Ultra-wide IF steel cold buckling defect detection method

Abstract

The invention belongs to the technical field of strip steel production and detection, and particularly relates to an ultra-wide IF steel cold buckling defect detection method which comprises the steps of stable bearing and conveying, surface image acquisition, intelligent defect identification, surface anomaly judgment, potential hazard decision and automatic alarm. The ultra-wide steel stable feeding and deviation correcting module ensures that a steel plate is always in the center of a detection area, the multi-view splicing imaging module realizes full-width and high-resolution dead-corner-free image acquisition of the steel plate, and the intelligent defect identification module accurately extracts a cold buckling defect contour. According to the method, whether the ultra-wide IF steel cold buckling defect is qualified or not is automatically judged on the basis of defect information, accurate and efficient detection of the ultra-wide IF steel cold buckling defect is achieved, risk signals prone to generating the cold buckling defect can be recognized in advance, double guarantees of real-time detection and active prevention and control are formed, and the detection efficiency is improved. The production quality control level and the production stability are comprehensively improved, and the production supervision difficulty is remarkably reduced.

Description

Technical Field

[0001] This invention relates to the field of strip steel production and testing technology, specifically a method for detecting cold bending defects in ultra-wide IF steel. Background Technology

[0002] Ultra-wide IF steel, with its excellent deep-drawing performance and formability, is widely used in high-end manufacturing fields such as automobiles and home appliances. Its production requires key processes such as degreasing, annealing, and leveling through continuous annealing units. However, due to factors such as the large width of the steel plate, the fluctuation of roll crown in the final cooling section and the over-aging section, uneven tension in the furnace, and temperature difference in the transverse plate, it is very easy to produce cold warping defects, which manifest as local uneven deformation on the surface. This not only leads to product downgrading, but in severe cases, it can also cause strip breakage and shutdown, resulting in significant economic losses. Therefore, accurate detection and risk control of cold warping defects in ultra-wide IF steel has become a core requirement for the quality control of strip steel production.

[0003] Currently, there are existing detection and control technologies for strip steel cold bending in the industry. For example, Chinese invention patent with publication number CN114990293A discloses an automatic control method for strip bending in the cold section of a vertical annealing furnace. This technology involves setting up high-definition cameras at key locations such as the rapid cooling section and OA cooling section of the vertical annealing furnace, and combining them with a strip steel bending visual automatic recognition system and an artificial intelligence main control system to identify strip steel bending and severe deviation from the roll in real time. Based on the deviation amount and the bending section, countermeasures are taken to reduce the risk of strip breakage caused by bending to a certain extent.

[0004] However, when the above-mentioned invention is applied to ultra-wide IF steel, it is difficult to accurately identify cold warping defects and reasonably assess the surface quality of ultra-wide IF steel. It can only be passively dealt with by stopping the machine or adjusting the process parameters after the warping defect occurs. It does not connect with the key operating parameters of the continuous annealing unit and build a hidden danger early warning mechanism, and cannot form a dual guarantee of real-time detection and active prevention and control. This is not conducive to comprehensively improving the production quality control level and production stability of ultra-wide IF steel. Therefore, a solution is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a method for detecting cold bending defects in ultra-wide IF steel, so as to solve the technical defects mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting cold-curved defects in ultra-wide IF steel, comprising the following steps: Step 1: The ultra-wide steel stable feeding and correction module provides stable load-bearing feeding and automatic and precise correction for ultra-wide IF steel, ensuring the positional stability of the steel plate in the detection area. Step 2: The multi-view stitching imaging module adopts a multi-camera stitching design to achieve full coverage and high-resolution acquisition of ultra-wide IF steel surface images; Step 3: The intelligent defect recognition module uses a customized algorithm combination to achieve image noise removal and accurate extraction of defect areas, and simultaneously outputs the defect information of the cold-rolled bend defect area; Step 4: The surface anomaly automatic judgment module judges the surface anomalies of the steel plate based on the cold bending defect information, and generates a surface unqualified signal or a surface qualified signal accordingly. Step 5: When a surface non-compliance signal is generated, the touch monitoring and early warning terminal issues a corresponding warning.

