A slab surface defect on-line detection and early warning system and method

Abstract

The present application relates to a kind of slab casting surface defect online detection and early warning system and method, belong to steel metallurgy continuous casting production equipment and method technical field.The technical scheme of the present application is: image acquisition module is photographed by high-speed industrial camera and light supplement component to the surface of continuous casting billet after straightening machine full coverage;Defect recognition module is based on deep learning algorithm, and the feature of collected image is extracted, and the type and position of defect are identified;Process linkage module according to defect type matches preset process adjustment scheme, and sends the adjustment instruction of pulling speed and cooling water volume to continuous casting machine control system;Early warning display module real-time shows defect information and process adjustment suggestion.The beneficial effects of the present application are: realize the online, accurate, fast detection of casting billet surface defect, reduce the labor cost, improve the stability of casting billet quality, especially suitable for the continuous casting production of vanadium-titanium slab.

Description

Technical Field

[0001] This invention relates to an online detection and early warning system and method for surface defects in slab castings, belonging to the technical field of continuous casting production equipment and methods in iron and steel metallurgy. Background Technology

[0002] The surface quality of continuously cast slabs directly determines the yield and final product performance of downstream rolling processes. This is especially true for vanadium-titanium steel slabs, which are prone to defects such as corner longitudinal cracks and surface inclusions during solidification due to the presence of vanadium and titanium. Currently, the surface inspection of slabs is mostly done manually by visual inspection. This method has disadvantages such as low inspection efficiency, high rate of missed inspections, and strong subjectivity. Furthermore, it cannot provide timely feedback on defects to guide process adjustments, resulting in a high proportion of defective slabs and affecting production efficiency.

[0003] Some steel companies have introduced single surface inspection equipment, but such equipment is mostly designed for ordinary carbon steel and has low accuracy in identifying defects in vanadium-titanium steel slabs. Furthermore, it lacks integration with the continuous casting machine's process control system, only providing defect alarms and failing to automatically adjust process parameters, thus failing to fundamentally solve the problem of defect generation. Therefore, developing an online detection and early warning system adapted to vanadium-titanium steel slabs and integrating detection with process linkage is of significant practical importance. Summary of the Invention

[0004] The purpose of this invention is to provide an online detection and early warning system and method for surface defects in slab castings. By employing multiple sets of high-speed industrial cameras and deep learning algorithms, the system achieves online, accurate, and rapid detection of surface defects in slab castings, reducing labor costs and improving the stability of slab quality. It is particularly suitable for the continuous casting production of vanadium-titanium steel slabs and effectively solves the aforementioned problems existing in the background technology.

[0005] The technical solution of this invention is: an online detection and early warning system for surface defects in slab castings, comprising an image acquisition module, a data transmission module, a defect recognition module, a process linkage module, and an early warning display module. The image acquisition module includes one or more symmetrically arranged high-speed industrial cameras and supplementary lighting components. The shooting area of ​​the high-speed industrial cameras covers the wide face, narrow face, and corners of the slab. The supplementary lighting components are adjustable, and their light intensity and angle are matched with the surface temperature of the casting and the shooting requirements. The data transmission module is connected to the image acquisition module, and the acquired image data is transmitted to the defect recognition module via 5G or industrial Ethernet. The defect recognition module has a built-in deep learning defect recognition model. The process linkage module is connected to both the defect recognition module and the continuous casting machine control system. The early warning display module is connected to both the defect recognition module and the process linkage module.

[0006] The high-speed industrial camera has a frame rate of no less than 100fps and a pixel resolution of no less than 12 million, and the lens of the high-speed industrial camera is equipped with a high-temperature resistant and dustproof protective cover.

[0007] The deep learning defect recognition model uses images of surface defects on vanadium-titanium steel slabs as training samples. The training samples include images of surface defects on slabs with different vanadium-titanium steel compositions and different process parameters, with a sample size of no less than 100,000 images and a model recognition accuracy of no less than 98%.

[0008] The process linkage module is also equipped with a manual intervention interface.

[0009] The early warning display module also has a data storage structure that records defect data and process adjustment data for 720 consecutive hours, and supports historical data query and trend analysis.

