Convex steel rolling plate shape control method and system based on deep learning

Abstract

The invention relates to the technical field of deep learning machine vision control embossed steel rolling, in particular to a deep learning-based embossed steel rolling plate shape control method and system.The method comprises the steps that a blank is heated and descaled, a first vision detection unit is arranged at an outlet of a heating furnace, and when the centering deviation is larger than a first threshold value, hydraulic deviation correction is triggered; a second visual detection unit is arranged on the roughing mill, a rolled piece surface oxide skin residual area is extracted through an image segmentation algorithm, and whether secondary descaling is triggered or not is detected; a third visual detection unit is arranged in the intermediate rolling mill, temperature distribution is uneven, flow distribution of cooling water of the finishing mill is adjusted, and if the fillet radius error exceeds a fourth threshold value, the side pressure amount of the vertical roller of the finishing mill is corrected; and a fourth visual detection unit is arranged on the finishing mill, when the filling rate is smaller than a fifth threshold value, the roll gap compensation amount is calculated, and the rolling reduction of the finishing mill is corrected in real time. The embossed steel is combined with deep learning and machine vision, so that the manufactured embossed steel is better in quality and more accurate in shape.

Description

Technical Field

[0001] The present invention relates to the technical field of deep learning machine vision controlled embossed steel rolling, and specifically to a method and system for controlling the plate shape of embossed steel rolling based on deep learning. Background Art

[0002] During the production of embossed steel, shape control is a key step in ensuring product quality and production efficiency. Traditional embossed steel shape control methods rely primarily on manual experience and relatively simple sensor monitoring methods, which have many limitations.

[0003] Traditional descaling methods often struggle to fully and accurately remove the oxide layer. The remaining oxide layer not only affects the friction distribution during rolling but can also cause surface defects, impacting the quality of the finished product. Furthermore, alignment methods suffer from low precision, making it difficult to accurately and promptly correct deviations. This can lead to inaccurate pass alignment during subsequent rolling, compromising the shape and dimensional accuracy of the rolled piece.

[0004] In the traditional rough rolling stage, there is a lack of real-time monitoring and precise adjustment means, making it difficult to ensure that the width expansion of the rolled piece meets the design requirements, which easily causes dimensional deviations of the rolled piece. In the intermediate rolling stage, the traditional detection method for the fillet radius of the transition zone between the flange and the waist of the rolled piece is not accurate enough, and it is impossible to detect and correct the fillet radius deviation in time, resulting in the shape of the rolled piece not meeting the design requirements. In the finishing rolling stage, there is a lack of effective real-time monitoring and precise adjustment mechanism, and the detection and processing of dimensional tolerances are also relatively lagging. When deviations are discovered, unqualified products have often been produced, resulting in waste of resources and reduced production efficiency. Summary of the Invention

[0005] The present invention provides a method and system for controlling the plate shape of convex steel rolling based on deep learning.

[0006] The technical solutions of the present invention are as follows:

[0007] The deep learning-based method for controlling the plate shape of embossed steel rolling includes the following steps:

[0008] S1. Heating the blank to reach the preset rolling temperature range, and using high-pressure water to descale and remove the surface oxide layer;

[0009] S2. A first visual inspection unit is installed above the conveyor roller between the heating furnace outlet and the slab mill to identify the edge position of the slab and calculate the offset of the slab centerline through 3D contour reconstruction. When the centering deviation exceeds a first threshold, hydraulic correction is triggered and the centering data is transmitted to the roughing mill in real time as a reference for the roughing mill pass alignment.

[0010] S3. A second visual inspection unit is installed at the exit of the roughing mill. The residual scale area on the surface of the rolled piece is extracted in real time using an image segmentation algorithm. When the residual area detected is greater than a second threshold, secondary descaling is triggered. The width spread is simultaneously calculated. If the deviation between the measured width spread and the theoretical value is greater than a third threshold, the side pressure of the vertical roll of the intermediate rolling mill is adjusted.

[0011] S4. A third visual inspection unit is provided at the outlet of the intermediate rolling mill to monitor the temperature distribution of the rolled piece in real time. When uneven temperature distribution is detected, the flow distribution of the cooling water of the finishing mill is adjusted. The transition area between the flange and the waist of the rolled piece is scanned. If the measured value of the fillet radius deviates from the designed value by more than a fourth threshold, the lateral pressure of the vertical roll of the finishing mill is corrected.

[0012] S5. A fourth visual inspection unit is set at the outlet side of the finishing mill. The flange cross-sectional profile is generated through three-dimensional point cloud reconstruction. The filling rate is calculated by comparing with the theoretical model. When the filling rate is less than the fifth threshold, the roll gap compensation amount is calculated and the finishing mill reduction amount is corrected in real time. At the same time, the cross-sectional dimension tolerance of the rolled piece is detected. If the flange deviation is greater than the sixth threshold, the rolled piece is sent back to the finishing mill and rolled again until the output flange deviation is within the qualified range.

