold material heating efficiency and partition contour matching degree, temperature anomaly gradient intensity and overheating area of each partition, and combine the dynamic weights adjusted according to the process stage, and substitute them into the preset comprehensive fitness function to calculate the comprehensive fitness of each partition; S4. If the fitness is lower than the preset fitness threshold, adjust the size and / or shape of the corresponding partition to achieve real-time change of the partition; S5. According to the adjusted partition, control the corresponding combustion air blower to rotate and change the air outlet angle, and the burner corresponding to the combustion air blower rotates at the same angle as the air outlet angle, so that the flame generated by the combustion air blower covers the adjusted partition.

[0005] Furthermore, The calculation of the cold material heating efficiency and partition contour matching degree includes the following steps: S30. Identify the molten zone, cold material zone, and overheating zone on the surface of the molten pool in the thermal image through image processing technology. Each zone is defined by a set of pixel points, corresponding to pixel groups with different temperature states on the surface of the molten pool, and accurate distinction is achieved through the temperature feature differences of pixel points: molten zone pixel set P melt ={((x, y), T)|T ∈ [T1, T2]}, cold material zone pixel set P cold ={((x, y), T)|T < T1}, overheating zone pixel set P over ={((x, y), T)|T > T2}, where T1 is the critical melting temperature of recycled aluminum melting and T2 is the preset temperature upper limit; S31. Calculate the cold material heating efficiency. The calculation formula for the cold material heating efficiency is: , where is the cold material heating efficiency, P cold is the cold material zone pixel set, is the current temperature of the cold material zone pixel, is the initial temperature of the cold material zone pixel, is the total energy consumption of the combustion air blower for this partition; S32. Use the image similarity algorithm to calculate the intersection and union ratio of the partition contour and the natural contour of the cold material zone based on the pixel point temperature mutation characteristics, and evaluate the rationality of the partition shape through the intersection and union ratio.

[0006] Furthermore, The calculation method of the temperature anomaly gradient intensity includes the following steps: B1. Perform pixel-level calculations on the thermal image to obtain the temperature gradient of each pixel. The temperature gradient includes the direction parameter and the degree parameter of temperature change. Specifically, the temperature gradient vector of each pixel is calculated using the basic gradient vector formula of the Sobel operator, and then the gradient magnitude is further calculated based on the temperature gradient vector. The gradient magnitude is used to quantify the degree of temperature change. B2. Compare the temperature gradient field of the current partition with the preset ideal gradient pattern pixel by pixel. The ideal gradient pattern is a gradient distribution pattern that points from the burner nozzle to the center of the unmelted solid cold material area and has a clear and smooth temperature gradient transition zone. B3. Pixels whose gradient magnitude exceeds a preset gradient threshold and whose direction differs from the ideal gradient pattern by an angle greater than a preset angle threshold are identified as abnormal gradient pixels. B4. Calculate the sum of the gradient magnitudes of all abnormal gradient pixels within the partition as the temperature anomaly gradient intensity, which is used to quantify the degree of disorder of the temperature gradient within the partition.

[0007] Furthermore, The comprehensive fitness function is: ; in, For overall compatibility, , , , The weights are dynamic and their sum is 1. The heating efficiency of the cold material. The intersection-union ratio (IUU) is used to characterize the fit of the partition contours. For the intensity of the temperature anomaly gradient, For pixel sets based on overheated areas P over The calculated overheated area of ​​the high-temperature zone, the dynamic weight is adjusted according to the process stage of recycled aluminum melting.

[0008] Furthermore, The partitioning adjustment described in S4 adopts one or more of the following re-partitioning strategies: targeted re-partitioning based on hot and cold region identification of thermal image pixel temperature data, global optimal re-partitioning based on optimization algorithm with the goal of maximizing comprehensive fit, and expert experience re-partitioning based on rule base combined with melting process rules. The strategy priority for partition adjustment is as follows: first, perform expert experience-based repartitioning based on rule base; then, optimize local partition details through targeted repartitioning based on hot and cold region identification; and finally, verify the overall adaptability with global optimal repartitioning based on optimization algorithm to ensure that each partition has no overlap and no coverage blind spots.