[0007] Furthermore, in step one, the ultra-wide steel stabilization and correction module adjusts the rotation speed of the precision conveying roller group according to the preset cycle time of the ultra-wide IF steel production line through the frequency conversion drive component, so that the conveying speed is stabilized within the range of 0.5 to 2 m / s, and the speed fluctuation is strictly controlled within ±0.01 m / s; the roller surface of the precision conveying roller group is precision ground and chrome plated, and the flatness is ≤0.02 mm. Meanwhile, the laser displacement sensing component collects position data of both sides of the steel plate in real time at a sampling frequency of 100Hz and transmits the data to the built-in control unit in real time. When the steel plate offset exceeds ±3mm, the control unit immediately sends a command to the servo correction execution component. The servo correction execution component drives the guide roller to finely adjust the angle through a high-precision servo motor to complete the correction action and make the steel plate return to the center axis of the detection area.

[0008] Furthermore, the multi-view stitching imaging module controls 3 to 6 line scan cameras to be stitched together along the width of the steel plate, with each camera having a pixel count ≥ 4096 and a shooting frame rate ≥ 500fps. The shooting areas of adjacent cameras are set with an overlap rate of 5% to 10%. The camera shooting frame rate is automatically adjusted according to the transmission speed signal.

[0009] Furthermore, during the image acquisition process, the images from adjacent cameras are stitched together in real time using an overlapping area feature matching algorithm to eliminate stitching gaps. The stitching error is ≤ ±0.1mm, forming complete surface image data covering the entire width of the steel plate. This data is then transmitted in real time to the defect intelligent recognition module via the EtherCAT protocol.

[0010] Furthermore, the specific operation process of the defect intelligent identification module is as follows: The system receives complete image data transmitted from the multi-view stitching imaging module and runs an image enhancement and denoising sub-algorithm through the industrial computing component. First, an improved 3×3 median filtering algorithm is used to smooth the image and remove interference noise. Then, an adaptive histogram equalization algorithm is used to enhance the grayscale contrast between the defect area and the normal surface. Subsequently, a defect area extraction sub-algorithm is run, which uses an adaptive Otsu threshold segmentation algorithm to automatically determine the grayscale threshold and segment the image into defect areas and normal surfaces. Then, the Canny edge detection algorithm is used to extract the complete contour of the defect area, and the defect information of the cold-curved defect area is transmitted to the surface anomaly automatic judgment module.

[0011] Furthermore, the determination and analysis process of the automatic surface anomaly determination module is as follows: The area of ​​the corresponding cold-rolled bend defect region is obtained and marked as the first feature value. The maximum and average values ​​of the actual concave-convex height difference of the corresponding cold-rolled bend defect region are marked as the second feature value and the third feature value, respectively. The abnormal feature coefficient is calculated by weighted summation of the first feature value, the second feature value and the third feature value. The abnormal feature coefficient is compared with the preset abnormal feature coefficient threshold. If there is a cold-rolled bend defect region with an abnormal feature coefficient exceeding the preset abnormal feature coefficient threshold, a surface unqualified signal is generated. If there is no cold-rolled bend defect region with an abnormal feature coefficient exceeding the preset abnormal feature coefficient threshold, a surface qualified signal is generated.

[0012] Furthermore, the surface anomaly automatic judgment module is connected to the potential hazard decision module. The surface anomaly automatic judgment module sends the surface qualified signal or the surface unqualified signal to the potential hazard decision module. If no surface unqualified signal is received within a unit time, the potential hazard of the continuous annealing unit is analyzed by the potential hazard decision module. The analysis is used to determine whether to output a signal that is prone to cold warping defects. When the touch monitoring and early warning terminal receives the signal that is prone to cold warping defects, it issues a corresponding early warning.