[0010] A method for online detection and early warning of surface defects in slab castings includes the following steps: S1: Surface image acquisition of billet, using multiple sets of high-speed industrial cameras and supplementary lighting components to capture full coverage images of the surface of the continuous casting billet after it exits the straightening machine; S2: Real-time image data transmission. The data transmission module transmits the image data to the backend defect recognition module for processing in real time. S3: Defect identification and judgment. The defect identification module is based on deep learning algorithms to extract features from the acquired images and identify the type and location of defects. S4: Process parameter linkage adjustment. The process linkage module matches the preset process adjustment scheme according to the defect type and sends adjustment instructions for casting speed and cooling water volume to the continuous casting machine control system. S5: Defect Information Early Warning and Data Storage. The early warning display module displays defect information and process adjustment suggestions in real time. The storage end of the early warning display module is connected to the business unit's cloud server to store historical data and support traceability analysis.

[0011] In step S1, the supplementary lighting component adaptively adjusts the intensity and angle of infrared supplementary lighting based on the surface temperature of the billet collected by the temperature sensor and the speed of the roller conveyor collected by the speed sensor.

[0012] In step S2, the data transmission module adopts a dual-mode transmission method of industrial Ethernet and 5G. When the industrial Ethernet fails, it automatically switches to 5G transmission.

[0013] In step S5, the early warning display module is an industrial touch screen installed in the main control room of the continuous casting operation area. It displays the defect image, defect type, size and location of each billet in real time; it also displays the adjustment instructions sent by the process linkage module and the real-time process parameters after adjustment; the storage terminal can record defect data and process adjustment data for 720 consecutive hours, supports querying historical data by time, furnace number or defect type, and generates quality trend curves.

[0014] The beneficial effects of this invention are: by employing multiple sets of high-speed industrial cameras and deep learning algorithms, online, accurate, and rapid detection of surface defects in cast billets is achieved, reducing labor costs and improving the quality stability of cast billets, which is especially suitable for the continuous casting production of vanadium-titanium steel slabs. Attached Figure Description

[0015] Figure 1 This is a structural block diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0016] To make the purpose, technical solutions, and advantages of the invention's embodiments clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only a small part of the embodiments of the present invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0017] An online detection and early warning system for surface defects in slab castings includes an image acquisition module, a data transmission module, a defect recognition module, a process linkage module, and an early warning display module. The image acquisition module includes one or more symmetrically arranged high-speed industrial cameras and supplementary lighting components. The shooting area of ​​the high-speed industrial cameras covers the wide face, narrow face, and corners of the slab. The supplementary lighting components have an adjustable structure, and their light intensity and angle are matched to the surface temperature of the casting and the shooting requirements. The data transmission module is connected to the image acquisition module, and the acquired image data is transmitted to the defect recognition module via 5G or industrial Ethernet. The defect recognition module has a built-in deep learning defect recognition model. The process linkage module is connected to both the defect recognition module and the continuous casting machine control system. The early warning display module is connected to both the defect recognition module and the process linkage module.

[0018] The high-speed industrial camera has a frame rate of no less than 100fps and a pixel resolution of no less than 12 million, and the lens of the high-speed industrial camera is equipped with a high-temperature resistant and dustproof protective cover.

[0019] The deep learning defect recognition model uses images of surface defects on vanadium-titanium steel slabs as training samples. The training samples include images of surface defects on slabs with different vanadium-titanium steel compositions and different process parameters, with a sample size of no less than 100,000 images and a model recognition accuracy of no less than 98%.

[0020] The process linkage module is also equipped with a manual intervention interface.

[0021] The early warning display module also has a data storage structure that records defect data and process adjustment data for 720 consecutive hours, and supports historical data query and trend analysis.

[0022] A method for online detection and early warning of surface defects in slab castings includes the following steps: S1: Surface image acquisition of billet, using multiple sets of high-speed industrial cameras and supplementary lighting components to capture full coverage images of the surface of the continuous casting billet after it exits the straightening machine; S2: Real-time image data transmission. The data transmission module transmits the image data to the backend defect recognition module for processing in real time. S3: Defect identification and judgment. The defect identification module is based on deep learning algorithms to extract features from the acquired images and identify the type and location of defects. S4: Process parameter linkage adjustment. The process linkage module matches the preset process adjustment scheme according to the defect type and sends adjustment instructions for casting speed and cooling water volume to the continuous casting machine control system. S5: Defect Information Early Warning and Data Storage. The early warning display module displays defect information and process adjustment suggestions in real time. The storage end of the early warning display module is connected to the business unit's cloud server to store historical data and support traceability analysis.

[0023] In step S1, the supplementary lighting component adaptively adjusts the intensity and angle of infrared supplementary lighting based on the surface temperature of the billet collected by the temperature sensor and the speed of the roller conveyor collected by the speed sensor.

[0024] In step S2, the data transmission module adopts a dual-mode transmission method of industrial Ethernet and 5G. When the industrial Ethernet fails, it automatically switches to 5G transmission.