[0009] Furthermore, The targeted re-division based on hot and cold region identification includes the following steps: C1. The thermal image is processed using an image segmentation algorithm to accurately identify the cold material zone, melting zone and overheated zone based on the temperature value of each pixel ((x, y), T); C2. Adjust the partitions according to the preset re-division principles, which include: a partition should cover an independent cold material zone as completely as possible. P cold The pixel set, partition boundaries and cold / hot areas are aligned with the natural contours formed by the temperature distribution of the pixels to avoid partitions with too few pixels or too narrow and elongated shapes. C3. By merging, splitting, or deforming and re-dividing, adjacent partitions with uniform internal pixel temperature ((x, y), T) and small temperature gradient are merged, containing multiple independent large-area cold material zones. P cold The pixel set is divided into partitions, and the partition boundaries are dynamically adjusted to achieve deformation based on the natural contour formed by the temperature change of the pixels in the cold material area.

[0010] Furthermore, The global optimal repartitioning based on the optimization algorithm includes the following steps: D1. Using the comprehensive fit function as the objective function, the objective is to maximize the overall fit of the entire furnace. D2. Set the optimization variable as the boundary position of the partition, and the constraints include: the partition covers the entire surface of the molten pool without overlap, the area of ​​the partition is not less than the minimum area effectively covered by the burner, and the shape of the partition does not appear excessively singular. D3. The optimal partitioning scheme is generated iteratively using a heuristic optimization algorithm, which may include a genetic algorithm or a particle swarm optimization algorithm.

[0011] Furthermore, The rule-based expert experience reclassification includes the following: Z1. If new cold material is detected near the furnace door, at least two zones near the furnace door will be merged, and the angle of the corresponding combustion fan will be adjusted so that the burner flame can cover the merged zone. Z2. If the temperature standard deviation of a certain partition is consistently higher than the preset temperature threshold, the partition will be split into at least two partitions. Z3. If the system is in the heat preservation stage, it will automatically switch to the preset large-area, low-power heat preservation partition template.

[0012] Furthermore, The control of the left, right, up, and down deflection angles of the combustion fan described in S5 includes the following steps: S50. Calculate the geometric center of the adjusted partition, and use the geometric center as the aiming target point of the burner flame. The combustion fan matches the air outlet angle with the target point as the reference. S51. Based on the shape, size and inherent parameters of the adjusted zone, calculate the required left and right deflection angles and up and down pitch angles of the combustion fan to ensure that the air outlet direction matches the burner flame jet direction and covers the entire zone. S52. Control the combustion fan to smoothly transition from the current angle to the calculated target angle, with the angle change rate not exceeding the preset angle change threshold, so that the burner flame stably covers the zone with the assistance of the combustion fan.

[0013] The beneficial effects of this invention are as follows: 1. Breaking away from the limitations of traditional fixed partition templates, this method calculates cold material heating efficiency based on temperature changes in the pixel set of the cold material zone and the energy consumption of the combustion fan. Simultaneously, it evaluates the fit between the partition outline and the natural outline of the cold material zone through cross-validation ratio (CGR), ensuring that the partitions accurately match the actual distribution of cold material. When the overall fit is insufficient, strategies such as targeted re-partitioning and expert-experienced re-partitioning dynamically adjust the size and shape of the partitions, ensuring that each partition completely covers the independent cold material zone, preventing insufficient heating of the cold material zone or overheating of the high-temperature zone. This "precise matching of cold material distribution + dynamic partition optimization" model allows the burner flame, after being adapted by the combustion fan, to precisely act on the target area, significantly improving cold material heating efficiency, shortening the melting cycle of recycled aluminum, and reducing ineffective heating caused by partition misalignment, thus lowering energy consumption.

[0014] 2. Traditional control methods lack precise control over temperature gradients, easily leading to uneven temperature distribution and localized temperature abrupt changes in the molten pool, which in turn causes defects in the recycled aluminum material. This invention calculates pixel-level temperature gradients using the Sobel operator, combines this with an ideal gradient pattern to identify abnormal gradient pixels, quantifies the intensity of abnormal temperature gradients, and effectively captures areas of disordered temperature gradients. Simultaneously, a comprehensive fit function uses the intensity of abnormal temperature gradients and the overheated area as deduction factors to drive zonal adjustments to eliminate temperature abrupt changes and optimize gradient distribution. This results in a smooth and orderly temperature gradient on the molten pool surface, avoiding localized overheating or unmelted dead zones, ensuring that recycled aluminum melts in a uniform temperature field, reducing problems such as component segregation and impurity residue caused by uneven temperature, and significantly improving melting quality and product consistency.