[0013] Furthermore, the specific analysis process of the potential hazard decision-making module is as follows: The operating parameters that are highly correlated with the cold bend defect during the operation of the continuous annealing unit are obtained and marked as highly correlated parameters i, where i is a natural number greater than 1; real-time data of highly correlated parameters i are collected and compared with the corresponding preset data requirements. If the real-time data does not meet the corresponding preset data requirements, it is determined that highly correlated parameters i are in a potential danger state. The total duration of high-correlation parameter i in a potential hazard state within a unit of time is obtained and marked as the high-correlation hazard time value. The average deviation and maximum deviation of the data when high-correlation parameter i is in a potential hazard state within a unit of time are marked as the high-correlation deviation value and the high-correlation deviation amplitude value, respectively. The parameter unfavorable coefficient is calculated by weighted summation of the high-correlation hazard time value, the high-correlation deviation value, and the high-correlation deviation amplitude value. The parameter unfavorable coefficient is compared with the corresponding preset parameter unfavorable coefficient threshold. If the parameter unfavorable coefficient exceeds the preset parameter unfavorable coefficient threshold, the high-correlation parameter i is marked as a defect cause object. If there is a defect-causing object within a unit of time, a signal indicating a tendency to generate cold-blown defects will be output.

[0014] Furthermore, if there is no defect-causing object within a unit of time, the unfavorable coefficient of the highly correlated parameter i is calculated by the ratio of the corresponding preset unfavorable coefficient threshold to obtain the unfavorable prediction value. Each highly correlated parameter is pre-set to correspond to a set of preset influence weight values. The unfavorable prediction value of the highly correlated parameter i is multiplied by the corresponding preset influence weight value to obtain the unfavorable risk value. The unfavorable risk values ​​of all highly correlated parameters are summed to obtain the potential hazard decision value. The potential hazard decision value is numerically compared with the preset potential hazard decision threshold. If the potential hazard decision value exceeds the preset potential hazard decision threshold, a signal indicating a tendency to generate cold-blown defects is output.

[0015] Compared with the prior art, the beneficial effects of the present invention are: In this invention, the combination of an ultra-wide steel stabilization and correction module and a multi-view splicing imaging module enables the acquisition of high-resolution, blind-angle-free images of the steel plate across its full width. The intelligent defect identification module accurately extracts the contour of the cold-rolled bend defect and automatically determines whether the cold-rolled bend defect of the ultra-wide IF steel is qualified based on the defect information, thereby achieving accurate and efficient detection of cold-rolled bend defects in ultra-wide IF steel.

[0016] In this invention, the potential hidden danger decision-making module further expands the control dimensions, identifies risk signals that are prone to cold bending defects in advance, and upgrades control from passively responding to defects to actively warning of hidden dangers. This makes it easier for managers to quickly locate problems and take control measures, greatly reducing the difficulty of production supervision and further ensuring the production quality and continuous stability of ultra-wide IF steel. Attached Figure Description

[0017] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: As Figure 1-2 As shown, the present invention proposes a method for detecting cold-curved defects in ultra-wide IF steel, comprising the following steps: Step 1: The ultra-wide steel stable feeding and correction module provides stable load-bearing feeding and automatic and precise correction of ultra-wide IF steel, ensuring the positional stability of the steel plate in the detection area and providing the basic conditions for subsequent imaging and detection. Specifically, the ultra-wide steel conveying and straightening module uses a frequency converter to adjust the rotation speed of the precision conveying rollers according to the preset cycle time of the ultra-wide IF steel production line, so that the conveying speed is stabilized within the range of 0.5 to 2 m / s, and the speed fluctuation is strictly controlled within ±0.01 m / s to avoid blurring of images due to speed fluctuations. The roller surface of the precision conveying rollers is precision ground and chrome-plated, with a flatness of ≤0.02 mm, which not only prevents scratches on the steel plate surface, but also reduces the vibration of the steel plate caused by unevenness of the roller surface. Meanwhile, the laser displacement sensing component collects position data of both sides of the steel plate in real time at a sampling frequency of 100Hz and transmits the data to the built-in control unit in real time. When the steel plate offset exceeds ±3mm, the control unit immediately sends a command to the servo correction execution component. The servo correction execution component drives the guide roller to fine adjust the angle through a high-precision servo motor and completes the correction action within 2 seconds, so that the steel plate returns to the center axis of the detection area. The correction accuracy is ≤±2mm, ensuring that the subsequent imaging module can accurately cover the full width range of the steel plate.