[0025] In step S5, the early warning display module is an industrial touch screen installed in the main control room of the continuous casting operation area. It displays the defect image, defect type, size and location of each billet in real time; it also displays the adjustment instructions sent by the process linkage module and the real-time process parameters after adjustment; the storage terminal can record defect data and process adjustment data for 720 consecutive hours, supports querying historical data by time, furnace number or defect type, and generates quality trend curves.

[0026] In practical applications, the connection relationships of the modules of this invention are as follows: The signal output terminal of the image acquisition module is connected to the signal input terminal of the data transmission module, and is used to transmit the acquired image data of the billet surface to the data transmission module; The signal output terminal of the data transmission module is connected to the signal input terminal of the defect recognition module, and is used to transmit image data to the back-end defect recognition module for processing in real time; The signal output terminal of the defect identification module is connected to the signal input terminals of the process linkage module and the early warning display module, respectively, and is used to synchronously send the identified defect information to the process linkage module and the early warning display module. The bidirectional communication terminal of the process linkage module is connected to the continuous casting machine control system, and its signal output terminal is connected to the early warning display module, which is used to send process adjustment commands and feed back the command information to the early warning display module. The storage end of the early warning display module is connected to the business unit's cloud server to store historical data and support traceability analysis.

[0027] The image acquisition module is installed above and on both sides of the conveyor rollers after the continuous casting billet exits the straightener. It includes six sets of high-speed industrial cameras and four sets of adjustable supplementary lighting components. Two sets of high-speed industrial cameras are used to capture the wide side of the billet, two sets are used to capture the narrow side, and two sets are used to capture the corners. The camera lenses are equipped with high-temperature resistant and dustproof protective covers to withstand the radiation temperature of the billet surface. The supplementary lighting components use infrared supplementary lighting technology. The light intensity and angle are adaptively adjusted based on the billet surface temperature and running speed data collected by the sensors to avoid image blurring caused by strong light reflection or insufficient light.

[0028] The data transmission module adopts a dual-mode transmission method of industrial Ethernet and 5G. When the industrial Ethernet fails, it automatically switches to 5G transmission to ensure the real-time and continuous transmission of image data, with a transmission delay of no more than 100ms.

[0029] The defect recognition module incorporates a deep learning defect recognition model based on an improved version of YOLOv8. The model training samples include 150,000 images of surface defects on slabs with different vanadium-titanium steel compositions and process parameters, covering common defects such as corner longitudinal cracks, surface slag inclusions, excessive vibration marks, and scratches. Through feature extraction and classifier optimization, the model achieves a defect recognition accuracy of over 98.5% for vanadium-titanium steel slabs and can accurately output defect type, location coordinates, and size parameters.

[0030] The process linkage module pre-stores a defect-process adjustment comparison table summarized from on-site process data. For example, when a corner longitudinal crack defect is detected, it automatically sends instructions to the continuous casting machine control system to reduce the casting speed by 0.1~0.2m / min and reduce the corner cooling water volume by 10%~15%. When a surface slag inclusion defect is detected, it sends instructions to increase the amount of mold flux added and adjust the molten steel casting speed. At the same time, the module is equipped with a manual intervention interface, allowing operators to modify or pause the automatic adjustment instructions according to the on-site working conditions.

[0031] The early warning display module is an industrial touch screen installed in the main control room of the continuous casting operation area. It displays the defect images, defect types, sizes and locations of each billet in real time; it also displays the adjustment instructions sent by the process linkage module and the real-time process parameters after adjustment; the module also has data storage and trend analysis functions, which can record defect data and process adjustment data for 720 consecutive hours, support querying historical data by time, furnace number and defect type, and generate quality trend curves.

[0032] This invention follows a closed-loop process of image acquisition → data transmission → defect identification → process linkage → early warning feedback, and the specific steps are as follows: Step 1: Image acquisition of the billet surface In the roller conveyor area 12m after the continuous casting billet exits the straightener (where the billet surface temperature is stable at 850~950℃), six sets of high-speed industrial cameras (two sets for the wide face, two sets for the narrow face, and two sets for the corners) are used to capture full coverage images of the billet surface. The supplementary lighting component adaptively adjusts the intensity and angle of the infrared supplementary light based on the billet surface temperature collected by the temperature sensor and the roller conveyor speed collected by the speed sensor. When the billet temperature is >900℃, the supplementary light intensity is reduced by 30%~40% to avoid reflections that cause image blurring. The camera frame rate is set to 120fps and the resolution is 12 megapixels to ensure that minute defects on the billet surface are captured.