[0015] 3. After the zoning adjustment, the geometric center of the zoning is used as the target point. Combined with the shape and size of the zoning and the inherent parameters of the burner, the left and right deflection and up and down pitch angles of the fan are accurately calculated. The fan is then controlled to smoothly transition from the current angle to the target angle, ensuring that the angle change rate does not exceed the threshold. This allows the burner flame to stably and completely cover the adjusted zoning with the assistance of the combustion fan, avoiding heating failure caused by flame deviation. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the logic flow of the present invention; Figure 2 This is a scene diagram of the self - adaptive zone combustion support of the melting furnace. Specific embodiments

[0017] The following will further elaborate on the present invention in conjunction with embodiments and the accompanying drawings. However, the implementation manners of the present invention are not limited thereto.

[0018] Embodiment, see Figures 1 to 2 : A method for controlling the self - adaptive zone combustion support of a melting furnace includes the following steps: S1. Obtain the thermal image of the melting furnace through a collection device. Each pixel point in the thermal image has a unique coordinate and a corresponding temperature value, and each pixel point is represented as ((x, y), T), where (x, y) is the two - dimensional plane coordinate of the pixel point, and T is the real - time temperature value at this coordinate; S2. Use a preset initial zoning template to zone the thermal image, number each zone, and the zone numbers correspond to the combustion - supporting blowers one by one. The combustion - supporting blower corresponding to the number is responsible for the combustion support of the corresponding zone; S3. As the temperature changes during the melting process of recycled aluminum, according to the temperature data in the thermal image, obtain the cold - charge heating efficiency and the zoning contour matching degree, the temperature anomaly gradient intensity and the over - heat area of each zone, and combine the dynamic weights adjusted according to the process stage, and substitute them into a preset comprehensive fitness function to calculate the comprehensive fitness of each zone; S4. If the fitness is lower than the preset fitness threshold, adjust the size and / or shape of the corresponding zone to achieve real - time change of the zone; S5. According to the adjusted zone, control the corresponding combustion - supporting blower to rotate and change the air outlet angle, and the burner corresponding to the combustion - supporting blower rotates at the same angle as the air outlet angle, so that the flame generated by the combustion - supporting blower covers the adjusted zone.

[0019] Furthermore, The calculation of the cold - charge heating efficiency and the zoning contour matching degree includes the following steps: S30. Identify the molten zone, cold - charge zone, and over - heat zone on the surface of the molten pool in the thermal image through image - processing technology. Each zone is defined by a set of pixel points, corresponding to pixel groups with different temperature states on the surface of the molten pool, and is accurately distinguished through the temperature - feature differences of pixel points: the pixel set of the molten zone P melt ={((x, y), T)|T ∈ [T1, T2]}, the pixel set of the cold - charge zone P cold ={((x, y), T)|T < T1}, the pixel set of the over - heat zoneP over ={((x, y), T)|T>T2}, where T1 is the critical melting temperature for recycled aluminum and T2 is the preset upper limit of temperature; S31. Calculate the heating efficiency of the cold material, wherein the formula for calculating the heating efficiency of the cold material is: ,in For cold material heating efficiency, P cold For the cold material area pixel set, This represents the current temperature of the pixels in the cold material zone. The initial temperature of the pixels in the cold material zone. This represents the total energy consumption of the combustion air blowers in this zone; S32. The intersection-union ratio of the partition contour and the natural contour of the cold material area based on the temperature change characteristics of the pixel is calculated using an image similarity algorithm. The rationality of the partition shape is evaluated by the intersection-union ratio.

[0020] The core advantage of calculating the cold material heating efficiency and the conformity of the partition contour is that it can accurately quantify the energy efficiency of cold material heating and the rationality of the partition shape, avoiding inefficient heating and the disconnect between the partition and the actual cold material area. In practice, image processing technology is used to accurately identify the molten area, cold material area and overheated area on the surface of the molten pool. The cold material heating efficiency is calculated based on the current and initial temperature difference of the cold material area pixel set and the total energy consumption of the combustion fan. Then, the intersection-union ratio (IoU) of the partition contour and the natural contour of the cold material area is obtained using an image similarity algorithm. For example, in the early stage of melting, the cold material heating efficiency of a certain partition is η=0.85 and the IoU is 0.8. These two indicators directly reflect the coverage and heating effect of the burner flame on the cold material area, which not only ensures that the cold material can be heated efficiently, but also avoids energy waste caused by unreasonable partition shape, providing accurate basic data for subsequent comprehensive adaptability judgment.