[0013] The specific operation of adjusting the side pressure of the vertical rolls in S3 is to calculate the side pressure adjustment amount of the vertical rolls:

[0014] ΔS=αΔW+C,

[0015] Among them, ΔS is the side pressure adjustment of the vertical roller, α is the width compensation coefficient, ΔW is the deviation between the measured width and the theoretical value, and C is the system error compensation term;

[0016]

[0017] Among them, σ s is the yield strength of the rolled material, β is the equilibrium coefficient, usually taken as 300.

[0018] The third visual inspection unit is also equipped with an infrared camera to detect the temperature difference of the rolled piece. If the temperature difference between the lowest temperature and the highest temperature of the rolled piece is greater than the maximum temperature difference threshold, the flow rate of cooling water in the high temperature area corresponding to the highest temperature is increased to Q new :

[0019] Q new =Q0×[1+0.02(ΔT-50)],

[0020] Among them, Q new is the cooling water flow after adjustment, Q0 is the original set cooling water flow, and ΔT is the temperature difference of the rolled piece;

[0021] At the same time, increase the rolling speed of the roughing mill to v new :

[0022] vnew =v0×[1+0.01(50-ΔT)],

[0023] Among them, v new is the adjusted rolling speed, and v0 is the original set rolling speed.

[0024] In S2, the three-dimensional contour reconstruction can also calculate the cross-sectional asymmetry of the rolled piece. If the cross-sectional asymmetry exceeds a sixth threshold, the rolled piece is returned to the heating furnace for reheating.

[0025] The specific operation of correcting the finishing mill reduction in S5 is to adjust the finishing mill reduction to Δh:

[0026]

[0027] Among them, Δh is the total reduction after compensation, h0 is the original set reduction, η is the flange filling rate, and K is the material deformation compensation coefficient.

[0028] After the reduction of the finishing mill is adjusted, the rolling speed is reduced by 10%-15% of the original rolling speed.

[0029] The fourth visual inspection unit identifies scale indentation defects on the rolled piece based on the spectral characteristics of the scale. If the scale indentation defect rate increases, the second threshold and the pressure of high-pressure water descaling are dynamically adjusted.

[0030] The fourth visual inspection unit can also detect surface defects of rolled products in real time when the flange deviation is within the qualified range. If surface defects exist, the rolled products will be automatically coded and marked for removal.

[0031] The deep learning-based embossed steel rolling shape control system includes:

[0032] The heating furnace heats the billet to reach the preset rolling temperature range, and uses high-pressure water to descale and remove the surface oxide layer;

[0033] The slab rolling mill performs preliminary rolling on the heated slab to form an intermediate slab;

[0034] Rough rolling mill, rolling the intermediate billet to initially form the basic shape of the rolled piece;

[0035] The intermediate rolling mill continues to roll the workpiece on the basis of rough rolling;

[0036] Finishing mill, which performs final rolling on the rolled product;

[0037] The first visual inspection unit identifies the edge position of the billet and calculates the centerline offset of the billet through 3D contour reconstruction. When the centering deviation exceeds a first threshold, hydraulic correction is triggered and the centering data is transmitted to the roughing mill in real time as a reference for the roughing mill pass alignment;

[0038] The second visual inspection unit uses an image segmentation algorithm to extract the residual area of ​​oxide scale on the surface of the rolled piece in real time. When the residual area detected is greater than the second threshold, secondary descaling is triggered. The width spread is calculated simultaneously. If the deviation between the measured width spread and the theoretical value is greater than the third threshold, the side pressure of the intermediate rolling mill roll is adjusted.

[0039] The third visual inspection unit monitors the temperature distribution of the rolled piece in real time. If uneven temperature distribution is detected, the flow distribution of the cooling water of the finishing mill is adjusted. The unit also scans the transition area between the flange and the waist of the rolled piece. If the measured value of the fillet radius deviates from the designed value by more than a fourth threshold, the side pressure of the vertical roller of the finishing mill is corrected.

[0040] The fourth visual inspection unit generates the flange cross-sectional profile through three-dimensional point cloud reconstruction, calculates the filling rate by comparing it with the theoretical model, and when the filling rate is less than the fifth threshold, calculates the roll gap compensation amount and corrects the reduction amount of the finishing mill in real time; at the same time, it detects the cross-sectional dimension tolerance of the rolled piece. If the flange deviation is greater than the sixth threshold, the rolled piece is sent back to the finishing mill and rolled again until the output flange deviation is within the qualified range.