[0020] Step 2: The multi-view stitching imaging module adopts a multi-camera stitching design to achieve full coverage and high-resolution acquisition of ultra-wide IF steel surface images, eliminate stitching gaps, and ensure the accuracy of image data. Specifically, the multi-view stitching imaging module controls 3 to 6 line scan cameras to be stitched together along the width of the steel plate, based on the width of the steel plate. Each camera has a resolution of ≥4096 pixels and a shooting frame rate of ≥500fps. The shooting areas of adjacent cameras are set with an overlap rate of 5% to 10%. The camera shooting frame rate is automatically adjusted according to the conveying speed signal to ensure that there is no image ghosting or missed shots when the steel plate is conveyed at high speed (up to 2m / s).

[0021] Furthermore, during the image acquisition process, the images from adjacent cameras are stitched together in real time using an overlapping area feature matching algorithm to eliminate stitching gaps. The stitching error is ≤ ±0.1mm, forming complete surface image data covering the entire width of the steel plate. This data is then transmitted in real time to the defect intelligent recognition module via the EtherCAT protocol.

[0022] Step 3: The intelligent defect recognition module uses a customized algorithm combination to achieve image noise removal and accurate extraction of defect areas, simultaneously outputting defect information of the cold-rolled curved defect area; the specific operation process is as follows: After receiving the complete image data transmitted by the multi-view stitching imaging module, the image enhancement and denoising sub-algorithm is first run through the industrial computing component (configured with CPU≥i7, memory≥16G, GPU≥RTX3060). The improved 3×3 median filtering algorithm is first used to smooth the image and remove interference noise such as dust and light fluctuations in the production environment. Then, the adaptive histogram equalization algorithm is used to enhance the grayscale contrast between the defect area and the normal surface, making the defect outline clearer. Subsequently, the defect region extraction sub-algorithm is run, and the adaptive Otsu threshold segmentation algorithm is used to automatically determine the grayscale threshold, segmenting the image into foreground (defect region) and background (normal surface). Then, the complete contour of the defect region is extracted by the Canny edge detection algorithm to eliminate small-area noise interference, and the defect information of the cold-curved defect region (such as the actual area of ​​the defect region and the height difference between concave and convex areas) is transmitted to the surface anomaly automatic judgment module.

[0023] Step 4: The automatic surface anomaly detection module determines surface anomalies of the steel plate based on the cold bending defect information, and generates surface non-conforming or conforming signals accordingly. The detection and analysis process is as follows: The area of ​​the corresponding cold-rolled bend defect region is obtained and marked as the first feature value. The maximum and average values ​​of the actual concave-convex height difference of the corresponding cold-rolled bend defect region are marked as the second feature value and the third feature value, respectively. The abnormal feature coefficient is obtained by weighted summation of the first feature value, the second feature value and the third feature value. That is, a preset weight coefficient is assigned to the first feature value, the second feature value and the third feature value, respectively. The first feature value, the second feature value and the third feature value are multiplied by the corresponding preset weight coefficient, and the sum of the three sets of product results is marked as the abnormal feature coefficient. It should be noted that the larger the value of the abnormal feature coefficient, the more severe the defect in the corresponding cold-rolled bend defect area. The abnormal feature coefficient is compared with the preset abnormal feature coefficient threshold. If there is a cold-rolled bend defect area with an abnormal feature coefficient exceeding the preset abnormal feature coefficient threshold, it indicates that the quality of the steel plate surface being inspected is poor, and a surface unqualified signal is generated. If there is no cold-rolled bend defect area with an abnormal feature coefficient exceeding the preset abnormal feature coefficient threshold, it indicates that the quality of the steel plate surface being inspected is good, and a surface qualified signal is generated.

