Ultra-long thin steel belt high-speed laser wire filling welding and on-line intelligent detection device and detection method

Abstract

The application discloses an ultra-long thin steel strip high-speed laser wire filling welding and on-line intelligent detection device and detection method, and belongs to the field of laser welding equipment. The device solves the problems of low weld seam tracking precision, unstable penetration, difficult defect early warning, non-closed loop quality verification, and non-traceable data of the ultra-long thin steel strip. A double-beam laser wire filling welding unit adopts a front high-power laser to realize high-speed deep penetration, and a rear swing laser to melt the welding wire and stir the molten pool. The two lasers work together in the molten pool. An on-line intelligent detection unit of the molten pool synchronously collects the crater shape, the molten pool appearance and the temperature field through a high-speed camera and an infrared thermal imaging system. A deep learning control unit establishes a defect correlation model based on the characteristics of the molten pool, and marks the suspected defect position in real time. An X-ray verification unit after welding automatically scans and verifies the marked area. The application realizes the high-speed, low-deformation and high-reliability continuous automatic welding production of the ultra-long thin steel strip with a length of greater than or equal to 50 m and a thickness of 1-6 mm.

Description

Technical Field

[0001] This invention belongs to the field of laser welding equipment technology, and in particular relates to a high-speed laser filler wire welding and online intelligent detection device and method for ultra-long thin steel strips. Background Technology

[0002] Welding ultra-long thin steel strips (over 50m in length and 2-6mm in thickness) faces technical bottlenecks such as long welding stroke, sensitivity to thermal deformation, high requirements for weld consistency, difficulty in online defect early warning, and difficulty in quality traceability. Traditional welding methods are inefficient and produce large deformations; conventional laser filler wire welding lacks pre-weld cleaning, weld tracking, dynamic monitoring and intelligent analysis of the molten pool, and is prone to defects such as porosity, incomplete penetration, and undercut; post-weld inspection relies on offline sampling, which cannot achieve online defect location, point verification, and closed-loop data traceability.

[0003] Existing technologies disclose laser filler wire welding, weld pool monitoring, weld seam tracking and X-ray inspection devices, but have not yet formed an integrated equipment system that includes pre-weld laser cleaning, real-time weld seam tracking, dual-beam high-speed welding, online intelligent detection of the weld pool, deep learning stability assessment, post-weld X-ray point verification, and full-process data traceability. This system cannot meet the industrial continuous production requirements of ultra-long thin steel strips with high speed, low deformation, high reliability and traceability.

[0004] Chinese Patent Publication No. CN 115096918 A, entitled "A High-Speed ​​Online Finished Product Defect X-ray Intelligent Inspection Equipment," uses X-ray inspection of products and employs intelligent grasping for conveyor belt loading and unloading, shortening the inspection cycle, but cannot achieve online product inspection. Chinese Patent Publication No. CN 121042877 A, entitled "An Intelligent Welding Inspection Integrated Device and Welding Inspection Method," uses a simultaneous welding and inspection mode to significantly shorten the production cycle, but cannot achieve immediate inspection when defects exist. Chinese Patent Authorization Announcement No. CN 120755509 B, entitled "A Laser Intelligent Welding Weld Quality Inspection Method and System," acquires multimodal signal data from the weld surface. Addressing issues such as insufficient adaptability to multiple materials and static parameter configuration, it uses a transfer learning framework and evolutionary algorithm to verify product quality using X-ray inspection, which is still necessary. Chinese Patent Publication No. CN116944684A, entitled "A High-Efficiency Laser-CMT Welding Method and Apparatus for Thin Plate Circumferential Welding," employs a dual-laser beam combined with CMT and TIG arc hybrid welding. The CMT and TIG arcs have relatively high heat input, which can easily cause significant deformation in thin plate welding. Chinese Patent Authorization Announcement No. CN107931835B, entitled "A Process for High-Speed ​​Laser Wire Filler Welding of High-Strength Duplex Steel Thin Plates," uses a pre-fabricated butt joint gap of 0.2mm~0.4mm, requiring high precision and unsuitable for mass production of long welds. Chinese Patent Authorization Announcement No. CN112238298 B, entitled "A Method for Large-Gap Butt Joint Welding of Aluminum Alloy Thin Plates with Oscillating Laser Wire Filler Welding," is applicable to aluminum alloys. This method uses a pre-fabricated butt joint gap of 0.8mm~1.4mm, also requiring high precision and unsuitable for mass production of long welds. Chinese Patent Publication No. CN109048059B, entitled "A Laser Scanning Wire Filler Welding Method for Thin Plates," is for non-penetration fillet welds. Chinese Patent Publication No. CN121223254A, entitled "A Wire Filler Welding Method Based on Composite Laser," is applicable to aluminum alloys. In this method, the distance between the two laser beams is small, and the first laser beam preheats the second. However, it does not mention penetration welding. Summary of the Invention