[0041] The beneficial effects of the present invention are:

[0042] Improve billet processing quality: The first visual inspection unit identifies the billet edge position in real time and calculates the centerline offset. When the centering deviation is greater than the first threshold, hydraulic correction is automatically triggered. The centering data is transmitted to the roughing mill in real time as a hole alignment reference, greatly improving the billet centering accuracy and ensuring accurate hole alignment in the subsequent rolling process, laying the foundation for high-quality rolling.

[0043] Optimizing rough rolling process control: A second visual inspection unit is installed at the roughing mill exit. Using an image segmentation algorithm, it extracts residual scale areas on the workpiece surface in real time, effectively resolving the scale residue issue and ensuring workpiece surface quality. Real-time calculation of width expansion allows for precise control of this expansion, improving workpiece dimensional accuracy.

[0044] Precise control of the intermediate rolling stage: The third visual inspection unit monitors the temperature distribution of the rolled piece in real time to ensure uniform temperature of all parts of the rolled piece, making the deformation of the rolled piece more uniform and helping to improve the plate quality.

[0045] Precisely control finishing quality: 3D point cloud reconstruction is used to generate the flange cross-sectional profile, and the fill rate is calculated by comparing it with a theoretical model to ensure that the flange shape of the rolled piece meets design requirements. Meanwhile, rolled pieces with excessive deviations are returned to the finishing mill for re-rolling until the flange deviation is within the acceptable range. This effectively avoids the production of substandard products and improves production efficiency and product quality.

[0046] Based on deep learning technology, this method and system realize real-time monitoring, intelligent analysis and automatic adjustment of the entire process of embossed steel rolling, reduce human intervention, improve the automation and intelligence level of production, reduce labor intensity, and at the same time improve production efficiency and product quality stability. Compared with the original process, the incidence of surface defects can be reduced by 40% after adopting the method of the present invention. DETAILED DESCRIPTION

[0047] Example 1

[0048] The technical solutions of the present invention are as follows:

[0049] S1. The blank is heated to reach a preset rolling temperature range, and high-pressure water is used to descale and remove the surface oxide layer.

[0050] Providing the required heat to the billet to reach the preset rolling temperature range ensures good plastic deformation during the subsequent rolling process, reduces rolling forces, and improves rolling efficiency and product quality. If the temperature is insufficient, the billet is prone to cracking and rolling resistance; excessive temperature may lead to coarse grains and severe oxidation and burning.

[0051] High-temperature oxidation of metal billets in a heating furnace will form a dense oxide layer. If these oxide layers are not removed, they will directly embed into the surface of the metal during subsequent rolling or forging, causing defects such as pitting, cracks, and delamination in the product, seriously affecting the mechanical properties.

[0052] High-pressure water descaling utilizes high-pressure water jets to rapidly cool and shrink the oxide layer, causing cracks to expand, warp, and peel. This effectively cleans the billet surface, improving the steel's surface quality; reduces roll wear and extends roll life; reduces pickling difficulty and acid consumption, saving production costs; and eliminates uneven metallographic structure and properties caused by the oxide layer, improving overall product quality.

[0053] S2. A first visual inspection unit is set above the conveyor roller between the heating furnace outlet and the slab mill to identify the edge position of the slab and calculate the offset of the slab centerline through three-dimensional contour reconstruction. When the centering deviation is greater than the first threshold, the hydraulic correction is triggered and the centering data is transmitted to the roughing mill in real time as a reference for the roughing mill hole alignment.

[0054] The first visual inspection unit has a dual-view deep learning machine vision function, which can identify the edge position of the billet and reconstruct the three-dimensional contour based on the billet edge position. It calculates the offset of the billet centerline based on the centerline of the roughing mill pass. When the offset Δx is greater than 0.5mm, hydraulic correction is triggered and the hydraulic correction stroke X is calculated:

[0055] X=mΔx,

[0056] Where m is the coefficient for compensating mechanical clearance.

[0057] Due to the inevitable mechanical gap in the rolling line conveyor roller, the hydraulic correction advance amount needs to additionally cover this part of the idle travel, and the hydraulic correction stroke must be greater than the offset; and as the billet temperature rises, it will also expand laterally. The actual width of the heated billet is larger than the cold width, so the advance amount needs to reversely offset this expansion.

[0058] This closed-loop control mechanism ensures that the rolls and the workpiece are always precisely matched, reducing dimensional deviations and surface defects caused by hole misalignment, and laying a quality foundation for subsequent finishing rolling processes.

[0059] 3D profile reconstruction also calculates the cross-sectional asymmetry of the rolled piece. If the asymmetry exceeds a sixth threshold, the piece is returned to the furnace for reheating. This effectively intercepts billets with severe internal defects, preventing rolling mill jams and strip breakage caused by cross-sectional asymmetry, while also reducing material waste and rework costs associated with scrapped products.