[0024] Step 5: When a surface non-conformity signal is generated, the touch monitoring and early warning terminal issues a corresponding warning to remind production management personnel to investigate and analyze the cause and quickly make reasonable improvement measures, which helps to ensure the quality of the processed products and significantly reduces the difficulty of production and processing supervision.

[0025] Example 2: Figure 1-2 As shown, the difference between this embodiment and embodiment one is that the surface anomaly automatic judgment module is communicatively connected to the potential hazard decision module. The surface anomaly automatic judgment module sends the surface qualified signal or the surface unqualified signal to the potential hazard decision module. If the surface unqualified signal is not received within a unit time, the potential hazard of the continuous annealing unit is analyzed by the potential hazard decision module. The analysis is used to determine whether to output a signal that is prone to cold warping defects. When the touch-screen monitoring and early warning terminal receives a signal that is prone to cold warping defects, it issues a corresponding warning. It can identify risk signals that are prone to cold warping defects in advance, upgrading the management from passively responding to defects to actively warning of potential hazards. This reminds production managers to take timely control measures, which helps to ensure the continuous and stable operation of the continuous annealing unit and effectively avoids the occurrence of cold warping defects in ultra-wide IF steel.

[0026] It should be noted that continuous annealing units are units that integrate the degreasing, annealing, and leveling processes of cold-rolled strip steel into a single production line for continuous production, which greatly shortens the annealing cycle compared to over-annealing. When producing ultra-wide and ultra-deep drawing steel with a width ≥1800mm, continuous annealing units are prone to warping defects in the final cooling section and over-aging section, which can lead to product downgrading, strip breakage, shutdown, and product delivery disruptions, resulting in huge economic losses. From the perspective of defect causes, the key to controlling defect formation lies in controlling the roll crown, strip strength, furnace tension, and transverse plate temperature difference in the final cooling section and over-aging section. The elimination of cold warping defects can be achieved through the coordination of various parameters.

[0027] The specific analysis process of the potential hidden danger decision module is as follows: obtain the operating parameters that are highly correlated with the cold bend defect during the operation of the continuous annealing unit, mark the corresponding operating parameters as highly correlated parameters i, and i is a natural number greater than 1; collect the real-time data of highly correlated parameters i, compare the real-time data with the corresponding preset data requirements, and if the real-time data does not meet the corresponding preset data requirements, then determine that highly correlated parameters i is in a hidden danger state. Obtain the total duration of high correlation parameter i in the hidden danger state within a unit time and mark it as the high correlation hidden danger time value. Also mark the average deviation and maximum deviation of the data when high correlation parameter i is in the hidden danger state within a unit time as the high correlation deviation value and the high correlation deviation value, respectively. The parameter unfavorable coefficient is calculated by weighting and summing the high-correlation time value, high-correlation partial table value, and high-correlation partial amplitude value. Specifically, each of the high-correlation time value, high-correlation partial table value, and high-correlation partial amplitude value is assigned a corresponding preset weight coefficient, and each of these values ​​is multiplied by its corresponding preset weight coefficient. The sum of the three products is then marked as the parameter unfavorable coefficient. Furthermore, the larger the parameter unfavorable coefficient, the more likely the current parameter status of the high-correlation parameter i is to lead to cold-blown defects. The parameter unfavorable coefficient is compared with the corresponding preset parameter unfavorable coefficient threshold. If the parameter unfavorable coefficient exceeds the preset parameter unfavorable coefficient threshold, it indicates that the current parameter status of the highly correlated parameter i is likely to cause cold-rolled warp defects. In this case, the highly correlated parameter i is marked as a defect cause object. If there is a defect cause object within a unit of time, it indicates that the current production and processing quality of the continuous annealing unit is high and cold-rolled warp defects are likely to occur. In this case, a signal indicating that cold-rolled warp defects are likely to occur is output.