[0021] Furthermore, The method for calculating the intensity of the temperature anomaly gradient includes the following steps: B1. Perform pixel-level calculations on the thermal image to obtain the temperature gradient of each pixel. The temperature gradient includes the direction parameter and the degree parameter of temperature change. Specifically, the temperature gradient vector of each pixel is calculated using the basic gradient vector formula of the Sobel operator, and then the gradient magnitude is further calculated based on the temperature gradient vector. The gradient magnitude is used to quantify the degree of temperature change. B2. Compare the temperature gradient field of the current partition with the preset ideal gradient pattern pixel by pixel. The ideal gradient pattern is a gradient distribution pattern that points from the burner nozzle to the center of the unmelted solid cold material area and has a clear and smooth temperature gradient transition zone. B3. Pixels whose gradient magnitude exceeds a preset gradient threshold and whose direction differs from the ideal gradient pattern by an angle greater than a preset angle threshold are identified as abnormal gradient pixels. B4. Calculate the sum of the gradient magnitudes of all abnormal gradient pixels within the partition as the temperature anomaly gradient intensity, which is used to quantify the degree of disorder of the temperature gradient within the partition.

[0022] The calculation of abnormal temperature gradient intensity can accurately identify the disorder of temperature gradient within a zone, avoiding uneven melting and localized furnace burn-out caused by sudden temperature changes. In practice, the Sobel operator is first used to perform pixel-level calculations on the thermal image to obtain the temperature gradient containing parameters of direction and intensity. Then, it is compared with the ideal gradient pattern of "smooth transition from the burner nozzle to the center of the cold material zone". Pixels with gradient magnitudes exceeding the threshold and large directional deviations are identified as abnormal gradient pixels. The sum of their magnitudes is the abnormal temperature gradient intensity. For example, if a zone has multiple temperature abrupt change points due to uneven distribution of cold material, the calculated abnormal temperature gradient intensity A=0.15. This value clearly quantifies the disorder of temperature changes within the zone, providing a clear target basis for subsequent zone adjustments and effectively avoiding melting quality problems caused by temperature gradient disorder.

[0023] Furthermore, The comprehensive fitness function is: ; in, For overall compatibility, , , , The weights are dynamic and their sum is 1. The heating efficiency of the cold material. The intersection-union ratio (IUU) is used to characterize the fit of the partition contours. For the intensity of the temperature anomaly gradient, For pixel sets based on overheated areas P over The calculated overheated area of ​​the high-temperature zone, the dynamic weight is adjusted according to the process stage of recycled aluminum melting.

[0024] The comprehensive fit function integrates four dimensions: cold material heating efficiency, zone contour fit, temperature anomaly gradient intensity, and overheated area, and is combined with dynamic weights that are adjusted according to the process stage. The aim is to achieve a multi-objective balanced evaluation so that the fit judgment fits the core requirements of different stages of recycled aluminum melting. For example, in the initial stage of melting, there is a large quantity of cold material, and the core objective is to rapidly heat the cold material. At this time, the dynamic weights W1 (cold material heating efficiency weight) are set to 0.4, W2 (cross-union ratio weight) to 0.3, and W3 (temperature anomaly gradient intensity weight) and W4 (overheated area weight) to 0.15 each. If the heating efficiency η = 0.85, the cross-union ratio IoU = 0.8, the temperature anomaly gradient intensity A = 0.08, and the overheated area S... over=0.06, the calculated overall fit is F=0.4×0.85+0.3×0.8-0.15×0.08-0.15×0.06=0.34+0.24-0.012-0.009=0.559, which is higher than the preset fit threshold of 0.5, so no adjustment is needed; As the melting process progresses into the middle stage, the amount of cold material decreases, and the core objective shifts to preventing overheating. At this point, the weights are adjusted to W1=0.2, W2=0.2, W3=0.3, and W4=0.3. If the S value of this partition is at this stage... over If A increases to 0.2 and A increases to 0.15, then F = 0.2 × 0.85 + 0.2 × 0.8 - 0.3 × 0.15 - 0.3 × 0.2 = 0.17 + 0.16 - 0.045 - 0.06 = 0.225, which is lower than the threshold, triggering an adjustment. The integration of dynamic weights and multiple indicators ensures that the core needs of each stage are met first, while also balancing heating efficiency, melting quality, and energy consumption control.