[0060] The first visual inspection unit is used to identify the edge position of the billet in real time and calculate the centerline offset. When the centering deviation is greater than the first threshold, hydraulic correction is automatically triggered, and the centering data is transmitted to the roughing mill in real time as a hole alignment reference, which greatly improves the centering accuracy of the billet, ensures accurate hole alignment in the subsequent rolling process, and lays the foundation for high-quality rolling.

[0061] S3. A second visual inspection unit is set up on the outlet side of the roughing mill. The residual area of ​​oxide scale on the surface of the rolled piece is extracted in real time through the image segmentation algorithm. When the residual area is detected to be greater than the second threshold, the secondary descaling is triggered; the width expansion is calculated simultaneously. If the deviation between the measured width expansion and the theoretical value is greater than the third threshold, the side pressure of the vertical roll of the intermediate rolling mill is adjusted.

[0062] In actual application, the second visual inspection unit is located 5m away from the exit of the roughing mill, avoiding the cooling water splashing area, and at a 45° downward angle to the rolling line of the roughing mill. The field of view covers the rolled product. The second visual inspection unit includes a high-speed industrial camera and a laser profile sensor.

[0063] A second visual inspection unit is installed at the roughing mill exit. Using an image segmentation algorithm, it extracts residual scale areas on the workpiece surface in real time, effectively resolving the scale residue issue and ensuring surface quality. Width expansion is calculated in real time, enabling precise control of this expansion and improving workpiece dimensional accuracy.

[0064] Specifically, this can be achieved through the following operations: first, reduce the noise of the rolled piece image through Gaussian filtering, then dynamically segment the scale area based on the Otsu algorithm, divide the number of scale pixels by the total area pixel number, and calculate the scale coverage rate, that is, the residual area. Depending on the composition of the billet, the second threshold is set between 3% and 5%. If it is greater than the second threshold, high-pressure water is triggered for secondary descaling, and the nozzle angle of the secondary descaling is 40°±5°.

[0065] The width expansion is calculated simultaneously. The width expansion represents the difference between the actual width expansion and the theoretical width expansion during the rolling process. It is a key indicator for measuring the accuracy of rolling dimension control. The specific calculation method is as follows:

[0066] ΔW=W 实测 —W 理论 ,

[0067] Where ΔW is the deviation between the measured width and the theoretical value, W 实测 W is the value obtained by the second visual inspection unit through deep learning to measure the actual width of the rolled piece after rough rolling. 理论 It is the width variation of the rolled piece under the ideal rolling state obtained in advance.

[0068] The width expansion at the rough rolling output directly affects the width of the billet at the intermediate rolling entrance. The intermediate rolling vertical roller actively controls the subsequent width expansion through the side pressure. Therefore, if the width expansion is too large, the side pressure of the intermediate rolling mill vertical roller should be adjusted. The specific operation is to calculate the side pressure adjustment amount of the vertical roller:

[0069] ΔS=αΔW+C,

[0070] Among them, ΔS is the side pressure adjustment of the vertical roller, α is the width compensation coefficient, and C is the system error compensation term;

[0071]

[0072] Among them, σ s is the yield strength of the rolled material, β is the equilibrium coefficient, usually taken as 300.

[0073] Furthermore, due to excessive rough rolling reduction, insufficient side pressure margin of the intermediate rolling vertical roll, or low rolling temperature and roll wear resulting in too wide a hole, the width expansion will be too small, which also harms the quality of the convex steel. The harm should be eliminated by reducing the pressure of the intermediate rolling vertical roll.

[0074] S4. A third visual inspection unit is set at the outlet side of the intermediate rolling mill to monitor the temperature distribution of the rolled piece in real time. When uneven temperature distribution is detected, the flow distribution of the cooling water of the finishing mill is adjusted; the transition area between the flange and the waist of the rolled piece is scanned. If the measured value of the fillet radius deviates from the design value and exceeds the fourth threshold, the lateral pressure of the vertical roller of the finishing mill is corrected.

[0075] The third visual inspection unit is comprised of a high-precision infrared camera and a high-speed visible light camera, working in conjunction with a lighting system. The infrared camera, with its highly sensitive temperature detection capabilities, accurately captures the surface temperature of the rolled piece; the high-speed visible light camera captures clear images of the rolled piece's appearance, assisting with subsequent geometric dimensional analysis. The inspection unit is secured by a stable mechanical structure, ensuring precise inspection position and angle throughout the production process, thereby obtaining accurate and reliable data.

[0076] As the workpiece passes through the intermediate mill exit, the third visual inspection unit continuously monitors the workpiece's temperature distribution in real time, rapidly acquiring numerous temperature data points and integrating and analyzing them using advanced image processing algorithms. This creates a thermal map of the workpiece's surface temperature distribution, visually demonstrating temperature differences across various regions.

