Integrated system and method for pretreatment of steel coils at the inlet of pickling mill

Abstract

An integrated system and method for pre-treating steel coils at the inlet of a pickling mill relate to the technical field of pickling mills. This system addresses the problem that most of the preparatory work for existing pickling mills requires manual operation and can only address a single issue, which is time-consuming and labor-intensive. The integrated system includes: a strapping unit that identifies the strap head and removes the straps, and sends a disassembly completion signal to a coil status information detection unit after the strapping is completed; the coil status information detection unit obtains the current coil status information; and sends the current coil status information to an end face defect detection unit; the end face defect detection unit detects coil end face quality defects and measures thickness based on the current coil status information, and sends the detection information to a code recognition unit; the code recognition unit confirms the status of the defective coil and identifies the image containing the code, detects and recognizes the text in the code image, and compares the recognized information with the ID read by the host computer. This system is used in the field of metallurgy.

Description

Technical Field

[0001] The present invention relates to the technical field of pickling mills, and in particular to an integrated pretreatment system for steel coils at an inlet of a pickling mill. Background Art

[0002] Handan Iron and Steel's cold-rolling and pickling mill produces approximately 170,000 tons of raw material per month, producing approximately 8,500 coils. Each coil has an average of two straps, and operators need to remove approximately 17,000 straps per month. The straps are typically 30 mm wide and 2 mm thick, making manual shearing very labor-intensive. Due to the high binding force exceeding 30,000 N, operators can be easily injured by the straps if they are not positioned properly during shearing, making the shearing process highly dangerous. Furthermore, the straps, especially those that are long, can easily fall onto the transport tracks, making the removal process extremely dangerous. Furthermore, the straps cannot be reused after manual removal and must be disposed of. The large number of straps and their irregular storage make on-site waste disposal difficult, impacting the site environment and increasing production costs.

[0003] The steel coil information identification and surface defect inspection of the walking beam are manually confirmed by on-site operators. During the confirmation, they need to pass through the walking beam operation area, lean over the walking beam and cross-operate with the running overhead crane. Operators are exposed to safety hazards such as being squeezed, falling, and being hit by objects. Manual inspection has subjective judgment factors, and there is a possibility of misjudgment or missed inspection, which affects the stable operation of the production line.

[0004] Based on preliminary research, it has been revealed that the technology currently used to detect defects in steel coil end surfaces cannot meet the requirements of current production lines. The challenge is to ensure accuracy while also taking into account the characteristics of hot-rolled coils and significantly improving detection speed to meet the high-speed production cycle. Furthermore, the existing cold rolling mill's pickling production line is imperfect and can only address and resolve single issues, which is time-consuming and labor-intensive. Summary of the Invention

[0005] The present invention addresses the problem that most of the preparatory work for the existing pickling mill requires manual operation and can only handle and solve a single problem, which is time-consuming and labor-intensive. An integrated system for pre-treatment of steel coils at the entrance of the pickling mill is proposed. The integrated system includes:

[0006] Unbundling unit, coil status information detection unit, end surface defect detection unit and inkjet code recognition unit;

[0007] The strapping unit is used to identify the strapping head and remove the strapping, and sends a disassembly completion signal to the coil status information detection unit after the strapping is completed;

[0008] The coil status information detection unit is used to receive the disassembly completion signal and obtain the current coil status information, including: coil width measurement, coil temperature measurement, coil collapse detection, and lead position determination; and send the current coil status information to the end surface defect detection unit;

[0009] The end surface defect detection unit is used to receive the status information of the current steel coil, perform steel coil end surface quality defect detection and thickness measurement based on the current steel coil status information, and send the detection information to the inkjet coding recognition unit; the steel coil end surface quality defect detection includes inner diameter detection, tower shape detection, end surface crack detection, loose coil detection, flat coil detection and staggered layer detection;

[0010] The inkjet coding recognition unit is used to receive defect information from the end face defect detection unit, confirm the status of the defective steel coil and determine the picture where the inkjet coding is located, detect and recognize the text of the inkjet coding picture, and compare the recognized information with the ID read by the host computer.

[0011] Furthermore, a preferred embodiment is proposed, wherein the strapping unit dismantles the strapping by an industrial robot, comprising:

[0012] The industrial robot is set on a slide, which drives the robot to move towards the steel coil until it reaches the preset optimal distance;

[0013] The industrial robot's built-in sensors monitor the distance to the steel coil in real time, ensuring it stops at the optimal distance;

[0014] The industrial robot moves the debundling head and laser ranging sensor directly above the steel coil. The laser sensor detects the distance between the debundling head and the surface of the steel coil in real time.

[0015] When the distance detected by the laser sensor reaches the optimal distance, the industrial robot stops moving downward;

[0016] The industrial robot carries the debundling head and moves horizontally along the axial direction of the steel coil, and detects the position of the strapping;

[0017] If the strap is successfully detected, the industrial robot moves down with the debundling head. The controller controls the debundling head to lightly press on the surface of the steel coil to ensure that the strap is in the middle of the fixed shear blade. The fixed shear blade is pushed by the cylinder to scoop up and clamp the strap. The movable shear blade then cuts the strap. If the strap is not cut successfully, an alarm is issued.

[0018] The slide drives the industrial robot back to its initial position, and the industrial robot clamps the cut strap and sends it to the strap reel; the reel curls and compresses the strap into a rolled disc, and the strapping motor is released, and the curled strap falls directly into the waste hopper.

[0019] Furthermore, a preferred embodiment is proposed, wherein the coil status information detection unit acquires the coil status information through a laser sensor, an infrared temperature sensor, and an intelligent camera;

[0020] Two sets of high-precision laser sensors are used to shoot at the two ends of the steel coil. An adjustable device is used to ensure that the two laser sensors are aligned in both vertical and horizontal positions. The distance between the two sensors is calculated by subtracting the distance between the steel coil ends collected by the sensors to obtain the steel coil width data.