[0025] Furthermore, The partitioning adjustment described in S4 adopts one or more of the following re-partitioning strategies: targeted re-partitioning based on hot and cold region identification of thermal image pixel temperature data, global optimal re-partitioning based on optimization algorithm with the goal of maximizing comprehensive fit, and expert experience re-partitioning based on rule base combined with melting process rules. The strategy priority for partition adjustment is as follows: first, perform expert experience-based repartitioning based on rule base; then, optimize local partition details through targeted repartitioning based on hot and cold region identification; and finally, verify the overall adaptability with global optimal repartitioning based on optimization algorithm to ensure that each partition has no overlap and no coverage blind spots.

[0026] The partitioning adjustment employs three strategies: targeted repartitioning, globally optimal repartitioning, and expert-experience repartitioning. It prioritizes "expert experience first, targeted optimization of details, and global verification for finalization," balancing response speed, local accuracy, and global optimality to ensure no overlapping partitions or blind spots. For example, when a new batch of cold material is detected near the furnace door, expert-experience repartitioning (Z1) is executed first, quickly merging two adjacent partitions near the furnace door to prevent the cold material area from being split and causing dispersed heating. Next, targeted repartitioning uses image segmentation algorithms to identify the natural contour of the new cold material area, adjusting the boundaries of the merged partitions to fully cover the pixel set of the cold material area while avoiding overly elongated partition shapes. Finally, a genetic algorithm is used for globally optimal repartitioning verification, ensuring that the merged partitions do not overlap with other partitions, and that the total area of ​​all partitions covers the entire molten pool surface. The area of ​​each partition is not less than the minimum area effectively covered by the burner, ultimately achieving an adjustment effect of "rapid response to changes in operating conditions + precise local adaptation + global overall optimality."

[0027] Furthermore, The targeted re-division based on hot and cold region identification includes the following steps: C1. The thermal image is processed using an image segmentation algorithm to accurately identify the cold material zone, melting zone and overheated zone based on the temperature value of each pixel ((x, y), T); C2. Adjust the partitions according to the preset re-division principles, which include: a partition should cover an independent cold material zone as completely as possible. P cold The pixel set, partition boundaries and cold / hot areas are aligned with the natural contours formed by the temperature distribution of the pixels to avoid partitions with too few pixels or too narrow and elongated shapes. C3. By merging, splitting, or deforming and re-dividing, adjacent partitions with uniform internal pixel temperature ((x, y), T) and small temperature gradient are merged, containing multiple independent large-area cold material zones. P cold The pixel set is divided into partitions, and the partition boundaries are dynamically adjusted to achieve deformation based on the natural contour formed by the temperature change of the pixels in the cold material area.

[0028] Targeted re-segmentation is based on image segmentation to identify hot and cold regions. Through merging, splitting, and deformation adjustments, it precisely matches the distribution of hot and cold regions, optimizes local partition details, and avoids the problem of "disconnection between partitions and actual temperature regions." For example, in the middle of a smelting furnace, there are two adjacent partitions. Image segmentation algorithms detect that the pixel temperatures within these two partitions are uniform, with minimal temperature gradients, indicating no clear hot and cold boundaries. They are merged according to the re-segmentation principle, reducing the control complexity of the combustion fan and improving heating uniformity. On the other side of the furnace, a certain partition contains three independent large-area cold material zones with pixel sets... P cold The components are independent of each other. At this point, the partition is divided into three smaller partitions, each corresponding to a cold material zone. Then, based on the natural contour of each cold material zone formed by temperature changes, the boundary of each smaller partition is dynamically adjusted so that the partition boundary is completely aligned with the contour of the cold material zone. This ensures that each cold material zone is accurately covered by the flame of the corresponding burner after the air outlet angle of the combustion fan is adapted, thus avoiding inadequate heating of some cold materials.