[0077] Temperature uniformity is a key factor in the stability of the rolling process. If the temperature distribution of the workpiece is uneven, the metal's resistance to deformation will vary from area to area, leading to fluctuations in rolling force. This can cause mill vibration and workpiece deviation, impacting the continuity and stability of production. Real-time temperature distribution monitoring can promptly identify potential problems and provide a basis for subsequent adjustments. Uneven temperatures can also lead to inconsistent deformation in different parts of the workpiece, making it prone to defects such as uneven thickness, wave formation, and warping during subsequent rolling, seriously affecting the mechanical properties and appearance quality of the product. By monitoring temperature distribution, measures can be taken in advance to prevent these defects and improve product qualification rates.

[0078] When the third visual inspection unit detects uneven temperature distribution in the workpiece through data analysis, it quickly transmits the relevant data to the finishing mill, which then controls the cooling water flow in each cooling zone. For hotter areas, the cooling water flow is increased to accelerate heat dissipation and lower the temperature; for cooler areas, the cooling water flow is appropriately reduced to avoid overcooling. During this adjustment process, the system monitors the cooling water flow and workpiece temperature in real time, dynamically adjusting the flow distribution based on this feedback to ensure that the workpiece temperature gradually becomes uniform.

[0079] Specifically, the third visual inspection unit is equipped with an infrared camera to detect the temperature difference of the rolled piece. If the temperature difference between the lowest temperature and the highest temperature of the rolled piece is greater than the maximum temperature difference threshold, the flow rate of cooling water in the high temperature zone corresponding to the highest temperature is increased to Q new :

[0080] Q new =Q0×[1+0.02(ΔT-50)],

[0081] Among them, Q new is the cooling water flow after adjustment, Q0 is the original set cooling water flow, and ΔT is the temperature difference of the rolled piece;

[0082] Increase the rolling speed of the roughing mill to v new :

[0083] v new =v0×[1+0.01(50-ΔT)],

[0084] Among them, v new is the adjusted rolling speed, and v0 is the original set rolling speed.

[0085] The third visual inspection unit monitors the temperature distribution of the rolled piece in real time to ensure that the temperature of each part of the rolled piece is uniform, making the deformation of the rolled piece more uniform, which is conducive to improving the plate shape quality.

[0086] Excessive temperature differences between the head and tail, edges and core of the rolled piece will lead to uneven distribution of metal deformation resistance, and fluctuations in rolling force will further lead to deviations in product thickness and shape. If the area is locally overcooled, cracks will easily occur in the local area, and the grains in the overheated area will be coarse.

[0087] Traditional solutions rely on manual adjustment of cooling water based on experience, resulting in delayed response and inability to accurately locate high-temperature areas. However, actual production requires precise temperature control, which is why a third visual inspection unit is employed. This unit includes an infrared camera that generates a temperature field cloud map in real time, identifying the precise location where the maximum temperature difference threshold has been exceeded.

[0088] The method adopted by the present invention can dynamically adjust the cooling water and the rolling speed, and in actual use, the temperature difference can be controlled within ±15°C.

[0089] Increasing cooling water flow in high-temperature zones reduces hot spots and prevents local overheating. Meanwhile, increasing the roughing mill's rolling speed reduces the dwell time of the workpiece in the roll gap in low-temperature zones, thus preventing excessive temperature drops. Pre-compensation for finishing rolling is implemented, pre-transmitting temperature difference data to the finishing model to optimize the distribution of vertical roll side pressure and detect edge temperature at the rolling exit.

[0090] For example, during production, the third visual inspection unit detected a temperature of 1050°C at the edge of the workpiece and 980°C at the core at the intermediate rolling exit, with a temperature difference of ΔT = 70°C. The specific operation of the present invention involves increasing the cooling water flow rate from 100 L / min to 140 L / min, raising the rolling speed from 2.0 m / s to 2.4 m / s, and adjusting the reduction from 12 mm to 10 mm in advance during the finishing process based on the temperature difference data. The ideal result is a temperature difference reduced to 30°C, crack-free edges on the finished product, and an improvement in thickness tolerance from ±0.5 mm to ±0.2 mm.

[0091] Uniform temperature distribution helps the rolled piece to achieve uniform deformation during finish rolling, reduces dimensional deviation and internal stress concentration caused by temperature differences, thereby improving the dimensional accuracy and mechanical properties of the product and meeting the high quality requirements of different users.

[0092] While monitoring the temperature distribution, the third visual inspection unit uses a high-speed visible light camera to perform high-precision scanning of the transition area between the flange and the waist of the rolled piece to ensure that every subtle geometric feature can be captured.

[0093] Using deep learning edge detection algorithms and 3D reconstruction technology, the system accurately extracts the contour information of the transition zone of the rolled piece and calculates the measured fillet radius. By comparing this with the pre-set design value, the system can accurately determine whether the fillet radius meets the requirements.