[0005] In view of this, in order to solve the problems of poor surface cleanliness, low weld tracking accuracy, unstable penetration, difficulty in early warning of defects, lack of closed-loop quality verification, and lack of data traceability in the welding process of ultra-long thin steel strips, this invention proposes a high-speed laser filler wire welding and online intelligent detection device and method for ultra-long thin steel strips. It is an integrated equipment for high-speed laser filler wire welding of ultra-long thin steel strips, online intelligent detection of the molten pool, and post-weld X-ray verification. It is suitable for continuous, efficient, and high-quality welding production of ultra-long thin steel strips with a length ≥50m and a thickness of 1~6mm in fields such as aerospace and shipbuilding.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a high-speed laser filler wire welding and online intelligent inspection device for ultra-long thin steel strip, comprising a steel strip conveying unit, a laser cleaning unit, a weld tracking unit, a dual-beam laser filler wire welding unit, a double-sided wire feeding unit, an online intelligent inspection unit for the molten pool, a deep learning control unit, a post-weld X-ray verification unit, and a data storage and traceability unit; The steel belt conveyor unit is compatible with thin steel belts with a length ≥50m and a thickness of 1~6mm, achieving constant tension conveying at 0.8~4m / min; The laser cleaning unit is located at the front end of the welding station and is used to remove oil and oxide film from the surface of the steel strip bevel. The weld tracking unit is located between the laser cleaning unit and the welding unit, and it detects the bevel profile, gap and misalignment in real time, and dynamically outputs the coordinates of the weld center. The dual-beam laser wire filler welding unit includes a front high-power laser and a rear oscillating laser. The front laser beam achieves high-speed deep penetration, while the rear laser beam oscillates to melt the welding wire and stir the molten pool. The two laser beams work together to melt the molten pool. The dual-sided wire feeding unit feeds wire symmetrically to ensure uniform weld filling. The online intelligent detection unit for the molten pool includes a high-speed camera system and an infrared thermal imaging system. Through the high-speed camera system and the infrared thermal imaging system, the keyhole shape, molten pool morphology and temperature field are collected simultaneously, and the characteristic parameters of the molten pool are extracted in real time. The deep learning control unit establishes a defect association model based on molten pool characteristics, and evaluates welding stability, provides early warning of defects, and marks the location of suspected defects in real time. The post-weld X-ray verification unit is located at the rear end of the welding station and automatically scans the marked suspected defect areas to verify the defect type, size and location. The data storage and traceability unit stores data from each stage and assigns a unique code to achieve full-process traceability.

[0007] Preferably, the power of the front laser beam is 6~10kW, the power of the rear laser beam is 2~4kW, and the distance between the two laser beams is 2~8mm.

[0008] Preferably, the rear laser beam oscillates in a circular or linear manner with an amplitude of 1-3 mm and a frequency of 50-200 Hz.

[0009] Preferably, the high-speed camera system has a camera frame rate of ≥2000fps and a spatial resolution of ≥5μm.

[0010] Preferably, the infrared thermal imaging system has a temperature measurement range of 600~2500℃ and a frame rate of ≥100fp.

[0011] Preferably, the weld seam tracking unit has an accuracy of ±0.1mm, which is suitable for gap fluctuations of 0~1.5mm.

[0012] Preferably, the deep learning control unit adopts a CNN+LSTM hybrid architecture, inputs molten pool images, temperature field data and process parameters, and automatically provides early warnings and fine-tunes laser power, wire feeding speed and beam oscillation parameters.

[0013] Preferably, the dual-beam laser wire filler welding unit further includes a protective gas device, which includes a front protective gas device and a root protective gas device. The front protective gas device is located above the welding area of ​​the workpiece, and the root protective gas device is located at the bottom of the workpiece.

[0014] Preferably, the diameter of the welding wire in the double-sided wire feeding unit is 0.8~1.6mm, the wire feeding speed is 2~8m / min, the angle between the welding wire and the welding direction is 10°~45°, and the angle between the welding wire and the steel strip is 15°~30°.