[0033] Step 2: Real-time transmission of image data The data transmission module adopts a dual-mode transmission method with industrial Ethernet (gigabit fiber) as the primary mode and 5G as the backup to transmit the collected image data to the main control room server. When the industrial Ethernet fails, it automatically switches to 5G transmission within 50ms to ensure that the transmission delay is stable within 80ms, meeting the real-time detection requirements.

[0034] Step 3: Intelligent Defect Identification and Judgment The defect identification module calls a deep learning model based on YOLOv8 to extract features from the received images and identify defects such as corner longitudinal cracks, surface slag inclusions, excessive vibration marks, and scratches. The model matches the extracted defect features with the training sample library (150,000 images of defects in vanadium-titanium steel slabs) and outputs the defect type, location coordinates (accuracy ±2mm), and size parameters. When the defect size exceeds the preset threshold (e.g., corner longitudinal crack length ≥50mm, vibration mark depth ≥0.5mm), it is determined to be an excessive defect, triggering subsequent linkage and early warning processes.

[0035] Step 4: Automatic adjustment of process parameters After receiving a defect exceeding the standard signal, the process linkage module retrieves the pre-stored "Vanadium-Titanium Steel Slab Defect-Process Adjustment Comparison Table" and generates targeted process adjustment instructions: - Excessive longitudinal cracks at the corner: Send instructions to reduce casting speed by 0.1~0.2m / min and reduce corner cooling water volume by 10%~15% to the continuous casting machine control system; - Excessive surface inclusions: Send instructions to increase the amount of protective slag added by 0.2~0.3 kg / min and reduce the casting speed by 0.05 m / min; Meanwhile, the module retains a manual intervention interface, allowing operators to modify or pause automatic commands according to on-site conditions.

[0036] Step 5: Defect Information Early Warning and Data Storage The early warning display module simultaneously receives defect identification information and process adjustment instructions, and displays defect images, types, locations, and process adjustment parameters in real time on industrial displays in the main control room and field operation room. At the same time, the system stores defect data, process adjustment data, and post-adjustment effect data to a cloud server, supporting continuous 720-hour data recording, multi-dimensional queries, and trend analysis. Example

[0037] 1. Application Environment and Conditions This invention is applied to the No. 2 continuous casting machine in the work area. This continuous casting machine produces vanadium-titanium steel slabs with a thickness of 200mm and a width of 1200~1800mm. The steel grade is Z-DD11, and the monthly production capacity is 35,000 tons. The production environment parameters during the application period are: slab exiting the straightener temperature of 850~950℃, roller speed of 0.8~1.4m / min, and crystallizer vibration frequency of 200~300Hz.

[0038] 2. Detailed Detection and Early Warning Process Example operating conditions: On [Date] at 14:30, Z-DD11 vanadium-titanium steel slab was cast on continuous casting machine No. 2. The casting speed was set to 1.4 m / min and the corner cooling water flow rate was 50 m³ / h.

[0039] 1. Image Acquisition: The six sets of high-speed industrial cameras (Hikvision MV-CH120-10GC) of the image acquisition module take full-coverage pictures of the billet of this furnace number. At this time, the surface temperature of the billet is 920℃. The supplementary lighting component automatically reduces the supplementary lighting intensity by 35%. The camera captures images of continuous slender textures at the corners of the billet.

[0040] 2. Data transmission: Image data is transmitted to the main control room server via industrial Ethernet with a transmission delay of 75ms, meeting real-time requirements.

[0041] 3. Defect Identification: The defect identification module processes the image and extracts a corner texture with a length of 80mm and a width of 1.2mm. The match rate with the corner longitudinal crack feature in the sample library reaches 99.2%, and it is determined to be a corner longitudinal crack exceeding the standard defect.

[0042] 4. Process linkage: The process linkage module retrieves the reference table and generates adjustment instructions: reduce casting speed by 0.15m / min (adjusted to 1.25m / min), reduce corner cooling water volume by 12% (adjusted to 44m³ / h), and sends them to the continuous casting machine PLC control system; the operator confirms that the on-site working conditions are normal and does not interfere with the automatic instructions.

[0043] 5. Early Warning and Storage: The early warning display module pops up a red warning in real time on both screens, showing "Furnace No. VD202X0518, corner longitudinal crack, length 80mm", and simultaneously displays the process adjustment parameters; the system stores the defect data of this furnace number, adjustment instructions and the detection data of the subsequent 20 billets to the cloud server.

