Visual identification intelligent head cutting method for wire flying shear
By deploying high-definition cameras and visual AI controllers on the wire rod production line, combined with deep learning models, defects at the head of the rolled piece can be identified in real time and the shearing length can be dynamically adjusted. This solves the problem of ineffective cutting in the traditional fixed-length cutting method, and improves the yield and production line stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUNAN VALIN XIANGTAN IRON & STEEL CO LTD
- Filing Date
- 2026-03-05
- Publication Date
- 2026-05-12
AI Technical Summary
In traditional wire rod production, the fixed-length cutting method results in a large amount of ineffective cuts, low yield, and an inability to adapt to individual differences in the actual defect length of rolled products, increasing scrap steel recycling costs and reducing production line efficiency.
Using a high-definition camera and a visual AI controller, combined with a deep learning model and a dynamic cutting length algorithm, defects at the head of the rolled piece are identified in real time, and the cutting length is dynamically adjusted to ensure accurate removal of defects and reduce ineffective cuts.
Significantly improves yield, reduces scrap steel generation, enhances production line stability and economic efficiency, and ensures accurate identification and removal of the actual defect length of each rolled piece through precise matching of visual recognition and AI algorithms.
Smart Images
Figure CN122007280A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of metal rolling technology, and in particular to a method for intelligent cutting head recognition using a wire shear. Background Technology
[0002] To ensure the smooth entry of rolled steel into subsequent rolling processes, high-speed wire rod production lines require the use of a No. 1 flying shear to cut irregular sections and low-temperature defects from the head of the rolled steel. Traditional processes use a fixed cutting length of 10-14cm, a length preset to the upper limit based on the properties and operating conditions of various steel grades, without considering individual differences in the actual defect length of each rolled steel piece. This results in the cutting length often far exceeding the actual defect requirements, leading to a large amount of usable steel being mistakenly cut into scrap. This increases scrap recycling costs and significantly reduces the yield, becoming a key issue restricting the improvement of production line efficiency. Summary of the Invention
[0003] The purpose of this invention is to at least solve one of the technical problems existing in the prior art, and to provide a method for intelligent cutting of wire shears by visual recognition, which can solve the problem that the head length far exceeds the actual defect requirements.
[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for intelligent cutting of wire shears using visual recognition, comprising: step one: hardware deployment and selection; step two: system calibration and parameter adaptation; step three: real-time acquisition of steel head images; step four: AI algorithm intelligent recognition and defect length calculation; step five: data transmission and cutting command issuance; step six: customized cutting execution; and step seven: effect verification and system iteration. Step one: hardware deployment and selection... In the mill exit inspection area before the No. 1 flying shear, 2-4 high-definition global exposure cameras are deployed. The cameras are distributed at an angle of 140°-170° along the circumference of the rolled piece to ensure coverage of the outer circumference of the rolled piece within a range of 280°-320°, reducing blind spots. The cameras are equipped with 12-20mm focal length, large aperture lenses, and image sensors with an effective pixel count of no less than 8 million. They have automatic exposure, automatic gain, and automatic white balance functions, and can adapt to the imaging requirements under high-temperature conditions. A visual AI controller is provided, requiring a multi-core CPU, 16-32GB of memory, 512GB-1TB of hard drive space, and an independent GPU computing card to ensure multi-task parallel processing capabilities. All hardware, power supply, and control modules are integrated into a cabinet and deployed in an operating area that is easily visible to workers. Image transmission uses a gigabit high-speed network.
[0005] Preferably, step two: system calibration and parameter adaptation: Based on the steel parts' movement speed (0-10m / s), rolling temperature (800℃-1200℃), and installation space on the production line, fine-tune the camera's installation position and optical parameters. After installation, conduct a comprehensive calibration of the system to clarify the conversion relationship between image pixels and actual physical length. The calibration cycle is set to 1-3 months / time to ensure that the detection accuracy adapts to changes in on-site working conditions. At the same time, preset dynamic parameter adjustment thresholds allow the camera to automatically adjust imaging parameters according to the steel parts' temperature and ambient light, ensuring uniform brightness of the steel parts' images under different working conditions.