[0021] The infrared temperature sensor is used to collect the single-point temperature signal on the end face of the steel coil and obtain the temperature value of the steel coil;

[0022] The inner coil of the steel coil is photographed and analyzed by a smart camera, the collected outline dimensions of the inner coil of the steel coil are processed, the minimum diameter of the inner ring of the steel coil is calculated, fitted and compared with the standard diameter, and an appropriate deviation value is set. If it exceeds the specified range, an alarm prompt will be given.

[0023] Furthermore, a preferred method is proposed, wherein the detection of cracks on both end faces includes:

[0024] The image information of the steel coil is collected by an optical camera, and the point cloud information of the local end surface is collected by a structured light detector;

[0025] De-noise and grid-align the collected point cloud information to obtain a point cloud image;

[0026] Convert the point cloud image into a grayscale depth map;

[0027] Perform instantiation segmentation on the depth map and perform regional separation on the parts whose segmentation size is larger than the set defect size;

[0028] The defect recognition model is trained based on the image data with the defect locations marked using the defect detection model;

[0029] Through data training, the final model parameters are determined, and the separated suspected defect areas are tested to obtain the final results.

[0030] Furthermore, a preferred embodiment is proposed, wherein the defect detection model comprises: an input end, a backbone network, a neck module, and a prediction module;

[0031] After the input end reads the color image or grayscale image, it performs Mosaic data enhancement, cmBN, and SAT self-adversarial training; the backbone network encodes the input image through CSPDarknet53, Mish activation function and Dropblock; the neck module pre-processes the features generated by the backbone network and strengthens the features; the prediction module prediction uses the loss function CIOU_Loss for training, and the prediction box screening uses DIOU_nms to screen and predict the final defects.

[0032] Furthermore, a preferred embodiment is proposed, wherein the steel coil loose coil detection includes:

[0033] Collect steel coil image information;

[0034] Set a threshold based on the depth value to remove the foreground and background;

[0035] The plane contour of the steel coil is obtained by detecting the contour detection algorithm;

[0036] By comparing the distances between the remaining detected contours and the contour with the largest area, a threshold is set to determine whether it is a loose coil; if it is determined to be a loose coil, the depth value of the loose coil and the depth value of the steel coil plane are detected to see if they exceed a threshold. If so, it is determined to be a misjudgment caused by the tower shape, otherwise it is determined to be a loose coil.

[0037] The inkjet code recognition unit consists of a main inkjet code recognition camera and a secondary inkjet code recognition camera. The main inkjet code recognition camera faces the steel coil and is used to collect the inkjet code on the right end face of the steel coil; the secondary inkjet code recognition camera collects the inkjet code image of the left end face of the saddle at an inclined perspective; and the inkjet code number is recognized and compared through image enhancement, correction and OCR technology.

[0038] Based on the same inventive concept, the present invention also proposes a method for pre-treating steel coils at the inlet of a pickling mill, the method comprising:

[0039] The strapping unit dismantles the strapping and sends a completion signal to the coil status information detection unit when the dismantling is completed;

[0040] The coil status information detection unit receives the completion signal and obtains the coil status information and sends it to the end surface defect detection unit;

[0041] The end surface defect detection unit determines whether the steel coil has defects based on the steel coil status information and sends the detection results to the inkjet coding recognition unit;

[0042] The coding unit obtains coding information based on the detection results and the optical camera, and feeds it back to the host computer.

[0043] Based on the same inventive concept, the present invention also proposes a computer device, including a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method for pre-treating steel coils at the entrance of the pickling mill described above.

[0044] Based on the same inventive concept, the present invention also proposes a computer-readable storage medium, which stores a computer program. When the computer program is run by a processor, the steps of the pickling mill inlet steel coil pretreatment method are executed as described above.

[0045] The present invention is beneficial in that:

[0046] While traditional workflows often address only a single issue, the integrated system proposed in this paper integrates multiple links into a complete processing chain. The coordinated operation of the strapping removal unit, status information detection unit, defect detection unit, and inkjet code recognition unit enables a more comprehensive and systematic solution to various issues encountered during the early stages of coil preparation. Signal transmission between the various components of the system enables real-time information feedback. The transmission of the removal completion signal, the collection of coil status information, and the transmission of defect detection results make the entire workflow more efficient and enable timely adjustment of processing strategies.

[0047] The end-face defect detection unit in the integrated system proposed by the present invention can perform a variety of quality defect detection (such as inner diameter detection, tower shape detection, etc.) by analyzing the status information of the steel coil. This multi-dimensional detection method can promptly discover and identify potential problems with the steel coil, improving product quality control. By reducing manual intervention and achieving rapid detection and identification, the system can significantly improve the operating efficiency of the production line. The time saved during operation can be converted into higher production capacity to meet changes in production needs. The introduction of the system eliminates the need for operators to perform tedious manual operations, thereby reducing labor intensity, alleviating the burden on workers, and improving the safety and comfort of the working environment. The application of the inkjet recognition unit can compare the detection results with the ID read by the host computer to form a complete data record. This data can not only be used for real-time monitoring, but also provide a basis for subsequent production optimization and quality improvement. The integrated system can quickly adapt to different steel coil specifications and production requirements, realize flexible production adjustments, and improve the adaptability and flexibility of the production line.

[0048] The integrated system proposed in the present invention improves the efficiency and accuracy of end face detection, and completes the detection process of a single coil (steel plate thickness 1.8-6.0mm, single coil size maximum outer diameter 2050mm, inner diameter 760mm, maximum width 1680mm) within 3 minutes, with a defect detection accuracy of 2mm; the steel coil defect recognition rate is above 95%; the supporting equipment integrity rate is 98%; the supporting mechanical equipment operation reliability is ≥95%; and the automation equipment and software operation reliability is ≥98%.