[0029] Furthermore, The global optimal repartitioning based on the optimization algorithm includes the following steps: D1. Using the comprehensive fit function as the objective function, the objective is to maximize the overall fit of the entire furnace. D2. Set the optimization variable as the boundary position of the partition, and the constraints include: the partition covers the entire surface of the molten pool without overlap, the area of ​​the partition is not less than the minimum area effectively covered by the burner, and the shape of the partition does not appear excessively singular. D3. The optimal partitioning scheme is generated iteratively using a heuristic optimization algorithm, which may include a genetic algorithm or a particle swarm optimization algorithm.

[0030] Global optimal repartitioning aims to maximize overall fit. It iteratively generates partitioning schemes that satisfy constraints through optimization algorithms, overcoming the limitations of local adjustments and achieving optimal heating of the entire furnace while ensuring that partitions comply with equipment and process constraints. For example, in the initial partitioning template of a certain smelting furnace, there are problems such as two partitions partially overlapping and one partition having an area smaller than the minimum effective coverage area of ​​the burner (resulting in incomplete flame coverage). Using the overall fit function as the objective function, constraints are set such as "no overlapping partitions, coverage of the entire molten pool, area meeting standards, and non-odd shapes." Particle swarm optimization is used for iterative optimization. After 50 iterations, the partition boundaries are redefined: overlapping parts are split, smaller partitions are merged and expanded, and ultimately all partitions satisfy the constraints. The overall fit of the entire furnace is improved compared to the initial state, achieving a unity of "individual partition fit" and "optimal overall heating."

[0031] Furthermore, The rule-based expert experience reclassification includes the following: Z1. If new cold material is detected near the furnace door, at least two zones near the furnace door will be merged, and the angle of the corresponding combustion fan will be adjusted so that the burner flame can cover the merged zone. Z2. If the temperature standard deviation of a certain partition is consistently higher than the preset temperature threshold, the partition will be split into at least two partitions. Z3. If the system is in the heat preservation stage, it will automatically switch to the preset large-area, low-power heat preservation partition template.

[0032] Based on a clear rule base set according to actual process scenarios, it can quickly respond to common operating condition changes without complex calculations, improving control response speed. At the same time, it is in line with industry practical experience. For example, during the melting process, if the temperature standard deviation of a certain zone is detected to be higher than the preset temperature threshold for 10 consecutive minutes, it indicates that the temperature distribution of that zone is extremely uneven, and there may be a situation of "one side overheating and one side not melting". According to the temperature line of sudden temperature change in the thermal image as the boundary, Z2 splits the zone into two zones, and controls the corresponding combustion fans respectively. The heating intensity is increased in the low temperature area and the power is reduced in the high temperature area, which quickly improves the temperature uniformity. When the recycled aluminum is completely melted and enters the heat preservation stage, Z3 automatically switches to the preset large area, low power heat preservation zone template, merging the original 6 small zones into 2 large area zones. The power of the combustion fans is reduced, which not only avoids local overheating caused by high power in small zones, but also saves energy and stably maintains the furnace temperature within the heat preservation range.

[0033] Furthermore, The control of the left, right, up, and down deflection angles of the combustion fan described in S5 includes the following steps: S50. Calculate the geometric center of the adjusted partition, and use the geometric center as the aiming target point of the burner flame. The combustion fan matches the air outlet angle with the target point as the reference. S51. Based on the shape, size and inherent parameters of the adjusted zone, calculate the required left and right deflection angles and up and down pitch angles of the combustion fan to ensure that the air outlet direction matches the burner flame jet direction and covers the entire zone. S52. Control the combustion fan to smoothly transition from the current angle to the calculated target angle, with the angle change rate not exceeding the preset angle change threshold, so that the burner flame stably covers the zone with the assistance of the combustion fan.

[0034] The combustion fan's outlet angle control is achieved by locating the geometric center of the partition, calculating the target angle based on the burner's inherent parameters, and smoothly transitioning the adjustment. This ensures that the burner flame, after being adapted by the combustion fan's outlet airflow, precisely covers the adjusted partition, preventing burner flame misalignment that could lead to heating failure. For example, if a partition is adjusted to be an ellipse with a major axis of 1.2 meters and a minor axis of 0.8 meters, its geometric center coordinates are calculated to be x=3.5m and y=2.8m. Combining this with the burner's installation position and rated flame diffusion angle, it is further calculated that the combustion fan needs to deflect 12° to the left and pitch 6° upwards to ensure the burner flame completely covers the elliptical partition with the fan's assistance. The control system then smoothly transitions the fan from the current angle to the target angle, ensuring that the burner flame always precisely points to the partition center, covering the cold material within the entire partition for uniform heating, while avoiding sudden angle changes that could impact the fan's shaft, gears, and other mechanical structures, thus extending the equipment's lifespan.