[0028] Furthermore, if there is no defect-causing object within a unit of time, the unfavorable coefficient of the highly correlated parameter i is calculated by the ratio of the corresponding preset unfavorable coefficient threshold to obtain the unfavorable prediction value. Each highly correlated parameter is pre-set to correspond to a set of preset influence weight values ​​with values ​​greater than zero. Moreover, the stronger the influence correlation of the highly correlated parameter i, the larger the value of the preset influence weight value that matches it. The unfavorable prediction value of the highly correlated parameter i is multiplied by the corresponding preset influence weight value to obtain the unfavorable risk value. The potential hazard decision value is obtained by summing the non-advantageous risk values ​​of all highly correlated parameters. The potential hazard decision value is then compared with the preset potential hazard decision threshold. If the potential hazard decision value exceeds the preset potential hazard decision threshold, it indicates that the current production and processing quality of the continuous annealing unit is highly vulnerable to cold warping defects. In this case, a signal indicating that cold warping defects are likely to occur is output.

[0029] The working principle of this invention is as follows: During use, the ultra-wide steel plate is stably conveyed by the ultra-wide steel plate stabilization and correction module to ensure that the steel plate is always centered in the detection area. The multi-view splicing imaging module realizes full-width, high-resolution, blind-angle-free image acquisition of the steel plate. The intelligent defect identification module accurately extracts the contour of the cold-rolled bend defect and outputs accurate defect information in real time. The surface anomaly automatic judgment module objectively quantifies the severity of the defect, realizing the automated judgment of whether the ultra-wide IF steel cold-rolled bend defect is qualified or not. This replaces the error of manual subjective judgment, improves the consistency and fairness of the judgment results, realizes the accurate and efficient detection of ultra-wide IF steel cold-rolled bend defects, comprehensively improves its production quality control level and production stability, and significantly reduces the difficulty of production supervision.

[0030] In this invention, the threshold, preset value, or preset range settings are for result comparison and analysis to determine whether the result is good or bad. The magnitude of these values ​​is determined by a combination of large-scale model analysis of sample data and human experience, and can also be appropriately adjusted based on seasonal or common-sense influence conditions. Similarly, the preset weight coefficients and influence factors are assigned specific values ​​based on the magnitude of each parameter's influence on the result, ultimately reflecting the impact on the result. These settings are also determined by a combination of large-scale model analysis of sample data and human experience, and can also be appropriately adjusted based on seasonal or common-sense influence conditions.

[0031] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, enabling those skilled in the art to better understand and utilize it. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for detecting cold-bending defects in ultra-wide IF steel, characterized in that, Includes the following steps: Step 1: The ultra-wide steel stable feeding and correction module provides stable load-bearing feeding and automatic and precise correction for ultra-wide IF steel, ensuring the positional stability of the steel plate in the detection area. Step 2: The multi-view stitching imaging module adopts a multi-camera stitching design to achieve full coverage and high-resolution acquisition of ultra-wide IF steel surface images; Step 3: The intelligent defect recognition module uses a customized algorithm combination to achieve image noise removal and accurate extraction of defect areas, and simultaneously outputs the defect information of the cold-rolled bend defect area; Step 4: The surface anomaly automatic judgment module judges the surface anomalies of the steel plate based on the cold bending defect information, and generates a surface unqualified signal or a surface qualified signal accordingly. Step 5: When a surface non-compliance signal is generated, the touch monitoring and early warning terminal issues a corresponding warning.

2. The method for detecting cold-rolled bend defects in ultra-wide IF steel according to claim 1, characterized in that, In step one, the ultra-wide steel stabilization and correction module adjusts the speed of the precision conveying roller group according to the preset cycle of the ultra-wide IF steel production line through the frequency conversion drive component, and the roller surface of the precision conveying roller group is precision ground and chrome plated. Meanwhile, the laser displacement sensing component collects position data of both sides of the steel plate in real time at a sampling frequency of 100Hz and transmits the data to the built-in control unit in real time. When the steel plate offset exceeds ±3mm, the control unit immediately sends a command to the servo correction execution component. The servo correction execution component drives the guide roller to finely adjust the angle through a high-precision servo motor to complete the correction action and make the steel plate return to the center axis of the detection area.