[0015] A method for high-speed laser filler wire welding and online intelligent inspection of ultra-long thin steel strips, using the above-mentioned device, includes the following steps: S1. Steel belt constant tension conveying procedure: A thin steel belt with a length ≥ 50m and a thickness of 1~6mm is clamped on the steel belt conveying unit 15, and the steel belt is continuously conveyed with constant tension at a speed of 0.8~4m / min. The position and timestamp of each unit are hard-synchronized and calibrated by an encoder. S2. Pre-welding laser cleaning procedure: 800~1200mm before the steel strip enters the welding station, the laser cleaning unit uses a pulsed fiber laser to scan and clean the bevel of the steel strip and the area 20~50mm on both sides to remove oxide film, oil and rust. S3. Real-time weld tracking steps: In the 300-500mm range after cleaning and before welding, the weld tracking unit uses a line laser sensor and a high-speed camera system to collect real-time images of the bevel contour. The image processing extracts the weld center coordinates, butt gap, and misalignment. When the gap is ≤0.2mm, the left and right positions of the filler nozzle are finely adjusted. When the gap is 0.2-0.8mm, the laser focus, wire feeding speed, and rear laser power are simultaneously finely adjusted. When the gap is 0.8-1.5mm, the system switches to a dual-sided wire feeding and rear laser enhanced stirring mode. When the misalignment is >0.5mm, the system issues a warning and reduces the speed, thereby ensuring that the laser beam and welding wire are aligned throughout the entire process and the tracking accuracy is ±0.1mm. S4. Dual-beam laser filler wire welding procedure: Using the weld center coordinates as a reference, start the dual-beam laser filler wire welding unit. The front laser beam focuses on the root of the weld with a power of 6~10kW and a small spot to achieve deep penetration and form a stable keyhole. The rear laser beam covers the upper part of the molten pool with a power of 2~4kW and a large spot, and oscillates in a circular or linear manner with an amplitude of 1~3mm and a frequency of 50~200Hz to melt the welding wire fed on both sides and stir the molten pool. At the same time, high-purity argon gas is introduced through the front shielding gas device and the root shielding gas device for full protection. S5. Online Intelligent Detection Steps for the Molten Pool: During the welding process, the online intelligent detection unit for the molten pool, integrated 50-150mm behind the welding head, synchronously collects data. A high-speed camera system captures the keyhole morphology and the surface morphology of the molten pool, while an infrared thermal imaging system captures the temperature field distribution of the molten pool. In real time, characteristic parameters of the molten pool, such as keyhole diameter, molten pool aspect ratio, edge fluctuation amplitude, oscillation frequency, temperature peak value, and temperature gradient, are extracted. Each characteristic data is accompanied by a millisecond-level timestamp and the absolute coordinates of the steel strip. S6. Deep Learning Stability Assessment and Defect Early Warning Steps: Input the molten pool feature parameters, temperature field data, and current process parameters into the CNN+LSTM hybrid architecture model of the deep learning control unit. The CNN extracts the spatial morphological features of the keyhole and the molten pool, and the LSTM captures their temporal evolution. The classification results and defect probabilities of the current welding state are output. When any defect probability is ≥30%, the absolute coordinates of the steel strip, timestamp, corresponding molten pool image, infrared image, feature parameters, and process parameters are packaged to generate defect marking information and written into the coordinate list to be re-inspected. At the same time, when the defect probability is <30% but the features deviate from the normal range, the laser power is automatically fine-tuned by ±5%, the wire feed speed by ±10%, or the oscillation parameters to stabilize the molten pool. S7. Post-weld X-ray point verification steps: After the steel strip continues to move forward through a 10-20m natural cooling section, it arrives at the post-weld X-ray verification unit; this unit receives the list of coordinates to be re-inspected, drives the X-ray generator and flat panel detector to move to the corresponding coordinates, and performs point scanning imaging on the marked area; the type, size and location of porosity, slag inclusion, incomplete penetration and cracks are identified by an automatic defect recognition algorithm, and the weld quality is graded according to GB / T 22085.1-2008 standard; if it is determined to be a Class III or Class IV defect exceeding the standard, the verification result is fed back to the deep learning control unit to incrementally update the model weights; S8. Data Storage and Full-Process Traceability Steps: The data storage and traceability unit will structure and bind all the data generated in steps S2 to S7, including laser cleaning parameters, original contour data of weld seam tracking and centering correction curves, welding process parameter time series, high-speed camera frame sequence and molten pool feature value sequence, infrared thermographic frame sequence, deep learning evaluation results, X-ray images and inspection reports, according to the unique product code of the steel strip. This will enable reverse traceability of quality and forward iteration of process at any time and any location from pre-weld preparation, welding process to post-weld inspection.

[0016] Compared with existing technologies, the beneficial effects of the high-speed laser filler wire welding and online intelligent inspection device and method for ultra-long thin steel strips described in this invention are: 1. High adaptability of ultra-long steel belt: constant tension conveying + high-precision weld seam tracking solves the problems of long-stroke welding deformation and centering deviation, and the welding length can reach hundreds of meters.

[0017] 2. Pre-welding cleaning + dual-beam high-quality and high-efficiency welding: Laser cleaning eliminates the causes of surface defects; front side penetration and rear side stirring enhance the stability of the keyhole in the eutectic pool; welding speed is 0.8~4m / min; narrow heat-affected zone and minimal deformation.

[0018] 3. Online intelligent inspection of the molten pool: High-speed imaging + infrared thermography are monitored simultaneously to accurately capture the dynamic evolution of the molten pool, extract key features in real time, realize the visualization of the welding process, and combine deep learning models to predict the occurrence of defects and mark their locations.

[0019] 4. X-ray fixed-point verification closed loop: Precise re-inspection of suspected defective areas, confirmation of defects and feedback to the process, reducing the missed detection rate, improving detection efficiency, and meeting the stringent quality requirements of high-end products.

[0020] 5. This invention enables automated welding production of ultra-long thin steel strips with high speed, low deformation, high consistency, and traceability. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a schematic diagram of the structure of the high-speed laser filler wire welding and online intelligent detection device for ultra-long thin steel strips described in this invention; In the diagram: 1-Front laser beam, 2-Rear laser beam, 3-Left wire feeder, 4-Right wire feeder, 5-Molten pool generated by the front laser, 6-Molten pool generated by the rear laser, 7-Front shielding gas device, 8-Root shielding gas device, 9-Laser cleaning unit, 10-High-speed camera system, 11-Infrared thermal imaging system, 12-Weld seam tracking unit, 13-Post-weld X-ray inspection unit, 14-Workpiece, 15-Steel strip conveyor unit. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the drawings, and not all of them. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein is for the purpose of describing specific embodiments only and is not intended to limit the invention.

[0023] See Figure 1 This embodiment describes a high-speed laser filler wire welding and online intelligent inspection device for ultra-long thin steel strips, comprising a steel strip conveying unit 15, a laser cleaning unit 9, a weld tracking unit 12, a dual-beam laser filler wire welding unit, a double-sided wire feeding unit, an online intelligent inspection unit for the molten pool, a deep learning control unit, a post-weld X-ray verification unit 13, and a data storage and traceability unit. The steel belt conveying unit 15 adopts a servo-driven roller conveyor and a tension control system, which is suitable for thin steel belts with a length ≥50m and a thickness of 1~6mm, and a welding speed of 0.8~4m / min, so as to realize continuous, stable and constant tension conveying of steel belts and suppress welding deformation.

[0024] The laser cleaning unit 9 is located at the front end of the welding station and includes a pulsed fiber laser, a scanning galvanometer, a focusing lens, and a fume purification device. Before welding, it removes oxide film, oil, and rust from the bevel of the steel strip and the 20-50mm area on both sides. The cleaning spot diameter is 5-15mm, and the scanning speed is 1-3m / s to ensure a clean welding interface and suppress porosity and inclusion defects from the source.