[0049] The invention is applied in the field of metallurgy. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a hardware architecture diagram of the unbundling strap and steel coil status information collection system described in Implementation Method 2;

[0051] Figure 2 This is the overall layout diagram of the unbundling strap system described in the second embodiment;

[0052] Figure 3 This is a schematic diagram of the robot's working range according to the second embodiment;

[0053] Figure 4 This is a schematic diagram of defect detection according to the fourth embodiment;

[0054] Figure 5 This is a schematic diagram of artificial edge crack / notch defect information according to the fourth embodiment;

[0055] Figure 6 This is a flow chart of the strapping unit according to the eleventh embodiment;

[0056] Figure 7 This is a flow chart of the steel coil status information collection system according to the eleventh embodiment;

[0057] Figure 8 This is a diagram showing the on-site equipment layout of the end face defect detection unit according to the eleventh embodiment;

[0058] Figure 9 This is a schematic diagram of the shooting effect of the inkjet code recognition camera according to the eleventh embodiment;

[0059] Figure 10 This is a schematic diagram of the defect detection model framework according to the eleventh embodiment;

[0060] Figure 11 Schematic diagram of detecting a tower-shaped defect according to the eleventh embodiment;

[0061] Figure 12 Schematic diagram of the overflow edge defect described in embodiment eleven. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in combination with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0063] Embodiment 1: The integrated system for pre-treatment of steel coils at the inlet of the pickling mill described in this embodiment comprises:

[0064] Unbundling unit, coil status information detection unit, end surface defect detection unit and inkjet code recognition unit;

[0065] The strapping unit is used to identify the strapping head and remove the strapping, and sends a disassembly completion signal to the coil status information detection unit after the strapping is completed;

[0066] The coil status information detection unit is used to receive the disassembly completion signal and obtain the current coil status information, including: coil width measurement, coil temperature measurement, coil collapse detection, and lead position determination; and send the current coil status information to the end surface defect detection unit;

[0067] The end surface defect detection unit is used to receive the status information of the current steel coil, perform steel coil end surface quality defect detection and thickness measurement based on the current steel coil status information, and send the detection information to the inkjet coding recognition unit; the steel coil end surface quality defect detection includes inner diameter detection, tower shape detection, end surface crack detection, loose coil detection, flat coil detection and staggered layer detection;

[0068] The inkjet coding recognition unit is used to receive defect information from the end face defect detection unit, confirm the status of the defective steel coil and determine the picture where the inkjet coding is located, detect and recognize the text of the inkjet coding picture, and compare the recognized information with the ID read by the host computer.

[0069] The integrated system described in this embodiment automates the entire process, from unbundling and condition detection to defect identification, through the linkage of various units, significantly reducing the need for manual operation. This high degree of automation not only improves work efficiency but also reduces the possibility of human error.

[0070] While traditional workflows often address only a single issue, the integrated system proposed in this embodiment integrates multiple links into a complete processing chain. The coordinated operation of the strapping removal unit, status information detection unit, defect detection unit, and inkjet code recognition unit enables a more comprehensive and systematic solution to various issues encountered during the early stages of coil preparation. Signal transmission between the various components of the system enables real-time information feedback. The transmission of the removal completion signal, the collection of coil status information, and the transmission of defect detection results make the entire workflow more efficient and enable timely adjustment of processing strategies.

[0071] In this embodiment, the end face defect detection unit can perform a variety of quality defect detection (such as inner diameter detection, tower type detection, etc.) by analyzing the status information of the steel coil. This multi-dimensional detection method can timely discover and identify potential problems of the steel coil and improve product quality control. By reducing manual intervention and achieving rapid detection and identification, the system can significantly improve the operating efficiency of the production line. The time saved during operation can be converted into higher production capacity to meet changes in production needs. The introduction of the system means that operators no longer need to perform tedious manual operations, thereby reducing labor intensity, alleviating the burden on workers, and improving the safety and comfort of the working environment. The application of the inkjet recognition unit can compare the test results with the ID read by the host computer to form a complete data record. These data can not only be used for real-time monitoring, but also provide a basis for subsequent production optimization and quality improvement. The integrated system can quickly adapt to different steel coil specifications and production requirements, realize flexible production adjustments, and improve the adaptability and flexibility of the production line.

[0072] Implementation method 2, see Figure 1 、 Figure 2 and Figure 3 This embodiment further defines the integrated system for pre-treatment of steel coils at the entrance of the pickling mill described in the first embodiment, wherein the strapping removal unit removes the straps using an industrial robot, and includes:

[0073] The industrial robot is set on a slide, which drives the robot to move towards the steel coil until it reaches the preset optimal distance;

[0074] The industrial robot's built-in sensors monitor the distance to the steel coil in real time, ensuring it stops at the optimal distance;

[0075] The industrial robot moves the debundling head and laser ranging sensor directly above the steel coil. The laser sensor detects the distance between the debundling head and the surface of the steel coil in real time.

[0076] When the distance detected by the laser sensor reaches the optimal distance, the industrial robot stops moving downward;

[0077] The industrial robot carries the debundling head and moves horizontally along the axial direction of the steel coil, and detects the position of the strapping;

[0078] If the strap is successfully detected, the industrial robot moves down with the debundling head. The controller controls the debundling head to lightly press on the surface of the steel coil to ensure that the strap is in the middle of the fixed shear blade. The fixed shear blade is pushed by the cylinder to scoop up and clamp the strap. The movable shear blade then cuts the strap. If the strap is not cut successfully, an alarm is issued.

[0079] The slide drives the industrial robot back to its initial position, and the industrial robot clamps the cut strap and sends it to the strap reel; the reel curls and compresses the strap into a rolled disc, and the strapping motor is released, and the curled strap falls directly into the waste hopper.

[0080] In this embodiment, the strapping unit is mainly composed of an industrial robot, a strapping cutting machine head, a strapping recovery coiler, a robot base and slide, a strapping waste collection hopper, an electronic control system, etc. The overall scheme and layout are as follows: Figure 2 The industrial robot model is ABB-IRB6700-150 / 3.20, including the robot body, controller, teaching pendant (PP) and cables. Its working range is as follows Figure 3 shown.