3. The method for detecting cold-rolled bend defects in ultra-wide IF steel according to claim 1, characterized in that, The multi-view stitching imaging module controls 3 to 6 line scan cameras to be stitched together along the width of the steel plate, based on the width of the steel plate. Each camera has a resolution of ≥4096 pixels and a shooting frame rate of ≥500fps. The shooting areas of adjacent cameras are set with an overlap rate of 5% to 10%. The camera shooting frame rate is automatically adjusted according to the transmission speed signal.

4. The method for detecting cold-rolled bend defects in ultra-wide IF steel according to claim 3, characterized in that, During image acquisition, images from adjacent cameras are stitched together in real time using an overlapping region feature matching algorithm to eliminate stitching breaks and form complete surface image data covering the entire width of the steel plate. This data is then transmitted in real time to the defect intelligent identification module via the EtherCAT protocol.

5. The method for detecting cold-rolled bend defects in ultra-wide IF steel according to claim 4, characterized in that, The specific operation process of the defect intelligent identification module is as follows: The system receives complete image data transmitted from the multi-view stitching imaging module, runs an image enhancement and denoising sub-algorithm through the industrial computing component, first uses a filtering algorithm to smooth the image, and then uses an adaptive histogram equalization algorithm to enhance the grayscale contrast between the defect area and the normal surface; then it runs a defect area extraction sub-algorithm, uses an adaptive Otsu threshold segmentation algorithm to automatically determine the grayscale threshold, segments the image into defect areas and normal surfaces, and then uses the Canny edge detection algorithm to extract the complete contour of the defect area, and transmits the defect information of the cold-curved defect area to the surface anomaly automatic judgment module.

6. The method for detecting cold-rolled bend defects in ultra-wide IF steel according to claim 5, characterized in that, The judgment and analysis process of the automatic surface anomaly judgment module is as follows: The area of ​​the corresponding cold-curved defect region is obtained and marked as the first feature value. The maximum and average values ​​of the actual concave-convex height difference of the corresponding cold-curved defect region are marked as the second feature value and the third feature value, respectively. The abnormal feature coefficient is calculated by weighted summation of the first feature value, the second feature value and the third feature value. If there is a cold-curved defect region with an abnormal feature coefficient exceeding the preset abnormal feature coefficient threshold, a surface unqualified signal is generated; otherwise, a surface qualified signal is generated.

7. The method for detecting cold-rolled bend defects in ultra-wide IF steel according to claim 6, characterized in that, The surface anomaly automatic judgment module communicates with the potential hazard decision module. If the potential hazard decision module does not receive a surface non-conformity signal within a unit of time, it will analyze the potential operational hazards of the continuous annealing unit. The analysis will determine whether to output a signal that is prone to cold warping defects. When the touch monitoring and early warning terminal receives the signal that is prone to cold warping defects, it will issue a corresponding early warning.

8. The method for detecting cold-rolled bend defects in ultra-wide IF steel according to claim 7, characterized in that, The specific analysis process of the potential hazard decision-making module is as follows: The operating parameters that are highly correlated with the cold-rolled bend defect during the operation of the continuous annealing unit are obtained, and the corresponding operating parameters are marked as highly correlated parameters i, where i is a natural number greater than 1; if there is a defect-causing object within a unit of time, a signal that is prone to cold-rolled bend defect is output.

9. A method for detecting cold-rolled bend defects in ultra-wide IF steel according to claim 8, characterized in that, If there is no defect-causing object within a unit of time, the non-advantageous risk values ​​of all highly correlated parameters are summed to obtain the potential hazard decision value. If the potential hazard decision value exceeds the preset potential hazard decision threshold, a signal that is prone to cold-blown defects is output.

Images