[0025] The weld seam tracking unit 12 is located between the laser cleaning unit 9 and the welding unit. It includes a line laser sensor, a high-definition industrial camera, and a real-time image processing module. It collects the bevel profile, butt gap, and misalignment in real time, dynamically outputs the weld seam center coordinates, and feeds them back to the welding unit. It adjusts the laser beam focus and filler wire position in real time, with a tracking accuracy of ±0.1mm. It is adaptable to gap fluctuations of 0~1.5mm and ensures the centering accuracy of long weld seams.

[0026] The dual-beam laser wire-filling welding unit includes a front high-power laser, a rear oscillating laser, a dual-sided wire feeding mechanism, an optical path adjustment component, and a protective gas device. The front laser beam 1 achieves high-speed deep penetration, while the rear laser beam 2 oscillates to melt the welding wire and stir the molten pool. The two laser beams work synergistically in the fusion pool. Front laser beam 1: small spot size, high energy density (power 6~10kW), focused on the weld root to achieve high-speed deep penetration, ensuring complete weld penetration of the steel strip and good root formation. Rear laser beam 2: large spot size, high-frequency oscillation (power 2~4kW), covering the upper part of the molten pool, melting the welding wire and stirring the molten pool, improving keyhole stability and suppressing porosity and cracks. The distance between the two laser beams is 2~8mm, and the fusion pool works synergistically. The rear laser beam 2 uses circular / linear oscillation with an amplitude of 1~3mm and a frequency of 50~200Hz. The protective gas is high-purity argon with a flow rate of 15~25L / min, achieving full protection on both sides.

[0027] The dual-sided wire feeding unit includes a symmetrical wire feeder, a wire feeding nozzle, a guide mechanism, and a wire feeding stability control module; the wire diameter is 0.8~1.6mm, and the wire feeding speed is 2~8m / min; the angle between the wire and the welding direction is 10°~45°, and the angle with the steel strip is 15°~30°. The symmetrical and uniform wire feeding on both sides ensures uniform weld filling and symmetrical formation, improving the consistency of long welds.

[0028] The online intelligent detection unit for the molten pool includes a high-speed camera system 10 and an infrared thermal imaging system 11. Through the high-speed camera system 10 and the infrared thermal imaging system 11, the keyhole shape, molten pool morphology and temperature field are collected simultaneously, and the characteristic parameters of the molten pool are extracted in real time. The online intelligent detection unit for the molten pool includes a high-speed camera system 10, an infrared thermal imaging system 11, a narrowband filter, an auxiliary light source, an image acquisition card, and a real-time feature extraction module. The high-speed camera system 10 has a frame rate ≥2000fps and a spatial resolution ≥5μm, and acquires keyhole morphology, molten pool surface morphology, and molten pool boundary evolution in real time. The infrared thermal imaging system 11 has a temperature measurement range of 600~2500℃ and a frame rate ≥100fps, and simultaneously acquires molten pool temperature field distribution, temperature gradient, and keyhole thermal radiation characteristics. It extracts key features such as keyhole diameter, molten pool aspect ratio, temperature peak, edge fluctuation amplitude, and oscillation frequency in real time, realizing intelligent monitoring of the dynamic evolution of the dual-beam co-melting pool throughout the entire process.

[0029] The deep learning control unit incorporates a molten pool feature extraction network, a welding stability assessment model, a defect correlation prediction model, and a process parameter optimization model (CNN+LSTM hybrid architecture). Using molten pool images, temperature field data, and process parameters as input, it automatically identifies normal molten pools, keyhole collapses, severe molten pool fluctuations, incomplete penetration, and porosity. It establishes a quantitative correlation model between molten pool features and defects such as porosity, cracks, incomplete penetration, and undercut, and outputs the presence of defects in real time. For suspected defect locations, it automatically records the three-dimensional coordinates, timestamps, and molten pool feature data, marking them as areas to be re-inspected by X-ray.

[0030] The deep learning control unit provides real-time warnings of defects and marks risk areas. It can automatically issue warnings and fine-tune laser power, wire feed speed, and beam oscillation parameters by inputting molten pool images, temperature field data, and process parameters.

[0031] The post-weld X-ray verification unit 13, located at the rear end of the welding station, includes an X-ray generator, a flat panel detector, an automatic scanning mechanism, an imaging analysis module, and a defect identification and annotation module. It receives the coordinates of suspected defects marked by the deep learning unit, automatically and accurately locates and scans the defects at specific points, automatically identifies and annotates the defect type, size, and location, and outputs a verification report, achieving a closed loop of online defect location, point-to-point verification, and result feedback. It automatically scans the marked suspected defect areas at specific points to verify the defect type, size, and location, and re-inspects suspected defects at specific points.

[0032] The data storage and traceability unit connects to various sensors, detection modules, and controllers, storing laser cleaning parameters, weld tracking data, welding process parameters, molten pool detection images, deep learning evaluation results, X-ray verification images, and reports throughout the entire process. It assigns a unique identification code to each steel strip, enabling full-process data traceability from pre-welding preparation and welding process to post-welding inspection, supporting process iteration optimization and quality accountability.

[0033] Example 1: The above-mentioned device was used to weld a 500m long, 3mm thick 304 stainless steel strip: The equipment is arranged along the direction of steel belt movement as follows: steel belt conveying unit 15, laser cleaning unit 9, weld seam tracking unit 12, dual-beam laser wire feeding welding unit, double-sided wire feeding unit, online intelligent detection unit for molten pool, and post-weld X-ray verification unit 13. The whole machine adopts servo synchronous bus control, and the linkage response time of each unit does not exceed 20 milliseconds, which can meet the requirements for continuous and stable production of 100-meter-long weld seams.