[0094] If the measured fillet radius deviates from the designed value by more than a fourth threshold, the third visual inspection unit immediately sends a correction command to the finishing mill control system. Upon receiving the command, the finishing mill control system rapidly calculates the adjustment for the vertical roll side pressure based on a pre-set correction model and algorithm.

[0095] To ensure product dimensional consistency, timely correction of the vertical roll side pressure can quickly correct the deviation of the rolled product corner radius, so that the rolled product can restore its normal geometric shape in the subsequent rolling process, reduce rework and scrap caused by dimensional deviation, improve production efficiency, and reduce production costs. At the same time, it avoids unnecessary wear on equipment such as rolls due to excessive adjustment of rolling parameters, extends equipment service life, and reduces equipment maintenance costs.

[0096] S5. A fourth visual inspection unit is set at the outlet side of the finishing mill. The flange cross-sectional profile is generated through three-dimensional point cloud reconstruction. The filling rate is calculated by comparing with the theoretical model. When the filling rate is less than the fifth threshold, the roll gap compensation amount is calculated and the finishing mill reduction amount is corrected in real time. At the same time, the cross-sectional dimension tolerance of the rolled piece is detected. If the flange deviation is greater than the sixth threshold, the rolled piece is sent back to the finishing mill and rolled again until the output flange deviation is within the qualified range.

[0097] As a comprehensive evaluation index, the flange filling rate can comprehensively consider the influence of various rolling factors, provide a unified reference standard for process optimization and control, and ensure that high-quality rolled products can be obtained under different rolling processes. Therefore, this index is used as the evaluation criterion in the final finishing rolling step.

[0098] Specifically, the operation of correcting the finishing mill reduction is as follows: adjusting the finishing mill reduction to Δh:

[0099]

[0100] Among them, Δh is the total reduction after compensation, h0 is the original set reduction, η is the flange filling rate, and K is the material deformation compensation coefficient.

[0101] Greater reduction requires longer time for deformation to transfer to the core of the workpiece. Reducing the rolling speed can prevent internal cracks caused by uneven deformation between the surface and the core. Therefore, after adjusting the reduction in the finishing mill, the rolling speed is reduced by 10%-15% of the original rolling speed. This speed reduction reduces dynamic rolling force fluctuations, prevents motor overload and roll damage, slows the temperature drop of the workpiece, maintains the final rolling temperature above the phase transition point, improves dimensional accuracy, and reduces scrap rates.

[0102] The flange cross-sectional profile is reconstructed from the 3D point cloud, and the fill rate is calculated by comparing it with the theoretical model to ensure that the flange shape of the rolled piece meets the design requirements. Meanwhile, rolled pieces with excessive deviations are sent back to the finishing mill for re-rolling until the flange deviation is within the acceptable range. This effectively avoids the production of substandard products and improves production efficiency and product quality.

[0103] At the same time, the cross-sectional dimension tolerance of the rolled piece is detected. If the flange deviation is greater than the sixth threshold, the rolled piece is sent back to the finishing mill and rolled again until a rolled piece with a flange deviation within the qualified range is output.

[0104] Flange size deviation directly affects assembly accuracy and structural strength. Immediate rework only increases energy consumption by 5-10%, while the cost of scrapping and re-rolling is 3-5 times higher.

[0105] It is required to have a reverse rolling mode, support the return of rolled pieces and automatic parameter tracing.

[0106] The fourth visual inspection unit identifies scale indentation defects on the rolled piece based on the spectral characteristics of the scale. If the scale indentation defect rate increases, the second threshold and the pressure of the high-pressure water descaling process are dynamically adjusted. Scale indentation can cause microcracks or reduced coating adhesion during subsequent processing. Early control can improve the qualified rate of finished products. Hard scale accelerates roll wear, and reducing its residue can extend the roll surface maintenance cycle and reduce production costs. For example, during production verification, when the fourth visual inspection unit detected a scale indentation defect and the scale indentation defect rate increased from 1.9% to 2.1%, the second threshold was adjusted from 5% to 3%, and the high-pressure water descaling pressure was increased from 18MPa to 22MPa. The scale indentation defect rate was reduced from 2.1% to 1.3%.

[0107] When the second threshold of secondary descaling is reduced, high-pressure water cleaning can be triggered earlier, reducing the total amount of oxide scale entering the finishing mill, thereby reducing the risk of press-in.

[0108] The fourth visual inspection unit can also detect surface defects of rolled products in real time when the flange deviation is within the qualified range. If surface defects exist, the rolled products will be automatically coded and marked for removal.

[0109] Prevents surface-defective rolled products from entering subsequent processes or the market, ensuring consistent product quality. Automatic inkjet marking and rejection establishes a unique "identity" for problem rolled products, facilitating accurate tracing of the causes of defects, such as raw material problems or abnormal rolling parameters, and facilitating targeted process improvements. Furthermore, rejecting defective rolled products prevents them from occupying subsequent processing resources, improving the continuity and efficiency of the production process and reducing rework and scrap costs associated with defective products.