[0081] In this embodiment, an industrial robot performs strapping removal, reducing reliance on manual labor and lowering labor costs. Furthermore, the robot can perform highly repetitive tasks, improving operational consistency and reliability. Built-in sensors monitor the distance to the coil in real time, ensuring the robot stops at the optimal distance. This not only improves operational precision but also enhances the system's intelligence. The slide's movement works in tandem with the robot, enabling rapid access to the coil, reducing wait time and operating intervals, and improving the efficiency of the entire removal process. A laser ranging sensor monitors the distance between the strapping head and the coil in real time, ensuring the head is in the optimal position, reducing manual adjustment time and improving work efficiency. Robotic operation reduces the risk of workers coming into direct contact with heavy equipment and hazardous materials, contributing to improved workplace safety. If the strapping is not successfully cut, the system issues an alarm, enabling the operator to intervene promptly to avoid accidents. The robot's controller precisely controls the position and force of the strapping head, ensuring the strap is centered between the fixed shear blades, reducing the possibility of cutting errors. Real-time feedback on the success of strap position detection ensures the reliability of subsequent operations. The winder curls and compresses the cut straps into a disc shape, which drops directly into the waste hopper, reducing the workload of manual cleaning and improving the cleanliness of the working environment.

[0082] Implementation method 3: This implementation method further limits the integrated system for pre-processing steel coils at the entrance of the pickling mill described in implementation method 1. The steel coil status information detection unit acquires the status information of the steel coils through a laser sensor, an infrared temperature sensor, and an intelligent camera.

[0083] Two sets of high-precision laser sensors are used to shoot at the two ends of the steel coil. An adjustable device is used to ensure that the two laser sensors are aligned in both vertical and horizontal positions. The distance between the two sensors is calculated by subtracting the distance between the steel coil ends collected by the sensors to obtain the steel coil width data.

[0084] The infrared temperature sensor is used to collect the single-point temperature signal on the end face of the steel coil and obtain the temperature value of the steel coil;

[0085] The inner coil of the steel coil is photographed and analyzed by a smart camera, the collected outline dimensions of the inner coil of the steel coil are processed, the minimum diameter of the inner ring of the steel coil is calculated, fitted and compared with the standard diameter, and an appropriate deviation value is set. If it exceeds the specified range, an alarm prompt will be given.

[0086] In this embodiment, the coil status information detection unit primarily comprises coil width measurement, coil temperature measurement, coil collapse detection, and lead position determination. This enables online detection of coil status information before it goes online, preventing unnecessary production line downtime due to incorrect coil type, excessive coil width, overheating, and internal coil collapse.

[0087] Given the coil width range of 930-1680mm, the coil width measurement function was implemented using a high-precision laser sensor solution. Two high-precision laser sensors were selected to illuminate the two ends of the coil. Adjustable devices were used to align the two laser sensors' vertical and horizontal heights. The coil width data was calculated by subtracting the distance between the two sensors from the coil end distance measured by the sensors. Based on the coil width range of 930-1680mm, the selected hardware parameters were as follows: model LR-TB5000C; reference distance 5000mm; measurement range 60-5000mm; and repeatability ±3mm.

[0088] The coil temperature measurement function uses a high-precision infrared temperature sensor, ensuring that the collected temperature values ​​are within a reliable accuracy range. Installed in a fixed position, it collects the temperature signal from a single point on the coil end surface and obtains the coil temperature value. Key hardware specifications are as follows: Model S50LRSF; spectral range 8-14μm; measurement range -40 to 300°C; sensor accuracy ±1% or ±2°C; temperature resolution: output resolution ±0.1°C.

[0089] The hardware parameters of the smart camera in this embodiment are as follows:

[0090]

[0091]

[0092] Implementation method 4, see Figure 4 and Figure 5 This embodiment further defines the integrated system for pre-treatment of steel coils at the inlet of the pickling mill described in the first embodiment, wherein the detection of cracks on both end faces includes:

[0093] The image information of the steel coil is collected by an optical camera, and the point cloud information of the local end surface is collected by a structured light detector;

[0094] De-noise and grid-align the collected point cloud information to obtain a point cloud image;

[0095] Convert the point cloud image into a grayscale depth map;

[0096] Perform instantiation segmentation on the depth map and perform regional separation on the parts whose segmentation size is larger than the set defect size;

[0097] The defect recognition model is trained based on the image data with the defect locations marked using the defect detection model;

[0098] Through data training, the final model parameters are determined, and the separated suspected defect areas are tested to obtain the final results.

[0099] like Figure 5 The image shows edge crack / notch defects on the end face of a steel coil simulated using steel coil strapping. The image contains 17 artificial defects, of which the five highlighted in red do not meet the defect criteria and are used to test the robustness of the algorithm. This means there are 12 end face defects to be detected. The raw defect detection results are shown in the appendix - Raw Edge Crack / Notch Detection Results. The image contains 48 defect images, showing 294 defects, of which 280 were detected, for an accuracy rate of 95.24%.

[0100] In this embodiment, an optical camera is used to capture image information of the steel coil, combined with point cloud information collected by a structured light detector, to provide richer surface feature data. This multimodal data acquisition method improves detection accuracy, making cracks and other defects easier to identify. Denoising the collected point cloud data reduces the impact of environmental noise or other interference on the detection results. Grid alignment ensures consistency between the point cloud data and the image data, enhancing the accuracy of subsequent analysis. Converting the point cloud image into a grayscale depth map helps extract more detailed surface information. The instanced segmentation technique of the depth map effectively locates and isolates potential defect areas, especially when the defects are large, effectively reducing the probability of false detection and missed detection. By inputting image data with defect locations into the defect detection model for training, the model continuously learns and optimizes. The trained model can more accurately identify various defects and is highly adaptable to different batches and types of steel coils. The entire inspection process is highly automated, from image acquisition and data processing to defect identification, reducing manual intervention and improving inspection efficiency and reliability. This integrated system can significantly shorten inspection time and accelerate production processes. By testing the isolated suspected defect areas, more accurate detection results can be obtained. This method not only improves the accuracy of defect identification, but also facilitates subsequent quality control and improvement measures.