[0034] The steel strip conveyor unit 15 adopts a three-section structure of active unwinding, tension buffer, and servo traction roller conveyor. The unwinding end is equipped with a magnetic powder brake to adjust the unwinding tension in real time, with a tension control accuracy of ±5%, effectively preventing the steel strip from stretching, deforming, or running off-track. The buffer section is equipped with floating rollers and tension sensors to absorb fluctuations in unwinding speed. The welding and inspection areas use paired servo-driven roller groups with wear-resistant rubber coating. The roller diameter is 120mm~200mm and the roller spacing is 400mm~800mm, providing stable clamping force. The welding speed is continuously adjustable within the range of 0.8m / min~4.0m / min, and it is suitable for ultra-long thin steel strips with a thickness of 2mm~6mm, a width of 1.2m~2.0m, and a length of 100m~1000m. The entire conveyor line is hard synchronized with the welding and inspection modules through encoders to achieve precise alignment of position, speed, and timestamp, providing a unified spatial reference for molten pool characteristics, defect coordinates, and X-ray scanning positions.

[0035] The laser cleaning unit 9 is installed 800mm~1200mm in front of the welding station. The entire gantry frame spans above the steel strip and can be electrically adjusted along the width direction. It is equipped with a 1064nm pulsed fiber laser with an average power of 150W~300W, a pulse width of 100 to 500 nanoseconds, a pulse frequency of 50kHz~200kHz, and a scanning speed of 1 m / s~3m / s. A single cleaning covers the bevel and an area of ​​20mm~50mm on both sides. It is equipped with a negative pressure fume purification device to effectively remove cleaning dust and metal vapor, protect optical lenses, and improve the on-site environment. After cleaning, the surface of the steel strip is free of oil, oxide film, and obvious color difference, providing a clean interface for welding.

[0036] The weld seam tracking unit 12 is located 300mm~500mm behind the laser cleaning unit. It uses an integrated sensor combining a line laser and a high-definition camera, mounted on a precision-adjustable three-dimensional slide. The horizontal adjustment range is ±20mm with a resolution of 0.01mm, the pitch angle is ±15° with a resolution of 0.1°, and the working distance is 150mm~300mm. The sensor uses a 650nm wavelength, 50mW power, and 0.1mm~0.3mm line laser, paired with a resolution of 2048×1536 and a frame rate of 50~100fps. An industrial camera with PS, global shutter, and strong light suppression and polarization filtering functions is used to capture the I-groove contour in real time. The image processing module outputs the weld center coordinates, the butt gap (0mm~1.5mm), the left and right misalignment (0mm~1.0mm), and the straightness of the groove edge within 20mm. The tracking control logic automatically adjusts the filler nozzle position, laser focus, wire feed speed, and rear laser power according to the gap change. When the gap change is less than 0.2mm, only the left and right positions of the filler nozzle are finely adjusted. When the gap is within 0.2mm~0.8mm, the laser focus, wire feed speed, and rear laser power are finely adjusted simultaneously. When the gap is within 0.8mm~1.5mm, it automatically switches to a dual-sided wire feed and rear laser enhanced stirring mode. When the misalignment is greater than 0.5mm, the system issues a warning and reduces the speed to prevent single-sided incomplete penetration. The entire process is continuously tracked without frame loss or drift, with a tracking accuracy of ±0.1mm, ensuring that the laser beam and welding wire of the 100-meter-long weld are always aligned.

[0037] The dual-beam laser wire-filling welding unit is positioned 400mm~600mm behind the tracker. It is equipped with two independent fiber lasers, an optical path system, a oscillation mechanism, double-sided wire feed nozzles, and a protective gas shield. The front laser is a 6kW~10kW fiber laser with a wavelength of 1070nm, operating in single-mode or low-order mode, with a spot diameter of 0.3mm, a defocusing amount of 0mm, and an incident angle of 85 degrees. It is used to form a deep-melt keyhole, ensuring 100% penetration of the steel strip and a full, non-collapsed root formation. The rear laser is a 2kW~4kW fiber laser with a wavelength of 1070nm, a spot diameter of 1.5~3.0mm, a defocusing amount of +5mm~+15mm, and an incident angle of 80°. It can oscillate in a circular or linear manner with an oscillation amplitude of 1~3... With a wavelength of 50-200 Hz, the laser beams are used to stabilize the melting wire, stir the molten pool, suppress keyhole collapse, and reduce porosity and cracks. The beam spacing is 2-8 mm, and the beams are symmetrically distributed on both sides of the weld centerline at 0-1 mm. The superposition of the two laser beams forms a single stable molten pool, which significantly improves the stability of the keyhole.

[0038] The dual-sided wire feeding system is symmetrically arranged on both sides, and is compatible with 304 and 316L stainless steel with diameters of 0.8mm to 1.6mm. The wire feeding speed is 2m / min to 8m / min, the wire extension is 10mm to 15mm, the angle between the wire feeding nozzle and the welding direction is 10 to 45°, the angle with the steel strip is 15 to 30°, and the distance between the end of the welding wire and the center of the laser spot on the rear side is 0 to 4mm, so as to accurately feed the wire into the molten pool.