[0006] Preferably, step three involves real-time acquisition of the head image of the steel component. When the head of the rolled piece enters the inspection area, the camera acquires images of the steel head in real time at a frame rate of 30-60 frames per second. During the acquisition process, water mist, water droplets, dust and other environmental interference factors are automatically filtered to ensure that the images are free of ghosting and frame loss. The acquired images are transmitted to the visual AI controller in real time via a high-speed network with a transmission delay controlled within 20ms to ensure data real-time performance.
[0007] Preferably, step four: AI algorithm intelligent identification and defect length calculation: The visual AI controller initiates a deep learning model and a dynamic cutting length algorithm to process the acquired images: First, it identifies the types of defects at the head of the steel piece, such as forks, cracks, and oxide scale coverage. Then, it calculates the actual defect length based on the defect distribution range. The algorithm processing time is controlled within 30-60ms (it needs to be earlier than the time when the first-level flying shear hot inspection detects the steel head). The defect length detection accuracy error is ≤±3-7mm, and the comprehensive recognition accuracy of defects and cutting length is ≥98.5%. Finally, it outputs the optimal cutting length for each rolled piece, with a length range of 5cm-15cm.
[0008] Preferably, step five: data transmission and cutting command issuance: The visual AI controller transmits the calculated optimal cutting length to the L1 automation system in real time via network messages or hardwired communication. This length replaces the original fixed cutting length setting of 10-14cm. If an excessively long defect (more than 14cm) is detected, an extended cutting command is automatically generated to prevent the defect from being carried into the downstream process.
[0009] Preferably, step six: customized cutting execution: After receiving the optimal shearing length data, the L1 automation system combines the original thermal detection signal and flying shear control logic to issue a precise shearing command to the flying shear equipment. The flying shear performs the shearing action according to the customized length, ensuring that defects at the head of the steel part are completely removed, while minimizing the amount of invalid cuts.
[0010] Preferably, step seven: effect verification and system iteration: A feedback detection device is set up at the actual cutting position of the flying shear to compare the defect length identified by AI with the actual cutting length in real time to verify the cutting effect; the yield rate is calculated after each batch of production to ensure that the yield rate is 0.8‰-1.2‰ higher than that of traditional processes; new working condition data and defect samples are collected regularly to update the deep learning model, optimize the dynamic cutting length algorithm, and continuously improve the system's recognition accuracy and adaptability.
[0011] Compared with the prior art, the beneficial effects of the present invention are: 1. This wire rod flying shear uses visual recognition for intelligent cutting, which reduces invalid cuts and significantly improves yield: By dynamically matching the actual defect length of each rolled piece with visual recognition and AI algorithms, a customized cutting length of 5cm-15cm is used instead of the traditional fixed length of 10-14cm. On average, this reduces invalid cuts by 3-8cm per rolled piece, significantly reducing scrap steel generation. Combined with the system's ≥98.5% recognition accuracy and ≤±3-7mm detection precision, it maximizes the retention of usable steel while ensuring complete defect removal, thus increasing the yield by 0.8‰-1.2‰ or more compared to traditional processes. Long-term large-scale production can create considerable economic value.
[0012] 2. The method of intelligent head cutting by visual recognition of wire shear avoids quality accidents and ensures rolling stability: The system can accurately identify defects such as forks, cracks, and iron oxide scale covering at the head of the rolled piece, and can detect extra-long defects exceeding 14cm in a timely manner and automatically generate extended cutting instructions. This completely solves the drawback of traditional fixed lengths that cannot adapt to extra-long defects, effectively avoids equipment failures or product quality problems caused by defective heads being carried into downstream stands, and greatly improves the stability and reliability of the production process.