[0101] Implementation mode 5: This implementation mode further limits the integrated system for pre-processing steel coils at the inlet of the pickling mill described in implementation mode 4. The defect detection model includes: an input end, a backbone network, a neck module, and a prediction module.

[0102] After the input end reads the color image or grayscale image, it performs Mosaic data enhancement, cmBN, and SAT self-adversarial training; the backbone network encodes the input image through CSPDarknet53, Mish activation function and Dropblock; the neck module pre-processes the features generated by the backbone network and strengthens the features; the prediction module prediction uses the loss function CIOU_Loss for training, and the prediction box screening uses DIOU_nms to screen and predict the final defects.

[0103] This embodiment can effectively increase sample diversity and improve the model's generalization ability for different defect scenarios by splicing the input images. During the training process, by mixing batch data of different categories, the deviation between categories is reduced, making the model more robust during the learning process. The use of SAT (self-adversarial training) enhances the model's robustness to noise and uncertainty, which helps to improve the reliability of detection. The CSPDarknet53 network structure combines CSP (CrossStage Partial) technology to reduce computational complexity by optimizing the connection of network layers while maintaining efficient feature extraction capabilities. Compared with traditional activation functions, Mish can provide smoother gradients, which helps the convergence of the model, accelerates the training process, and thus improves detection accuracy. The Dropblock regularization method randomly discards regional features during training, enhances the generalization ability of the model, and prevents overfitting. The neck module (Neck) optimizes and strengthens the features output by the backbone network through pre-processing functions, so that the subsequent prediction module can capture defect features more accurately, thereby improving the accuracy of defect detection. The CIOU_Loss loss function not only considers bounding box overlap but also incorporates factors such as center point distance and aspect ratio, significantly improving the accuracy of predicted boxes and reducing false positives and false negatives. The DIOU_nms screening mechanism uses a non-maximum suppression algorithm that comprehensively evaluates distance and overlap, effectively suppressing redundant predicted boxes and improving the quality of the final detection results. The integration of these modules forms an efficient workflow, from data input to feature extraction to prediction output. The modules are organically linked, improving overall processing speed.

[0104] Embodiment 6: This embodiment further defines the integrated system for pre-treatment of steel coils at the entrance of the pickling mill described in embodiment 1. The steel coil loose coil detection includes:

[0105] Collect steel coil image information;

[0106] Set a threshold based on the depth value to remove the foreground and background;

[0107] The plane contour of the steel coil is obtained by detecting the contour detection algorithm;

[0108] By comparing the distances between the remaining detected contours and the contour with the largest area, a threshold is set to determine whether it is a loose coil; if it is determined to be a loose coil, the depth value of the loose coil and the depth value of the steel coil plane are detected to see if they exceed a threshold. If so, it is determined to be a misjudgment caused by the tower shape, otherwise it is determined to be a loose coil.

[0109] By setting a depth threshold and comparing contours, the probability of misjudgment is effectively reduced. Depth information provides additional validation, especially in situations where other objects (such as tower structures) may interfere. This method not only identifies loose coils but also adapts to coils of varying shapes and sizes, enhancing the system's ability to handle a wide variety of products.

[0110] Implementation method seven. This implementation method further limits the integrated system for pre-processing steel coils at the entrance of the pickling mill described in implementation method one. The inkjet code recognition unit is composed of a main inkjet code recognition camera and an auxiliary inkjet code recognition camera. The main inkjet code recognition camera is facing the steel coil and is used to collect the inkjet code on the right end face of the steel coil; the auxiliary inkjet code recognition camera collects the inkjet code image of the left end face of the saddle at an inclined perspective; and the inkjet code number is identified and compared through image enhancement, correction and OCR technology.

[0111] Embodiment 8: A method for pretreating steel coils at an inlet of a pickling mill, the method comprising:

[0112] The strapping unit dismantles the strapping and sends a completion signal to the coil status information detection unit when the dismantling is completed;

[0113] The coil status information detection unit receives the completion signal and obtains the coil status information and sends it to the end surface defect detection unit;

[0114] The end surface defect detection unit determines whether the steel coil has defects based on the steel coil status information and sends the detection results to the inkjet coding recognition unit;

[0115] The coding unit obtains coding information based on the detection results and the optical camera, and feeds it back to the host computer.

[0116] Embodiment 9. A computer device comprises a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method for pre-treating steel coils at the entrance of a pickling mill as described in embodiment 8.

[0117] Embodiment 10: A computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method for pre-treating steel coils at the inlet of a pickling mill as described in embodiment 8 are executed.

[0118] Implementation method 11, see Figure 6 、 Figure 7 、 Figure 8 、 Figure 9 、 Figure 10 、 Figure 11 and Figure 12This embodiment provides a specific example of the integrated system for pre-treatment of steel coils at the entrance of the pickling mill described in the first embodiment, and is also used to explain the second to seventh embodiments. Specifically:

[0119] The integrated system includes: a strapping unit, a coil status information detection unit, an end surface defect detection unit and a coding recognition unit;

[0120] The walking beam transport system transports the steel coil to the unbundling operation station and locks it, and sends the "start unbundling" signal to the unbundling belt unit. The unbundling belt unit controls the rotation of the ground roller based on the belt head position information sent back by the belt head tracking system, and places the belt head position at 7:00 o'clock on the steel coil (if at this position, the ground roller rotation is not controlled). After confirmation by the self-built system, the system starts the unbundling operation.

[0121] The slide drives the robot to the optimal distance toward the coil. The robot then moves the debundling head and laser ranging sensor above the coil and vertically descends. The laser sensor measures the distance between the debundling head and the coil surface in real time. The robot stops descending when it reaches the optimal distance from the sensor.

