Search

CN-122007280-A - Intelligent end cutting method for visual identification of wire flying shears

CN122007280ACN 122007280 ACN122007280 ACN 122007280ACN-122007280-A

Abstract

The invention discloses a method for visually identifying intelligent crop heads by using wire flying shears, and relates to the technical field of metal rolling processes. The method for intelligently identifying the wire flying shears comprises the steps of hardware deployment and selection, system calibration and parameter adaptation, real-time acquisition of steel part head images, intelligent identification and defect length calculation by an AI algorithm, data transmission and shearing instruction issuing, customized shearing execution and effect verification and system iteration, wherein the method reduces invalid crop quantity, remarkably improves the yield, dynamically matches the actual defect length of each rolled piece through the visual identification and the AI algorithm, replaces the traditional 10-14cm fixed length with the customized shear length of 5cm-15cm, reduces the invalid crop of 3-8cm on average for each rolled piece, greatly reduces the waste steel production amount, combines the identification accuracy of the system of more than or equal to 98.5% and the detection accuracy of less than or equal to +/-3-7 mm, and maximizes the reserved effective steel materials while ensuring the complete removal of the defects.

Inventors

  • FU CHUAN
  • LIU ZHANG
  • HUANG CHENGYONG

Assignees

  • 湖南华菱湘潭钢铁有限公司

Dates

Publication Date
20260512
Application Date
20260305

Claims (7)

  1. 1. A method for identifying intelligent crop by wire flying shears comprises the steps of hardware deployment and selection, system calibration and parameter adaptation, real-time acquisition of steel part head images, intelligent identification by an AI algorithm and defect length calculation, data transmission and shearing instruction issuing, customized shearing execution and effect verification and system iteration, and is characterized in that the step one comprises hardware deployment and selection: The method comprises the steps of arranging 2-4 high-definition global exposure cameras in a rolling mill outlet detection area in front of a No. 1 flying shear, distributing the cameras at an included angle of 140-170 degrees along the circumferential direction of a rolled piece, ensuring to cover a range of 280-320 degrees of the circumferential direction of the rolled piece, reducing detection blind areas, arranging a 12-20mm focal length and large aperture lens on the camera, enabling effective pixels of an image sensor to be not lower than 800 ten thousand, having the functions of automatic exposure, automatic gain and automatic white balance, adapting to imaging requirements under high-temperature working conditions, matching a vision AI controller, requiring a CPU to be a multi-core processor, a memory of 16-32GB and a hard disk of 512GB-1TB, carrying an independent GPU operation card, ensuring multi-task parallel processing capability, integrating all hardware, a power supply and a control module into a cabinet, arranging the cabinet in an operation area convenient for workers to check, and adopting a gigabit-level high-speed network for image transmission.
  2. 2. The method for visually identifying intelligent crop of wire flying shears according to claim 1, wherein the second step comprises the following steps of system calibration and parameter adaptation: The method comprises the steps of carrying out fine adjustment on the installation position and optical parameters of a camera according to the movement speed (0-10 m/s) of a steel piece, the rolling temperature (800-1200 ℃) of the steel piece and the installation space of a field production line, carrying out comprehensive calibration on a system after the installation, determining the conversion relation between image pixels and the actual physical length, setting the calibration period to be 1-3 months/time, ensuring the detection precision to adapt to the change of the field working condition, presetting a dynamic parameter adjustment threshold value, enabling the camera to automatically adjust the imaging parameters according to the temperature of the steel piece and the ambient light, and ensuring the uniform and consistent image brightness of the steel piece under different working conditions.
  3. 3. The method for visually identifying intelligent crop by using the wire flying shears according to claim 1, wherein the third step is that the head image of the steel part is acquired in real time: When the head of the rolled piece enters the detection area, the camera acquires the image of the head of the steel piece in real time according to the frame rate of 30-60 frames/second, the environmental interference factors of water mist, water drops and dust are automatically filtered in the acquisition process, the image is ensured to be free from smear and frame loss, the acquired image is transmitted to the vision AI controller in real time through a high-speed network, the transmission delay is controlled within 20ms, and the real-time property of the data is ensured.
  4. 4. The method for visually identifying intelligent crop of wire flying shears according to claim 1, wherein the AI algorithm intelligent identification and defect length calculation: The vision AI controller starts a deep learning model and a dynamic cutting algorithm to process the acquired images, namely firstly identifying the type of the split, crack and iron oxide scale coverage defect of the head of the steel piece, and then calculating the actual defect length based on the defect distribution range, wherein the algorithm processing time is controlled within 30-60ms (the time required to be earlier than the time required to detect the head of the steel by the first-stage flying shear thermal inspection), the defect length detection precision error is less than or equal to +/-3-7 mm, the defect and shearing length comprehensive identification precision is more than or equal to 98.5%, and finally the optimal shearing length of each rolled piece is output, wherein the length range is 5cm-15cm.
  5. 5. The method for visually identifying intelligent crop of wire flying shears according to claim 1, wherein the fifth step comprises the following steps of: The vision AI controller transmits the calculated optimal shearing length to the L1 automation system in real time in a network message or hard wire communication mode, the length replaces the original fixed-length shearing length setting of 10-14cm, and if the ultra-long defect (more than 14 cm) is detected, a lengthening shearing instruction is automatically generated, so that the defect is prevented from being brought into a downstream process.
  6. 6. The method for visually identifying intelligent crop of wire flying shears according to claim 1, wherein the step six is to customize the shear execution: After the L1 automation system receives the optimal shearing length data, the original thermal detection signal and the flying shear control logic are combined to give an accurate shearing instruction to the flying shear equipment, and the flying shear executes shearing action according to the customized length to ensure that the head defect of the steel part is completely cut off, and meanwhile, the ineffective head cutting amount is reduced to the maximum extent.
  7. 7. The method for visually identifying intelligent crop by using the wire flying shears according to claim 1, wherein the step seven is that the effect verification and the system iterate: The method comprises the steps of setting a feedback detection device at the actual shearing position of the flying shear, comparing the defect length identified by the AI with the actual shearing length in real time, verifying the shearing effect, counting the yield after each batch of production, ensuring that the yield is improved by more than 0.8 per mill-1.2 per mill compared with the traditional process, periodically collecting new working condition data and defect samples, updating a deep learning model, optimizing a dynamic shearing algorithm, and continuously improving the identification precision and the adaptation capability of the system.