[0013] 3. The method of intelligent cutting head recognition for wire shear is highly adaptable and continuously optimized over the long term: the system hardware can adapt to complex working conditions with steel movement speeds of 0-10m / s and rolling temperatures of 800℃-1200℃. Through dynamic parameter adjustment algorithms, it filters out interference from water mist and dust environments, demonstrating outstanding robustness. At the same time, it supports regular calibration and model iteration every 1-3 months, and can continuously optimize recognition accuracy and algorithm performance as production data accumulates, helping the production line maintain a high-efficiency and low-consumption operating state in the long term. Attached Figure Description
[0014] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart illustrating the deep learning model training and workflow of the present invention. Figure 2 This is a schematic diagram of the flying shear vision optimization system of the present invention. Detailed Implementation
[0015] This section will describe in detail specific embodiments of the present invention. Preferred embodiments of the present invention are shown in the accompanying drawings. The purpose of the drawings is to supplement the textual description with graphics, so that people can intuitively and vividly understand each technical feature and overall technical solution of the present invention, but they should not be construed as limiting the scope of protection of the present invention.
[0016] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.
[0017] In the description of this invention, terms such as greater than, less than, and exceeding are understood to exclude the stated number, while terms such as above, below, and within are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.
[0018] In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.
[0019] Please see Figure 1-2 This invention provides a technical solution: a method for intelligent wire cutting using visual recognition, comprising the following steps: Step 1: Hardware deployment and selection: In the mill exit inspection area before the No. 1 flying shear, 2-4 high-definition global exposure cameras are deployed. These cameras are arranged at an angle of 140°-170° along the circumference of the rolled piece, ensuring coverage of a 280°-320° range around the outer perimeter to reduce blind spots. The cameras are equipped with 12-20mm focal length, large aperture lenses, and image sensors with at least 8 million effective pixels. They feature automatic exposure, automatic gain control, and automatic white balance, adaptable to imaging requirements under high-temperature conditions. A supporting visual AI controller is provided, requiring a multi-core CPU, 16-32GB of RAM, 512GB-1TB of hard drive space, and an independent GPU to ensure multi-task parallel processing capabilities. All hardware, power supply, and control modules are integrated into a cabinet and deployed in an easily accessible operating area. Image transmission utilizes a gigabit-level high-speed network (fiber optic or wired Ethernet). Step Two: System Calibration and Parameter Adaptation Based on the steel parts' movement speed (0-10m / s), rolling temperature (800℃-1200℃), and installation space on the production line, fine-tune the camera's installation position and optical parameters. After installation, conduct a comprehensive system calibration to clarify the conversion relationship between image pixels and actual physical length. The calibration cycle is set to 1-3 months per calibration to ensure that the detection accuracy adapts to changes in on-site working conditions. At the same time, preset dynamic parameter adjustment thresholds allow the camera to automatically adjust imaging parameters according to the steel parts' temperature and ambient light, ensuring uniform brightness of the steel parts' images under different working conditions. Step 3: Real-time image acquisition of the steel component head: When the head of the rolled piece enters the inspection area, the camera acquires images of the steel head in real time at a frame rate of 30-60 frames per second. During the acquisition process, water mist, water droplets, dust and other environmental interference factors are automatically filtered to ensure that the images are free of ghosting and frame loss. The acquired images are transmitted to the visual AI controller in real time via a high-speed network with a transmission latency controlled within 20ms to ensure data real-time performance. Step 4: AI Algorithm Intelligent Identification and Defect Length Calculation: The visual AI controller initiates a deep learning model and a dynamic cutting length algorithm to process the acquired images: First, it identifies the types of defects at the head of the steel piece, such as forks, cracks, and oxide scale coverage. Then, it calculates the actual defect length based on the defect distribution range. The algorithm processing time is controlled within 30-60ms (it needs to be earlier than the time when the first-stage flying shear hot inspection detects the steel head). The defect length detection accuracy error is ≤±3-7mm, and the comprehensive recognition accuracy of defects and shear length is ≥98.5%. Finally, it outputs the optimal shear length for each rolled piece, with a length range of 5cm-15cm (dynamically adjusted according to the actual defects). Step 5: Data transfer and cut command issuance: The visual AI controller transmits the calculated optimal cutting length to the L1 automation system in real time via network packets or hardwired communication. This length replaces the original fixed cutting length setting of 10-14cm. If an excessively long defect (more than 14cm) is detected, an extended cutting command is automatically generated to prevent the defect from being carried into the downstream process. Step Six: Customized Cutting Execution: After receiving the optimal shearing length data, the L1 automation system combines the original thermal detection signal and flying shear control logic to issue a precise shearing command to the flying shear equipment; the flying shear performs the shearing action according to the customized length to ensure that defects at the head of the steel part are completely removed, while minimizing the amount of invalid cuts. Step Seven: Effect Verification and System Iteration A feedback detection device is set up at the actual cutting position of the flying shear to compare the defect length identified by AI with the actual cutting length in real time to verify the cutting effect; the yield rate is calculated after each batch of production to ensure that the yield rate is 0.8‰-1.2‰ higher than that of traditional processes; new working condition data and defect samples are collected regularly to update the deep learning model, optimize the dynamic cutting length algorithm, and continuously improve the system's recognition accuracy and adaptability. Furthermore, this method reduces the amount of ineffective cuts and significantly improves the yield: by dynamically matching the actual defect length of each rolled piece with visual recognition and AI algorithms, a customized cutting length of 5cm-15cm is used instead of the traditional fixed length of 10-14cm. On average, this reduces the amount of ineffective cuts by 3-8cm per rolled piece, significantly reducing the amount of scrap steel generated. Combined with the system's recognition accuracy of ≥98.5% and detection precision of ≤±3-7mm, it maximizes the retention of effective steel while ensuring complete removal of defects, thus increasing the yield by more than 0.8‰-1.2‰ compared to traditional processes. Long-term large-scale production can create considerable economic value. Furthermore, this method avoids quality accidents and ensures rolling stability: the system can accurately identify defects such as forks, cracks, and iron oxide scale covering at the head of the rolled piece, and can detect ultra-long defects exceeding 14cm in a timely manner and automatically generate an extension shearing command. This completely solves the drawback of traditional fixed lengths being unable to adapt to ultra-long defects, effectively avoids equipment failures or product quality problems caused by defect heads being carried into downstream stands, and significantly improves the stability and reliability of the production process. Furthermore, this method is highly adaptable and offers continuous long-term benefits: the system hardware can adapt to complex working conditions with steel movement speeds of 0-10 m / s and rolling temperatures of 800℃-1200℃, and the dynamic parameter tuning algorithm filters out interference from water mist and dust environments, demonstrating outstanding robustness; at the same time, it supports periodic calibration and model iteration every 1-3 months, and can continuously optimize recognition accuracy and algorithm performance as production data accumulates, helping the production line maintain a high-efficiency and low-consumption operating state in the long term.
[0020] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.
Claims
1. A method for intelligent cutting of wire shears using visual recognition, comprising the following steps: Step 1: hardware deployment and selection; Step 2: system calibration and parameter adaptation; Step 3: real-time acquisition of steel head images; Step 4: AI algorithm intelligent recognition and defect length calculation; Step 5: data transmission and cutting command issuance; Step 6: customized cutting execution; and Step 7: effect verification and system iteration, characterized in that: Step 1: Hardware Deployment and Selection In the mill exit inspection area before the No. 1 flying shear, 2-4 high-definition global exposure cameras are deployed. The cameras are distributed at an angle of 140°-170° along the circumference of the rolled piece to ensure coverage of the outer circumference of the rolled piece within a range of 280°-320°, reducing blind spots. The cameras are equipped with 12-20mm focal length, large aperture lenses, and image sensors with an effective pixel count of no less than 8 million. They have automatic exposure, automatic gain, and automatic white balance functions, and can adapt to the imaging requirements under high-temperature conditions. A visual AI controller is provided, requiring a multi-core CPU, 16-32GB of memory, 512GB-1TB of hard drive space, and an independent GPU computing card to ensure multi-task parallel processing capabilities. All hardware, power supply, and control modules are integrated into a cabinet and deployed in an operating area that is easily visible to workers. Image transmission uses a gigabit high-speed network.