[0122] The robot, carrying the debundling head, then moves horizontally along the coil's axis, simultaneously detecting the position of the strap. If the strap is not detected, it returns to its pre-translation position and rechecks. If the strap is still not detected after three attempts, an alarm will sound. After detecting the strap, the robot, carrying the debundling head, moves downward. Controlled by a proximity switch, the debundling head gently presses against the coil's surface. At this point, the strap is positioned between the fixed shear blades. The fixed shear blades, driven by a pneumatic cylinder, scoop up and clamp the strap, and the movable shear blades then cut the strap. If the strap is not effectively cut, an alarm will sound.

[0123] The slide drives the robot back, and it then carries the cut strap to the rewinder, where it is compressed into a coiled disc. Simultaneously, the robot returns to its initial position. The rewinding motor releases, and the coiled strap falls directly into the waste hopper.

[0124] After the entire unbundling process is completed, the unbundling belt unit will feed back the signal to the strap flattening system to perform the strap flattening operation.

[0125] The strapping removal unit uses a combination of vision and robotics to identify and remove the strapping head. It automatically identifies the strapping head position, and the robot scans the strapping position before automatically removing it, eliminating the complexity of manual removal and effectively improving efficiency. The vision component uses image processing tools and algorithm selection to identify and determine the strapping head position. Traditional image processing methods are difficult to identify the coiled strapping head, requiring deep learning to train and learn various coiled strapping heads. After defining a model, collecting and analyzing data, injecting training data into the improved model, and outputting the results, the process is repeated to continuously improve accuracy. Robotic strapping removal requires a slide to transport the robot to the appropriate position. Sensor signals are transmitted between the robot and the strapping removal head to locate the strap, remove it, and place it on the winding machine. The operator simply activates the work button, and the system automatically begins operation. Once the coil is in place, the robot begins removing the strapping head, requiring no human intervention.

[0126] In addition to fully automatic unbundling, semi-automatic unbundling is also possible. In this state, the operator can manually control the walking beam to transport the steel coils to the unbundling station. After the operator confirms that the unbundling system is ready, he turns the control cabinet knob to semi-automatic mode and manually clicks the "Operator Manual Bundle Cutting" button to start unbundling. The process is as follows Figure 6 shown.

[0127] The coil status information detection unit receives the signal that the steel coil is in place from the stepping beam, and reads the current coil number information, and controls the label recognition unit, collapsed coil recognition unit, temperature measurement unit and width measurement unit to perform the detection operation. The label recognition unit uses the label information collected by OCR technology and compares it with the coil number information of the stepping beam. If there is a difference, a second re-inspection and comparison will be carried out. If the results are still inconsistent, an alarm will be issued; the collapsed coil recognition unit will perform contour recognition processing on the collected image of the inner circle of the steel coil, find the inner circle contour line and fit it with the standard contour. If it exceeds the set threshold, an alarm will be issued; the temperature measurement unit reads the single-point temperature information of the end face of the steel coil. If the temperature value exceeds the maximum temperature limit, an alarm will be issued and uploaded to the stepping beam control system and displayed on the HMI screen; the width measurement unit will calculate and process the collected data to obtain the width information of the steel coil. If it exceeds the width range of this type of steel coil, an alarm will be processed and uploaded to the stepping beam control system and displayed on the HMI screen. The specific flow chart is as follows Figure 7 shown.

[0128] In this embodiment, the stepping beam is arranged in an L-shape. To facilitate the subsequent design description, the vertical section in the bird's-eye view is named Section A and the horizontal section is named Section B. The image is collected through the OCR module, where the resolution of the OCR module meets 300M and the frame rate meets 10fps. The OCR module is divided into two groups: the main inkjet recognition camera and the auxiliary inkjet recognition camera, which are installed near saddle No. 5 and on the left side of saddle No. 7 respectively. The main inkjet recognition camera is facing the steel coil of saddle No. 5 and can shoot the inkjet code on the right end face of the steel coil. The shooting effect is as follows: Figure 9 The auxiliary inkjet recognition camera takes a photo of the left side of saddle No. 7 at an oblique angle, and the effect is shown in the left picture. Figure 9 As shown in the right picture, the inkjet code number is then identified and compared through image enhancement, correction and OCR technology.

[0129] OCR equipment parameters are as follows:

[0130]

[0131] The structured light detector is installed at the end of the robotic arm through a mechanical connection component. By controlling the movement of the robotic arm, the structured light detector is moved to the target position to collect local point cloud information of the end face; based on the designed scanning path, the global scanning information of the end face can be obtained.

[0132] Structured light detector parameters:

[0133]

[0134]

[0135] Structured light detector calibration method:

[0136] 1. Prepare a checkerboard calibration plate and calibrate using Zhang Zhengyou's calibration method;

[0137] 2. Place the camera at a distance of 600-700mm from the calibration plate and take several pictures of the calibration plate from different angles.

[0138] 3. Detect the feature points of each image, calculate the homography matrix and optimize it using LMA;

[0139] 4. Solve the intrinsic and extrinsic parameters of the camera under ideal distortion-free conditions;

[0140] 5. Apply the least squares method to find the actual distortion coefficient;

[0141] 6. Integrate internal parameters, external parameters, and distortion coefficients, use LMA, optimize estimation, and improve estimation accuracy;

[0142] 7. Obtain camera intrinsic parameters and extrinsic parameters as distortion coefficients.

[0143] In this embodiment, the main inkjet code recognition includes: determining the center of the steel coil + polar coordinate transformation + OCR;

[0144] Among them, polar coordinate transformation: after the perspective transformation of the image, it can be regarded as taking a picture at a vertical angle to the end face, obtaining a perfect circle that approximates the true shape of the end face, detecting the center position of the circle in the image and the outermost outer diameter of the end face, and performing polar coordinate transformation on the end face image with the center of the end face circle as the center. The annular end face is flattened, and the arc-shaped inkjet code on the end face can be transformed into linear text, which is more conducive to text detection and recognition.

[0145] In this embodiment, the secondary inkjet code recognition includes: curved image correction + OCR;

[0146] Curved image correction includes: perspective transformation + polar coordinate transformation + conventional image enhancement;

[0147] Use perspective transformation to re-project images taken at an oblique angle to a normal viewing angle. The vertex position parameters of the transformation can be adjusted according to the specific angle of the camera installed on site. After the transformation, the coding font size becomes uniform and the text quality can be further enhanced.