Description

Intelligent end cutting method for visual identification of wire flying shears Technical Field The invention relates to the technical field of metal rolling processes, in particular to a method for visually identifying intelligent crop heads by wire flying shears. Background In order to ensure that the rolled piece smoothly enters the subsequent rolling process, the high-speed wire production line needs to cut the irregular part and the low-temperature defect part of the head of the rolled piece through a No. 1 flying shear. The traditional process adopts fixed-length crop setting of 10 cm to 14cm, the length is an upper limit value preset based on the performances and working conditions of various steel grades, and the individual difference of the actual defect length of a single rolled piece is not considered. The length of the cut end is far beyond the actual defect requirement in most cases, a large amount of effective steel materials are mistakenly cut into scrap steel, the recovery cost of the scrap steel is increased, the yield is obviously reduced, and the method becomes a key problem for restricting the benefit improvement of a production line. Disclosure of Invention The invention aims to at least solve one of the technical problems in the prior art, and provides a method for visually identifying intelligent cutting heads by using wire flying shears, which can solve the problem that the head length is far beyond the actual defect requirement. The invention provides a method for intelligently cutting a head through visual identification of wire flying shears, which comprises the following technical scheme that the method comprises the steps of hardware deployment and selection, system calibration and parameter adaptation, steel head image real-time acquisition, AI algorithm intelligent identification and defect length calculation, data transmission and shearing instruction issuing, customized shearing execution and effect verification and system iteration, wherein the step comprises the steps of hardware deployment and selection: The method comprises the steps of arranging 2-4 high-definition global exposure cameras in a rolling mill outlet detection area in front of a No. 1 flying shear, distributing the cameras at an included angle of 140-170 degrees along the circumferential direction of a rolled piece, ensuring to cover a range of 280-320 degrees of the circumferential direction of the rolled piece, reducing detection blind areas, arranging a 12-20mm focal length and large aperture lens on the camera, enabling effective pixels of an image sensor to be not lower than 800 ten thousand, having the functions of automatic exposure, automatic gain and automatic white balance, adapting to imaging requirements under high-temperature working conditions, matching a vision AI controller, requiring a CPU to be a multi-core processor, a memory of 16-32GB and a hard disk of 512GB-1TB, carrying an independent GPU operation card, ensuring multi-task parallel processing capability, integrating all hardware, a power supply and a control module into a cabinet, arranging the cabinet in an operation area convenient for workers to check, and adopting a gigabit-level high-speed network for image transmission. Preferably, the second step comprises the steps of system calibration and parameter adaptation: The method comprises the steps of carrying out fine adjustment on the installation position and optical parameters of a camera according to the movement speed (0-10 m/s) of a steel piece, the rolling temperature (800-1200 ℃) of the steel piece and the installation space of a field production line, carrying out comprehensive calibration on a system after the installation, determining the conversion relation between image pixels and the actual physical length, setting the calibration period to be 1-3 months/time, ensuring the detection precision to adapt to the change of the field working condition, presetting a dynamic parameter adjustment threshold value, enabling the camera to automatically adjust the imaging parameters according to the temperature of the steel piece and the ambient light, and ensuring the uniform and consistent image brightness of the steel piece under different working conditions. Preferably, the third step is that the head image of the steel part is acquired in real time: When the head of the rolled piece enters the detection area, the camera acquires the image of the head of the steel piece in real time according to the frame rate of 30-60 frames/second, the environmental interference factors of water mist, water drops and dust are automatically filtered in the acquisition process, the image is ensured to be free from smear and frame loss, the acquired image is transmitted to the vision AI controller in real time through a high-speed network, the transmission delay is controlled within 20ms, and the real-time property of the data is ensured. Preferably, the AI algorithm intelligently identifies a