[0044] This invention employs multiple sets of high-speed industrial cameras and deep learning algorithms to achieve online, accurate, and rapid detection of surface defects in vanadium-titanium steel slabs, with an accuracy rate of up to 98.5%, far exceeding the levels of manual inspection and traditional equipment inspection, thus reducing the rate of missed detections and false judgments.

[0045] By connecting the process linkage module with the continuous casting machine control system, closed-loop control of defect detection and process adjustment is achieved. This enables timely adjustment of process parameters in the early stages of defect occurrence, reducing the continuous generation of defects from the source and improving the stability of billet quality.

[0046] The early warning display module has data storage and trend analysis functions, which can provide data support for process optimization and quality traceability in the work area, and help the work area to upgrade digitally and intelligently.

Claims

1. An online detection and early warning system for surface defects in slab castings, characterized in that: The system includes an image acquisition module, a data transmission module, a defect identification module, a process linkage module, and an early warning display module. The image acquisition module comprises one or more symmetrically arranged high-speed industrial cameras and supplementary lighting components. The high-speed industrial cameras cover the wide, narrow, and corner areas of the slab. The supplementary lighting components are adjustable, and their light intensity and angle are matched to the surface temperature of the slab and the shooting requirements. The data transmission module is connected to the image acquisition module, and the acquired image data is transmitted to the defect identification module via 5G or industrial Ethernet. The defect identification module has a built-in deep learning defect identification model. The process linkage module is connected to both the defect identification module and the continuous casting machine control system. The early warning display module is connected to both the defect identification module and the process linkage module.

2. The online detection and early warning system for surface defects in slab castings according to claim 1, characterized in that: The high-speed industrial camera has a frame rate of no less than 100fps and a pixel resolution of no less than 12 million, and the lens of the high-speed industrial camera is equipped with a high-temperature resistant and dustproof protective cover.

3. The online detection and early warning system for surface defects in slab castings according to claim 1, characterized in that: The deep learning defect recognition model uses images of surface defects on vanadium-titanium steel slabs as training samples. The training samples include images of surface defects on slabs with different vanadium-titanium steel compositions and different process parameters, with a sample size of no less than 100,000 images and a model recognition accuracy of no less than 98%.

4. The online detection and early warning system for surface defects in slab castings according to claim 1, characterized in that: The process linkage module is also equipped with a manual intervention interface.

5. The online detection and early warning system for surface defects in slab castings according to claim 1, characterized in that: The early warning display module also has a data storage structure that records defect data and process adjustment data for 720 consecutive hours, and supports historical data query and trend analysis.

6. A method for online detection and early warning of surface defects in slab castings, characterized in that... Includes the following steps: S1: Surface image acquisition of billet, using multiple sets of high-speed industrial cameras and supplementary lighting components to capture full coverage images of the surface of the continuous casting billet after it exits the straightening machine; S2: Real-time image data transmission. The data transmission module transmits the image data to the backend defect recognition module for processing in real time. S3: Defect identification and judgment. The defect identification module is based on deep learning algorithms to extract features from the acquired images and identify the type and location of defects. S4: Process parameter linkage adjustment. The process linkage module matches the preset process adjustment scheme according to the defect type and sends adjustment instructions for casting speed and cooling water volume to the continuous casting machine control system. S5: Defect Information Early Warning and Data Storage. The early warning display module displays defect information and process adjustment suggestions in real time. The storage end of the early warning display module is connected to the business unit's cloud server to store historical data and support traceability analysis.

7. The method for online detection and early warning of surface defects in slab castings according to claim 6, characterized in that: In step S1, the supplementary lighting component adaptively adjusts the intensity and angle of infrared supplementary lighting based on the surface temperature of the billet collected by the temperature sensor and the speed of the roller conveyor collected by the speed sensor.

8. The method for online detection and early warning of surface defects in slab castings according to claim 6, characterized in that: In step S2, the data transmission module adopts a dual-mode transmission method of industrial Ethernet and 5G. When the industrial Ethernet fails, it automatically switches to 5G transmission.

9. The method for online detection and early warning of surface defects in slab castings according to claim 6, characterized in that: In step S5, the early warning display module is an industrial touch screen installed in the main control room of the continuous casting operation area. It displays the defect image, defect type, size and location of each billet in real time; it also displays the adjustment instructions sent by the process linkage module and the real-time process parameters after adjustment; the storage terminal can record defect data and process adjustment data for 720 consecutive hours, supports querying historical data by time, furnace number or defect type, and generates quality trend curves.

Images