[0148] Polar coordinate transformation: After the perspective transformation of the image, it can be regarded as a photo taken at a vertical angle to the end face, obtaining a perfect circle that approximates the true shape of the end face. The center position of the circle and the outermost diameter of the end face in the image are detected. The end face image is then polar-coordinate transformed with the center of the circle as the center. The annular end face is flattened, and the arc-shaped inkjet code on the end face can be transformed into linear text, which is more conducive to text detection and recognition.

[0149] OCR recognition algorithms include:

[0150] The character area was cropped from the straightened image, and then the character recognition model was used to perform character recognition on the cropped image. ABINet was used as the OCR recognition solution for this experiment, and three advanced academic models, CRNN, SAR, and VisionLAN, were selected as control models for experimental comparison. A total of 372 experimental images were used, and a 4-fold validation method was used. This means that a quarter of the images were randomly selected as test images each time, and the remaining images were used as model training images. The experiment was repeated four times, and the test results are shown below:

[0151] Model Average accuracy ABINet 95.7 CRNN 85.22 SAR 72.72 VisionLAN 90.22

[0152] From the above test results, it can be seen that the final detection results of the ABINet model are significantly better than those of the other three models. Therefore, this method is selected as the OCR recognition algorithm in the detailed design plan. At the same time, the final effect of the model is directly related to the number of training images, and subsequent performance improvements depend on the addition of a large number of new images.

[0153] The edge crack / notch defect detection principle of the end face defect detection unit in this embodiment is as follows:

[0154] The collected point cloud is denoised and grid-aligned; the point cloud image is converted into a grayscale depth map; the image is instantiated and segmented, and the area of ​​the segmentation size larger than the set defect size is separated; the defect recognition model is trained based on the image data with the defect location marked and the defect detection model; through data training, the final model parameters are determined, and the separated suspected defect areas are detected to obtain the final result. Among them, the defect recognition model is as follows Figure 10 shown.

[0155] The thickness detection principle of the end face defect detection unit in this embodiment is specifically as follows: grid alignment of the collected point cloud; point cloud clustering using adaptive machine learning; morphological segmentation of the clustering results; end face tomography reconstruction; spatial local layer thickness component analysis and calculation; parallel differential single layer search; and global average layer thickness calculation and verification.

[0156] Tower / overflow edge defect detection method:

[0157] Because the defects of the tower and overflow edge are similar and the experimental site size is limited, the end face of the small steel coil in the simulation experiment is placed parallel to the ground, and then the inspection is carried out using the following algorithm:

[0158] 1. Set a threshold based on the depth value to remove the foreground and background;

[0159] 2. Count the depth values ​​and select the depth value with the most points as the coil plane. Set another threshold so that depth values ​​within the range of [ad, a+d] are considered to be the coil plane range.

[0160] 3. Check the image data in [0, ad] and [a+d, 255], and determine whether it is a pyramid / overflow edge based on whether the area is higher than a certain threshold.

[0161] According to the depth value of the coil plane, the area beyond a certain height of the coil plane is extracted, and it is judged whether the area of ​​the area exceeds the threshold. If it exceeds a certain threshold, it is considered to be a tower shape, such as Figure 11 shown.

[0162] Based on the width of the detected defect feature and the thickness of the steel coil, it can be determined whether the defect is a tower type or an overflow edge, such as Figure 12 Shown is an overflow edge defect.

[0163] In this embodiment, each unit communicates via industrial Ethernet to meet the requirements of long-distance transmission and shield electromagnetic interference.

[0164] Specifically:

[0165] Interlocking communication between the robot arm and the walking beam: The PLC and the walking beam PLC are connected using hard-wired IO to ensure that the robot arm will not start when the walking beam is moving; the walking beam will not start when the robot arm is working, thereby ensuring the safe operation of the equipment.

[0166] Information transmission of steel coils corresponding to saddles 5#, 7#, and 8#: A sensor for collecting temperature and coil width information is installed next to saddle 5#, and this information is collected using a PLC and transmitted to a logic processing module (PLC) via Ethernet communication. The information collection sensors for steel coils on saddles 7# and 8# include optical cameras and structured light detectors, and the image information they collect is transmitted to an industrial computer via Ethernet communication for information processing.

[0167] Robotic arm status communication: The robot arm is configured with PLC and industrial computer through switches and Ethernet to monitor the robot arm's operating status and control the detection process;

[0168] Test result display: The test results processed by the industrial computer are remotely connected to the display in the main control room through Ethernet communication to display the test results and system status.

[0169] Steel coil communication between the walking beam and the host computer: When a new steel coil is transported to section A of the walking beam or moves to section B of the walking beam, the walking beam PLC sends the location information and steel coil ID to the host computer via Ethernet. After receiving the information, the host computer will feedback to the walking beam PLC.

[0170] The integrated system described in this embodiment also includes a power supply system, which powers the OCR module through DC12V, powers the structured light detector through AC220V, powers the host computer through AC220V, powers the robotic arm through AC380V, and powers the logic control device through AC220V.

[0171] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0172] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. The integrated system for pre-treatment of steel coils at the inlet of the pickling mill is characterized by: The integrated system comprises: Unbundling unit, coil status information detection unit, end surface defect detection unit and inkjet code recognition unit; The strapping unit is used to identify the strapping head and remove the strapping, and sends a disassembly completion signal to the coil status information detection unit after the strapping is completed; The coil status information detection unit is used to receive the disassembly completion signal and obtain the current coil status information, including: coil width measurement, coil temperature measurement, coil collapse detection, and lead position determination; and send the current coil status information to the end surface defect detection unit; The end surface defect detection unit is used to receive the status information of the current steel coil, perform steel coil end surface quality defect detection and thickness measurement based on the current steel coil status information, and send the detection information to the inkjet coding recognition unit; the steel coil end surface quality defect detection includes inner diameter detection, tower shape detection, end surface crack detection, loose coil detection, flat coil detection and staggered layer detection; The inkjet coding recognition unit is used to receive defect information from the end face defect detection unit, confirm the status of the defective steel coil and determine the picture where the inkjet coding is located, detect and recognize the text of the inkjet coding picture, and compare the recognized information with the ID read by the host computer.