[0110] The deep learning-based embossed steel rolling shape control system includes:

[0111] The heating furnace heats the billet to reach the preset rolling temperature range, and uses high-pressure water to descale and remove the surface oxide layer;

[0112] The slab rolling mill performs preliminary rolling on the heated slab to form an intermediate slab;

[0113] Rough rolling mill, rolling the intermediate billet to initially form the basic shape of the rolled piece;

[0114] The intermediate rolling mill continues to roll the workpiece on the basis of rough rolling;

[0115] Finishing mill, which performs final rolling on the rolled product;

[0116] The first visual inspection unit identifies the edge position of the billet and calculates the centerline offset of the billet through 3D contour reconstruction. When the centering deviation exceeds a first threshold, hydraulic correction is triggered and the centering data is transmitted to the roughing mill in real time as a reference for the roughing mill pass alignment;

[0117] The second visual inspection unit uses an image segmentation algorithm to extract the residual area of ​​oxide scale on the surface of the rolled piece in real time. When the residual area detected is greater than the second threshold, secondary descaling is triggered. The width spread is calculated simultaneously. If the deviation between the measured width spread and the theoretical value is greater than the third threshold, the side pressure of the intermediate rolling mill roll is adjusted.

[0118] The third visual inspection unit monitors the temperature distribution of the rolled piece in real time. If uneven temperature distribution is detected, the flow distribution of the cooling water of the finishing mill is adjusted. The unit also scans the transition area between the flange and the waist of the rolled piece. If the measured value of the fillet radius deviates from the designed value by more than a fourth threshold, the side pressure of the vertical roller of the finishing mill is corrected.

[0119] The fourth visual inspection unit generates the flange cross-sectional profile through three-dimensional point cloud reconstruction, calculates the filling rate by comparing it with the theoretical model, and when the filling rate is less than the fifth threshold, calculates the roll gap compensation amount and corrects the reduction amount of the finishing mill in real time; at the same time, it detects the cross-sectional dimension tolerance of the rolled piece. If the flange deviation is greater than the sixth threshold, the rolled piece is sent back to the finishing mill and rolled again until the output flange deviation is within the qualified range.

[0120] Based on deep learning technology, this method and system realize real-time monitoring, intelligent analysis and automatic adjustment of the entire process of embossed steel rolling, reducing manual intervention, improving the automation and intelligence level of production, reducing labor intensity, and at the same time improving production efficiency and product quality stability.

Claims

1. A method for controlling the shape of convex steel rolling based on deep learning, characterized in that: The steps include: S1. Heating the blank to reach the preset rolling temperature range, and using high-pressure water to descale and remove the surface oxide layer; S2. A first visual inspection unit is installed above the conveyor roller between the heating furnace outlet and the slab mill to identify the edge position of the slab and calculate the offset of the slab centerline through 3D contour reconstruction. When the centering deviation exceeds a first threshold, hydraulic correction is triggered and the centering data is transmitted to the roughing mill in real time as a reference for the roughing mill pass alignment. S3. A second visual inspection unit is installed at the exit of the roughing mill. The residual scale area on the surface of the rolled piece is extracted in real time using an image segmentation algorithm. When the residual area detected is greater than a second threshold, secondary descaling is triggered. The width spread is simultaneously calculated. If the deviation between the measured width spread and the theoretical value is greater than a third threshold, the side pressure of the vertical roll of the intermediate rolling mill is adjusted. S4. A third visual inspection unit is provided at the outlet of the intermediate rolling mill to monitor the temperature distribution of the rolled piece in real time. When uneven temperature distribution is detected, the flow distribution of the cooling water of the finishing mill is adjusted. The transition area between the flange and the waist of the rolled piece is scanned. If the measured value of the fillet radius deviates from the designed value by more than a fourth threshold, the lateral pressure of the vertical roll of the finishing mill is corrected. S5. A fourth visual inspection unit is set at the outlet side of the finishing mill. The flange cross-sectional profile is generated through three-dimensional point cloud reconstruction. The filling rate is calculated by comparing with the theoretical model. When the filling rate is less than the fifth threshold, the roll gap compensation amount is calculated and the finishing mill reduction amount is corrected in real time. At the same time, the cross-sectional dimension tolerance of the rolled piece is detected. If the flange deviation is greater than the sixth threshold, the rolled piece is sent back to the finishing mill and rolled again until the output flange deviation is within the qualified range.