[0039] The protective gas system adopts a combination of front coaxial and side blowing protection and back bottom gas groove protection. The protective gas is high-purity argon with a purity of 99.99%, with a front flow rate of 15~25L / min and a back flow rate of 15~25L / min, which isolates the molten pool and high-temperature weld from air throughout the process to prevent oxidation and nitriding.

[0040] The online intelligent detection unit for the weld pool is integrated 50-150 mm behind the welding head. It adopts a coaxial arrangement of a high-speed industrial camera, an infrared thermal imager, a narrow-band filter, and an auxiliary cold light source. A water-cooled protective cover isolates it from high-temperature splashes and fumes. The high-speed industrial camera has a resolution of 2048×1536, a frame rate of 2000-5000 frames per second, and an exposure time of 1-5 microseconds. It is paired with an 880±20 nm narrow-band filter to suppress arc light and laser stray light, with a field of view of 15×12. The system covers the keyhole and molten pool area with a mm coverage area, and outputs keyhole diameter, molten pool aspect ratio, edge contour, spatter distribution, and surface fluctuation frequency in real time. The infrared thermal imager uses the 800~1400nm band, a temperature measurement range of 600~2500 degrees Celsius, a frame rate of 100~200fps, and a resolution of 640×480. It outputs the highest temperature of the molten pool, temperature gradient, keyhole thermal radiation intensity, high-temperature zone area, and cooling rate in real time. The image processing unit completes the analysis of one frame within 10ms, extracting more than 20 molten pool features such as keyhole diameter, molten pool length and width, edge fluctuation standard deviation, oscillation frequency, highest temperature, and temperature gradient. All feature data are accompanied by millisecond-level timestamps and absolute coordinates of the steel strip, and are synchronously stored in the database, realizing full-process intelligent monitoring of the dynamic evolution process of the dual-beam eutectic pool.

[0041] The deep learning control unit adopts an industrial-grade edge computing platform with no less than 16GB of video memory, and has a built-in hybrid architecture model of CNN and LSTM. The training dataset contains more than 150,000 sets of images, temperature field data and process parameters of normal molten pool, porosity initiation, keyhole collapse, molten pool fluctuation, incomplete penetration, and undercut. The defect recognition accuracy is no less than 98.5%. The real-time inference process takes molten pool image, infrared thermogram and current process parameters as input, extracts spatial features through CNN and captures temporal evolution through LSTM, and outputs information on four types of defects: porosity, cracks, incomplete penetration and undercut. The system automatically records the absolute coordinates of the steel strip, millisecond-level timestamp, molten pool image, infrared image, feature parameters and process parameters for areas with a defect probability of no less than 30%, generates unique defect information and marks it as an area to be re-inspected by X-ray. The coordinate data is sent to the X-ray verification unit in real time to realize intelligent assessment of welding process stability and accurate defect location and early warning.

[0042] The post-weld X-ray verification unit 13 is located 10-20m after the welding cooling section. It uses an X-ray machine paired with a flat panel detector, capable of moving 0-2000mm along the width of the steel strip and moving synchronously with it. The X-ray tube voltage is 80-150 kV, current is 5-20mA, scanning speed is 0.5-2.0m, positioning accuracy is ±0.5 mm, and imaging area is 400×300 mm. The fixed-point verification process first receives defect information and a coordinate list from the deep learning unit, then moves to the target coordinates for precise alignment, activates the X-ray for fixed-point scanning, and acquires images of the weld interior. An automatic defect recognition algorithm identifies defects such as porosity, slag inclusions, incomplete penetration, and cracks, and outputs the defect type, size, and location, in accordance with GB / T... The 22085.1-2008 standard classifies weld quality into grades I, II, III, and IV. After imaging inspection, X-ray images, inspection reports, and judgment results are bound to defect information and stored in the database. If grade III or IV defects are found to exceed the standard, the results are fed back to the deep learning unit to update the model and optimize subsequent process parameters, forming a closed-loop control system of online monitoring, intelligent early warning, fixed-point verification, and process optimization.

[0043] The data storage and traceability unit stores all data from the sensors, cameras, and controllers of the entire machine in real time. This includes all operational data such as laser cleaning parameters, weld tracking data, welding process parameters, molten pool images, infrared images, deep learning evaluation results, X-ray images and reports, steel strip position, timestamps, ambient temperature and humidity, and shielding gas flow rate. Each roll of steel strip is assigned a product number, and by querying the number, information such as production time, equipment number, process parameters, cleaning records, tracking curves, molten pool characteristic curves, defect markings, X-ray inspection reports, and operators can be fully traced, meeting the stringent requirements for quality control and accountability traceability of high-end products.

[0044] Example 2: In a specific application, taking a 500m long and 3mm thick 304 stainless steel strip as an example, 0.8mm diameter 308L welding wire was used. The laser cleaning power was 200W, the scanning speed was 2m / s, and the spot diameter was 10mm. The weld seam tracking was adapted to a gap of 0~1mm with a tracking accuracy of ±0.1mm. The welding speed was 3.0m / min. The front laser power was 7kW, the defocusing amount was 0mm, and the incident angle was 85°. The rear laser power was 3kW, the defocusing amount was +10mm, the incident angle was 80°, the circular swing amplitude was 2mm, and the frequency was 120Hz. The distance between the two beams was 4mm, the wire feeding speed on both sides was 5m / min, and the shielding gas flow rate was 20L / min. The molten pool was inspected using a 3000Fps high-speed camera and a 150Fps infrared thermal imager to extract more than ten feature parameters in real time. Two porosity risk areas were marked. X-ray fixed-point scanning confirmed a micro-porosity of 0.3mm in diameter, which was judged as Class II qualified. Finally, the weld seam was smooth without undercut, and the steel strip deformation was less than 1mm. The weld quality meets the Class B standard for mm / m and X-ray flaw detection. The weld structure is dense and the impact toughness meets the standard. No grinding treatment is required after welding. This fully verifies the advanced nature and practicality of the equipment of this invention in high-speed welding of ultra-long thin steel strips, online intelligent detection and post-weld quality verification.