2. The integrated system for pre-treatment of steel coils at the inlet of the pickling mill according to claim 1 is characterized in that: The strap removal unit removes the straps by an industrial robot, comprising: The industrial robot is set on a slide, which drives the robot to move towards the steel coil until it reaches the preset optimal distance; The industrial robot's built-in sensors monitor the distance to the steel coil in real time, ensuring it stops at the optimal distance; The industrial robot moves the debundling head and laser ranging sensor directly above the steel coil. The laser sensor detects the distance between the debundling head and the surface of the steel coil in real time. When the distance detected by the laser sensor reaches the optimal distance, the industrial robot stops moving downward; The industrial robot carries the debundling head and moves horizontally along the axial direction of the steel coil, and detects the position of the strapping; If the strap is successfully detected, the industrial robot moves down with the debundling head. The controller controls the debundling head to lightly press on the surface of the steel coil to ensure that the strap is in the middle of the fixed shear blade. The fixed shear blade is pushed by the cylinder to scoop up and clamp the strap. The movable shear blade then cuts the strap. If the strap is not cut successfully, an alarm is issued. The slide drives the industrial robot back to its initial position, and the industrial robot clamps the cut strap and sends it to the strap reel; the reel curls and compresses the strap into a rolled disc, and the strapping motor is released, and the curled strap falls directly into the waste hopper.

3. The integrated system for pre-treatment of steel coils at the inlet of the pickling mill according to claim 1 is characterized in that: The steel coil status information detection unit acquires the status information of the steel coil through a laser sensor, an infrared temperature sensor, and an intelligent camera; Two sets of high-precision laser sensors are used to shoot at the two ends of the steel coil. An adjustable device is used to ensure that the two laser sensors are aligned in both vertical and horizontal positions. The distance between the two sensors is calculated by subtracting the distance between the steel coil ends collected by the sensors to obtain the steel coil width data. The infrared temperature sensor is used to collect the single-point temperature signal on the end face of the steel coil and obtain the temperature value of the steel coil; The inner coil of the steel coil is photographed and analyzed by a smart camera, the collected outline dimensions of the inner coil of the steel coil are processed, the minimum diameter of the inner ring of the steel coil is calculated, fitted and compared with the standard diameter, and an appropriate deviation value is set. If it exceeds the specified range, an alarm prompt will be given.

4. The integrated system for pre-treatment of steel coils at the inlet of the pickling mill according to claim 1, characterized in that: The two end face crack detection includes: The steel coil image information is collected by an optical camera, and the local point cloud information of the end surface is collected by a structured light detector; De-noise and grid-align the collected point cloud information to obtain a point cloud image; Convert the point cloud image into a grayscale depth map; Perform instantiation segmentation on the depth map and perform regional separation on the parts whose segmentation size is larger than the set defect size; The defect recognition model is trained based on the image data with the defect locations marked using the defect detection model; Through data training, the final model parameters are determined, and the separated suspected defect areas are tested to obtain the final results.

5. The integrated system for pre-treatment of steel coils at the inlet of the pickling mill according to claim 4 is characterized in that: The defect detection model includes: an input end, a backbone network, a neck module, and a prediction module; After the input end reads the color image or grayscale image, it performs Mosaic data enhancement, cmBN, and SAT self-adversarial training; the backbone network encodes the input image through CSPDarknet53, Mish activation function and Dropblock; the neck module pre-processes the features generated by the backbone network and strengthens the features; the prediction module prediction uses the loss function CIOU_Loss for training, and the prediction box screening uses DIOU_nms to screen and predict the final defects.

6. The integrated system for pre-treatment of steel coils at the inlet of the pickling mill according to claim 1, characterized in that: The steel coil loose coil detection includes: Collect steel coil image information; Set a threshold based on the depth value to remove the foreground and background; The plane contour of the steel coil is obtained by detecting the contour detection algorithm; By comparing the distances between the remaining detected contours and the contour with the largest area, a threshold is set to determine whether it is a loose coil; if it is determined to be a loose coil, the depth value of the loose coil and the depth value of the steel coil plane are detected to see if they exceed a threshold. If so, it is determined to be a misjudgment caused by the tower shape, otherwise it is determined to be a loose coil.

7. The integrated system for pre-treatment of steel coils at the inlet of the pickling mill according to claim 1, characterized in that: The inkjet code recognition unit consists of a main inkjet code recognition camera and a secondary inkjet code recognition camera. The main inkjet code recognition camera faces the steel coil and is used to collect the inkjet code on the right end face of the steel coil; the secondary inkjet code recognition camera collects the inkjet code image of the left end face of the saddle at an inclined perspective; and the inkjet code number is recognized and compared through image enhancement, correction and OCR technology.

8. A method for pre-treating steel coils at the inlet of a pickling mill, characterized in that: The method is implemented based on the integrated system for pre-treatment of steel coils at the entrance of the pickling mill according to any one of claims 1 to 7, and the method comprises: The strapping unit dismantles the strapping and sends a completion signal to the coil status information detection unit when the dismantling is completed; The coil status information detection unit receives the completion signal and obtains the coil status information and sends it to the end surface defect detection unit; The end surface defect detection unit determines whether the steel coil has defects based on the steel coil status information and sends the detection results to the inkjet coding recognition unit; The coding unit obtains coding information based on the detection results and the optical camera, and feeds it back to the host computer.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method for pre-treating steel coils at the entrance of the pickling mill according to claim 8.

10. A computer-readable storage medium, characterized in that The computer readable storage medium stores a computer program, which, when executed by a processor, executes the steps of the method for pre-treating steel coils at the inlet of a pickling mill as claimed in claim 8.

Images