[0035] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. An adaptive zoned combustion control method for a smelting furnace, characterized in that, Includes the following steps: S1. Acquire a thermal image of the smelting furnace through a data acquisition device. Each pixel in the thermal image has a unique coordinate and a corresponding temperature value. Each pixel is represented as ((x, y), T), where (x, y) is the two-dimensional plane coordinate of the pixel and T is the real-time temperature value at that coordinate. S2. The thermal image is divided into partitions using a preset initial partition template, each partition is numbered, and the partition number corresponds one-to-one with the combustion fan. The combustion fan corresponding to the number is responsible for combustion in the corresponding partition. S3. As the temperature changes during the melting of recycled aluminum, based on the temperature data in the thermal image, the cold material heating efficiency and the degree of conformity of the partition contour, the intensity of the temperature anomaly gradient and the overheated area of ​​each partition are obtained. Combined with the dynamic weights adjusted with the process stage, the results are substituted into the preset comprehensive fit function to calculate the comprehensive fit of each partition. S4. If the overall adaptability is lower than the preset adaptability threshold, the size and / or shape of the corresponding partition is adjusted to achieve real-time changes in the partition; S5. Based on the adjusted zone, control the rotation of the corresponding combustion fan to change the air outlet angle. The burner corresponding to the combustion fan rotates at the same angle as the air outlet angle, so that the flame generated by the combustion fan covers the adjusted zone.

2. The adaptive zone combustion control method for a smelting furnace according to claim 1, characterized in that, The calculation of the cold material heating efficiency and the degree of conformity with the partition profile includes the following steps: S30. Identify the molten zone, cold material zone, and overheated zone on the surface of the molten pool in the thermal image through image processing technology. Each zone is defined by a set of pixel points, corresponding to pixel groups with different temperature states on the surface of the molten pool, and is accurately distinguished through the temperature feature differences of pixel points: the pixel set of the molten zone P melt ={(x, y), T | T ∈ [T1, T2]}, the pixel set of the cold material zone ={(x, y), T | T < T1}, the pixel set of the overheated zone P over ={(x, y), T | T > T2}, where T1 is the critical melting temperature for the melting of recycled aluminum, and T2 is the preset upper temperature limit; S31. Calculate the heating efficiency of the cold material, wherein the formula for calculating the heating efficiency of the cold material is: ,in For cold material heating efficiency, For the pixel set of the cold material area, This represents the current temperature of the pixels in the cold material zone. The initial temperature of the pixels in the cold material zone. This represents the total energy consumption of the combustion air blowers in the corresponding zones; S32. The intersection-union ratio of the partition contour and the natural contour of the cold material area based on the temperature change characteristics of the pixel is calculated using an image similarity algorithm. The rationality of the partition shape is evaluated by the intersection-union ratio.

3. The adaptive zone combustion control method for a smelting furnace according to claim 1, characterized in that, The calculation of the temperature anomaly gradient intensity includes the following steps: B1. Perform pixel-level calculations on the thermal image to obtain the temperature gradient of each pixel. The temperature gradient includes the direction parameter and the degree parameter of temperature change. Specifically, the temperature gradient vector of each pixel is calculated using the basic gradient vector formula of the Sobel operator, and then the gradient magnitude is further calculated based on the temperature gradient vector. The gradient magnitude is used to quantify the degree of temperature change. B2. Compare the temperature gradient field of the current partition with the preset ideal gradient pattern pixel by pixel. The ideal gradient pattern is a gradient distribution pattern that points from the burner nozzle to the center of the unmelted solid cold material area and has a clear and smooth temperature gradient transition zone. B3. Pixels whose gradient magnitude exceeds a preset gradient threshold and whose direction differs from the ideal gradient pattern by an angle greater than a preset angle threshold are identified as abnormal gradient pixels. B4. Calculate the sum of the gradient magnitudes of all abnormal gradient pixels within the partition as the temperature anomaly gradient intensity, which is used to quantify the degree of disorder of the temperature gradient within the partition.