[0045] A method for inspection using high-speed laser filler wire welding of ultra-long thin steel strips and an online intelligent inspection device: S1. Steel strip constant tension conveying step: A thin steel strip with a length ≥50m and a thickness of 1~6mm is clamped on the steel strip conveying unit 15. The steel strip is continuously conveyed at a constant tension at a speed of 0.8~4m / min by a servo-driven roller conveyor and tension control system. The position and timestamp of each unit are hard-synchronized by an encoder. S2. Pre-welding laser cleaning steps: 800~1200mm before the steel strip enters the welding station, the laser cleaning unit 9 uses a pulsed fiber laser to scan and clean the bevel of the steel strip and the area 20~50mm on both sides to remove oxide film, oil and rust. The cleaning spot diameter is 5~15mm, the scanning speed is 1~3m / s, and the cleaning power is 150~300W. S3. Real-time weld tracking steps: In the 300~500mm range after cleaning and before welding, the weld tracking unit 12 uses a line laser sensor and a high-speed camera system 10 to collect bevel contour images in real time. The weld center coordinates, butt gap, and misalignment are extracted through image processing. When the gap is ≤0.2mm, the left and right positions of the filler nozzle are finely adjusted. When the gap is 0.2~0.8mm, the laser focus, wire feeding speed, and rear laser power are simultaneously finely adjusted. When the gap is 0.8~1.5mm, the system switches to a dual-sided wire feeding and rear laser enhanced stirring mode. When the misalignment is >0.5mm, the system issues a warning and reduces the speed, thereby ensuring that the laser beam and welding wire are aligned throughout the process and the tracking accuracy is ±0.1mm. S4. Dual-beam laser filler wire welding steps: Using the weld center coordinates as a reference, start the dual-beam laser filler wire welding unit. The front laser beam 1, with a power of 6~10kW and a small spot, is focused on the root of the weld to achieve deep penetration and form a stable keyhole. The rear laser beam 2, with a power of 2~4kW and a large spot, covers the upper part of the molten pool and oscillates in a circular or linear manner with an amplitude of 1~3mm and a frequency of 50~200Hz to melt the welding wire fed on both sides and stir the molten pool. The two laser beams are spaced 2~8mm apart and work together in the eutectic pool. At the same time, high-purity argon gas is introduced through the front shielding gas device and the root shielding gas device for full protection. S5. Online Intelligent Detection Steps for the Molten Pool: During the welding process, the online intelligent detection unit for the molten pool, integrated 50-150mm behind the welding head, synchronously collects data. A high-speed camera system captures the keyhole shape and molten pool surface morphology at a frame rate of ≥2000fps with a narrow-band filter, while an infrared thermal imaging system captures the molten pool temperature field distribution at a frame rate of ≥100fps. Molten pool characteristic parameters such as keyhole diameter, molten pool aspect ratio, edge fluctuation amplitude, oscillation frequency, temperature peak, and temperature gradient are extracted in real time. Each characteristic data point is accompanied by a millisecond-level timestamp and the absolute coordinates of the steel strip. S6. Deep Learning Stability Assessment and Defect Early Warning Steps: Input the molten pool feature parameters, temperature field data, and current process parameters into the CNN+LSTM hybrid architecture model of the deep learning control unit. The CNN extracts the spatial morphological features of the keyhole and molten pool, and the LSTM captures their temporal evolution. The classification results and defect probabilities of the current welding state are output. The identified states include normal molten pool, porosity initiation, keyhole collapse, severe molten pool fluctuation, incomplete penetration, and undercut. When the probability of any defect is ≥30%, the absolute coordinates of the steel strip, timestamp, corresponding molten pool image, infrared image, feature parameters, and process parameters are packaged to generate defect marking information and written into the coordinate list to be re-inspected. At the same time, when the defect probability is <30% but the features deviate from the normal range, the laser power is automatically fine-tuned by ±5%, the wire feed speed by ±10%, or the oscillation parameters to stabilize the molten pool. S7. Post-weld X-ray point verification steps: After the steel strip continues to move forward through a 10-20m natural cooling section, it arrives at the post-weld X-ray verification unit 13; this unit receives the list of coordinates to be re-inspected, drives the X-ray generator and flat panel detector to move to the corresponding coordinates, and performs point scanning imaging on the marked area with a tube voltage of 80-150kV and a tube current of 5-20mA; the type, size and location of porosity, slag inclusion, incomplete penetration and cracks are identified by an automatic defect identification algorithm, and the weld quality is graded according to GB / T 22085.1-2008 standard; if it is determined to be a Class III or Class IV defect exceeding the standard, the verification result is fed back to the deep learning control unit to incrementally update the model weights, forming a closed-loop control of "online monitoring → intelligent early warning → point verification → parameter optimization"; S8. Data Storage and Full-Process Traceability Steps: The data storage and traceability unit will structure and bind all the data generated in steps S2 to S7—including laser cleaning parameters, original weld seam tracking contour data and centering correction curves, welding process parameter time series, high-speed camera frame series and molten pool feature value series, infrared thermographic frame series, deep learning evaluation results, X-ray images and inspection reports—according to the unique product code of the steel strip. This will enable reverse quality traceability and forward process iteration at any time and location from pre-weld preparation, welding process to post-weld inspection.