2. The method for controlling the shape of convex steel rolling based on deep learning according to claim 1, characterized in that: The specific operation of adjusting the side pressure of the vertical rolls in S3 is to calculate the side pressure adjustment amount of the vertical rolls: ΔS=αΔW+C, Among them, ΔS is the side pressure adjustment of the vertical roller, α is the width compensation coefficient, ΔW is the deviation between the measured width and the theoretical value, and C is the system error compensation term; Among them, σ s is the yield strength of the rolled material, and β is the equilibrium coefficient.

3. The method for controlling the shape of convex steel rolled products based on deep learning according to claim 1, characterized in that: The third visual inspection unit is also equipped with an infrared camera to detect the temperature difference of the rolled piece. If the temperature difference between the lowest temperature and the highest temperature of the rolled piece is greater than the maximum temperature difference threshold, the flow rate of cooling water in the high temperature area corresponding to the highest temperature is increased to Q new : Q new =Q0×[1+0.02(ΔT-50)], Among them, Q new is the adjusted cooling water flow rate, Q0 is the original set cooling water flow rate, and ΔT is the temperature difference of the rolled piece.

4. The method for controlling the shape of convex steel rolling based on deep learning according to claim 3, characterized in that: At the same time as increasing the cooling water flow rate in the high temperature zone corresponding to the highest temperature, the rolling speed of the roughing mill is increased to v new : v new =v0×[1+0.01(50-ΔT)], Among them, v new is the adjusted rolling speed, and v0 is the original set rolling speed.

5. The method for controlling the shape of convex steel rolling based on deep learning according to claim 1, characterized in that: In S2, the three-dimensional contour reconstruction can also calculate the cross-sectional asymmetry of the rolled piece. If the cross-sectional asymmetry exceeds a sixth threshold, the rolled piece is returned to the heating furnace for reheating.

6. The method for controlling the shape of convex steel rolling based on deep learning according to claim 1, characterized in that: The specific operation of correcting the finishing mill reduction in S5 is to adjust the finishing mill reduction to Δh: Among them, Δh is the total reduction after compensation, h0 is the original set reduction, η is the flange filling rate, and K is the material deformation compensation coefficient.

7. The method for controlling the shape of convex steel rolled products based on deep learning according to claim 4, characterized in that: After the reduction of the finishing mill is adjusted, the rolling speed is reduced by 10%-15% of the original rolling speed.

8. The method for controlling the shape of convex steel rolling based on deep learning according to claim 1, characterized in that: The fourth visual inspection unit can also detect surface defects of rolled products in real time when the flange deviation is within the qualified range. If surface defects exist, the rolled products will be automatically coded and marked for removal.

9. The method for controlling the shape of convex steel rolling based on deep learning according to claim 1, characterized in that: The fourth visual inspection unit identifies scale indentation defects on the rolled piece based on the spectral characteristics of the scale. If the scale indentation defect rate increases, the second threshold and the pressure of high-pressure water descaling are dynamically adjusted.

10. The convex steel rolling plate shape control system based on deep learning is characterized by: include: The heating furnace heats the billet to reach the preset rolling temperature range, and uses high-pressure water to descale and remove the surface oxide layer; The slab rolling mill performs preliminary rolling on the heated slab to form an intermediate slab; Rough rolling mill, rolling the intermediate billet to initially form the basic shape of the rolled piece; The intermediate rolling mill continues to roll the workpiece on the basis of rough rolling; Finishing mill, which performs final rolling on the rolled product; The first visual inspection unit identifies the edge position of the billet and calculates the centerline offset of the billet through 3D contour reconstruction. When the centering deviation exceeds a first threshold, hydraulic correction is triggered and the centering data is transmitted to the roughing mill in real time as a reference for the roughing mill pass alignment; The second visual inspection unit uses an image segmentation algorithm to extract the residual area of ​​oxide scale on the surface of the rolled piece in real time. When the residual area detected is greater than the second threshold, secondary descaling is triggered. The width spread is calculated simultaneously. If the deviation between the measured width spread and the theoretical value is greater than the third threshold, the side pressure of the intermediate rolling mill roll is adjusted. The third visual inspection unit monitors the temperature distribution of the rolled piece in real time. If uneven temperature distribution is detected, the flow distribution of the cooling water of the finishing mill is adjusted. The unit also scans the transition area between the flange and the waist of the rolled piece. If the measured value of the fillet radius deviates from the designed value by more than a fourth threshold, the side pressure of the vertical roller of the finishing mill is corrected. The fourth visual inspection unit generates the flange cross-sectional profile through three-dimensional point cloud reconstruction, calculates the filling rate by comparing it with the theoretical model, and when the filling rate is less than the fifth threshold, calculates the roll gap compensation amount and corrects the reduction amount of the finishing mill in real time; at the same time, it detects the cross-sectional dimension tolerance of the rolled piece. If the flange deviation is greater than the sixth threshold, the rolled piece is sent back to the finishing mill and rolled again until the output flange deviation is within the qualified range.

Images