4. The adaptive zone combustion control method for a smelting furnace according to claim 2 or 3, characterized in that, The comprehensive fitness function is: ; in, For overall compatibility, , , , The weights are dynamic and their sum is 1. The heating efficiency of the cold material. The intersection-union ratio (IUU) is used to characterize the fit of the partition contours. For the intensity of the temperature anomaly gradient, For pixel sets based on overheated areas P over The calculated overheated area of ​​the high-temperature zone, the dynamic weight is adjusted according to the process stage of recycled aluminum melting.

5. The adaptive zone combustion control method for a smelting furnace according to claim 1, characterized in that, The partitioning adjustment described in S4 adopts one or more of the following re-partitioning strategies: targeted re-partitioning based on hot and cold region identification of thermal image pixel temperature data, global optimal re-partitioning based on optimization algorithm with the goal of maximizing comprehensive fit, and expert experience re-partitioning based on rule base combined with melting process rules. The strategy priority for partition adjustment is as follows: first, perform expert experience-based repartitioning based on rule base; then, optimize local partition details through targeted repartitioning based on hot and cold region identification; and finally, verify the overall adaptability with global optimal repartitioning based on optimization algorithm to ensure that each partition has no overlap and no coverage blind spots.

6. The adaptive zone combustion control method for a smelting furnace according to claim 5, characterized in that, The targeted re-division based on hot and cold region identification includes the following steps: C1. The thermal image is processed using an image segmentation algorithm to accurately identify the cold material zone, melting zone and overheated zone based on the temperature value of each pixel ((x, y), T); C2. Adjust the partitions according to the preset re-division principles, which include: a partition should cover an independent cold material zone as completely as possible. P cold The pixel set, partition boundaries and cold / hot areas are aligned with the natural contours formed by the temperature distribution of the pixels to avoid partitions with too few pixels or too narrow and elongated shapes. C3. By merging, splitting, or deforming and re-dividing, adjacent partitions with uniform internal pixel temperature ((x, y), T) and small temperature gradient are merged, containing multiple independent large-area cold material zones. P cold The pixel set is divided into partitions, and the partition boundaries are dynamically adjusted to achieve deformation based on the natural contour formed by the temperature change of the pixels in the cold material area.

7. The adaptive zone combustion control method for a smelting furnace according to claim 5, characterized in that, The global optimal repartitioning based on the optimization algorithm includes the following steps: D1. Using the comprehensive fit function as the objective function, the objective is to maximize the overall fit of the entire furnace. D2. Set the optimization variable as the boundary position of the partition, and the constraints include: the partition covers the entire surface of the molten pool without overlap, the area of ​​the partition is not less than the minimum area of ​​the effective flame coverage of the burner, and the shape of the partition does not appear excessively singular. D3. The optimal partitioning scheme is generated iteratively using a heuristic optimization algorithm, which may include a genetic algorithm or a particle swarm optimization algorithm.

8. The adaptive zone combustion control method for a smelting furnace according to claim 5, characterized in that, The rule-based expert experience reclassification includes the following: Z1. If new cold material is detected near the furnace door, at least two zones near the furnace door will be merged, and the angle of the corresponding combustion fan will be adjusted so that the burner flame can cover the merged zone. Z2: If the temperature standard deviation of a certain partition is consistently higher than the preset temperature threshold, the partition will be split into at least two partitions. Z3: If the system is in the heat preservation stage, it will automatically switch to the preset large-area, low-power heat preservation partition template.

9. The adaptive zone combustion control method for a smelting furnace according to claim 1, characterized in that, The control of the left, right, up, and down deflection angles of the combustion fan described in S5 includes the following steps: S50. Calculate the geometric center of the adjusted partition, and use the geometric center as the aiming target point of the burner flame. The combustion fan matches the air outlet angle with the target point as the reference. S51. Based on the shape, size and inherent parameters of the adjusted zone, calculate the required left and right deflection angles and up and down pitch angles of the combustion fan to ensure that the air outlet direction matches the burner flame jet direction and covers the entire zone. S52. Control the combustion fan to smoothly transition from the current angle to the calculated target angle, with the angle change rate not exceeding the preset angle change threshold, so that the burner flame stably covers the zone with the assistance of the combustion fan.

Images