2. The method for intelligent cutting head recognition of wire flying shears according to claim 1, characterized in that: Step two: System calibration and parameter adaptation: Based on the steel parts' movement speed (0-10m / s), rolling temperature (800℃-1200℃), and installation space on the production line, fine-tune the camera's installation position and optical parameters. After installation, conduct a comprehensive calibration of the system to clarify the conversion relationship between image pixels and actual physical length. The calibration cycle is set to 1-3 months / time to ensure that the detection accuracy adapts to changes in on-site working conditions. At the same time, preset dynamic parameter adjustment thresholds allow the camera to automatically adjust imaging parameters according to the steel parts' temperature and ambient light, ensuring uniform brightness of the steel parts' images under different working conditions.
3. The method for intelligent cutting head recognition of wire flying shears according to claim 1, characterized in that: Step 3: Real-time image acquisition of the steel component head: When the head of the rolled piece enters the inspection area, the camera acquires images of the steel head in real time at a frame rate of 30-60 frames per second. During the acquisition process, water mist, water droplets, dust and other environmental interference factors are automatically filtered to ensure that the images are free of ghosting and frame loss. The acquired images are transmitted to the visual AI controller in real time via a high-speed network with a transmission delay controlled within 20ms to ensure data real-time performance.
4. The method for intelligent cutting head recognition of wire flying shears according to claim 1, characterized in that: Step four: AI algorithm intelligent identification and defect length calculation: The visual AI controller initiates a deep learning model and a dynamic cutting length algorithm to process the acquired images: First, it identifies the types of defects at the head of the steel piece, such as forks, cracks, and oxide scale coverage. Then, it calculates the actual defect length based on the defect distribution range. The algorithm processing time is controlled within 30-60ms (it needs to be earlier than the time when the first-level flying shear hot inspection detects the steel head). The defect length detection accuracy error is ≤±3-7mm, and the comprehensive recognition accuracy of defects and cutting length is ≥98.5%. Finally, it outputs the optimal cutting length for each rolled piece, with a length range of 5cm-15cm.
5. The method for intelligent cutting head recognition of wire flying shears according to claim 1, characterized in that: Step 5: Data transmission and cut command issuance: The visual AI controller transmits the calculated optimal cutting length to the L1 automation system in real time via network messages or hardwired communication. This length replaces the original fixed cutting length setting of 10-14cm. If an excessively long defect (more than 14cm) is detected, an extended cutting command is automatically generated to prevent the defect from being carried into the downstream process.
6. The method for intelligent cutting head of wire flying shear with visual recognition according to claim 1, characterized in that: Step six: Customized cut execution: After receiving the optimal shearing length data, the L1 automation system combines the original thermal detection signal and flying shear control logic to issue a precise shearing command to the flying shear equipment. The flying shear performs the shearing action according to the customized length, ensuring that defects at the head of the steel part are completely removed, while minimizing the amount of invalid cuts.
7. The method for intelligent cutting head recognition of wire flying shears according to claim 1, characterized in that: Step seven: Effect verification and system iteration: A feedback detection device is set up at the actual cutting position of the flying shear to compare the defect length identified by AI with the actual cutting length in real time to verify the cutting effect; the yield rate is calculated after each batch of production to ensure that the yield rate is 0.8‰-1.2‰ higher than that of traditional processes; new working condition data and defect samples are collected regularly to update the deep learning model, optimize the dynamic cutting length algorithm, and continuously improve the system's recognition accuracy and adaptability.