[0046] The embodiments of the present invention disclosed above are merely illustrative of the invention. These embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A high-speed laser filler wire welding and online intelligent inspection device for ultra-long thin steel strips, characterized in that: It includes a steel strip conveying unit (15), a laser cleaning unit (9), a weld tracking unit (12), a dual-beam laser wire feeding welding unit, a double-sided wire feeding unit, an online intelligent detection unit for the molten pool, a deep learning control unit, a post-weld X-ray verification unit (13), and a data storage and traceability unit; The steel belt conveyor unit (15) is compatible with thin steel belts with a length ≥50m and a thickness of 1~6mm, and can achieve constant tension conveying of 0.8~4m / min; The laser cleaning unit (9) is located at the front end of the welding station and is used to remove oil and oxide film from the surface of the steel strip bevel. The weld tracking unit (12) is located between the laser cleaning unit (9) and the welding unit, and detects the bevel profile, gap and misalignment in real time, and dynamically outputs the center coordinates of the weld. The dual-beam laser filler wire welding unit includes a front high-power laser and a rear oscillating laser. The front laser beam (1) achieves high-speed deep penetration, and the rear laser beam (2) oscillates to melt the welding wire and stir the molten pool. The two laser beams work together to melt the molten pool. The dual-sided wire feeding unit feeds wire symmetrically to ensure uniform weld filling. The online intelligent detection unit for the molten pool includes a high-speed camera system (10) and an infrared thermal imaging system (11). Through the high-speed camera system (10) and the infrared thermal imaging system (11), the keyhole shape, molten pool morphology and temperature field are collected simultaneously, and the characteristic parameters of the molten pool are extracted in real time. The deep learning control unit establishes a defect association model based on molten pool characteristics, and evaluates welding stability, provides early warning of defects, and marks the location of suspected defects in real time. The post-weld X-ray verification unit (13) is located at the rear end of the welding station and automatically scans the marked suspected defect areas to verify the defect type, size and location. The data storage and traceability unit stores data from each stage and assigns a unique code to achieve full-process traceability.

2. The high-speed laser filler wire welding and online intelligent inspection device for ultra-long thin steel strips according to claim 1, characterized in that: The power of the front laser beam (1) is 6~10kW, the power of the rear laser beam (2) is 2~4kW, and the distance between the two laser beams is 2~8mm.

3. The high-speed laser filler wire welding and online intelligent inspection device for ultra-long thin steel strips according to claim 1, characterized in that: The rear laser beam (2) oscillates in a circular or linear manner with an amplitude of 1-3 mm and a frequency of 50-200 Hz.

4. The high-speed laser filler wire welding and online intelligent inspection device for ultra-long thin steel strips according to claim 1, characterized in that: The high-speed camera system (10) has a camera frame rate of ≥2000fps and a spatial resolution of ≥5μm.

5. The high-speed laser filler wire welding and online intelligent inspection device for ultra-long thin steel strips according to claim 1, characterized in that: The infrared thermal imaging system (11) has a temperature measurement range of 600~2500℃ and a frame rate of ≥100fp.

6. The high-speed laser filler wire welding and online intelligent inspection device for ultra-long thin steel strips according to claim 1, characterized in that: The weld seam tracking unit (12) has an accuracy of ±0.1mm and is suitable for gap fluctuations of 0~1.5mm.

7. The high-speed laser filler wire welding and online intelligent inspection device for ultra-long thin steel strips according to claim 1, characterized in that: The deep learning control unit adopts a CNN+LSTM hybrid architecture. It takes molten pool images, temperature field data and process parameters as input, and automatically issues warnings and fine-tunes laser power, wire feeding speed and beam oscillation parameters.

8. The high-speed laser filler wire welding and online intelligent inspection device for ultra-long thin steel strips according to claim 1, characterized in that: The dual-beam laser filler wire welding unit also includes a protective gas device, which includes a front protective gas device (7) and a root protective gas device (8). The front protective gas device (7) is located above the welding part of the workpiece (14), and the root protective gas device (8) is located at the bottom of the workpiece (14).

9. The high-speed laser filler wire welding and online intelligent inspection device for ultra-long thin steel strips according to claim 1, characterized in that: The diameter of the welding wire in the double-sided wire feeding unit is 0.8~1.6mm, and the wire feeding speed is 2~8m / min; the angle between the welding wire and the welding direction is 10°~45°, and the angle between the welding wire and the steel strip is 15°~30°.

10. A method for high-speed laser filler wire welding and online intelligent inspection of ultra-long thin steel strips, employing the apparatus described in any one of claims 1-9, characterized in that, Includes the following steps: S1. Steel belt constant tension conveying steps: A thin steel belt with a length ≥ 50m and a thickness of 1~6mm is clamped on the steel belt conveying unit (15), and the steel belt is continuously conveyed with constant tension at a speed of 0.8~4m / min. The position and timestamp of each unit are hard-synchronized by an encoder. S2. Pre-welding laser cleaning steps: 800~1200mm before the steel strip enters the welding station, the laser cleaning unit (9) uses a pulsed fiber laser to scan and clean the steel strip bevel and the area 20~50mm on both sides to remove oxide film, oil and rust. S3. Real-time tracking steps for weld seam: In the 300~500mm range between cleaning and welding, the weld seam tracking unit (12) uses a line laser sensor and a high-speed camera system (10) to collect the bevel contour image in real time. The center coordinates of the weld seam, the butt gap and the misalignment are extracted through image processing. When the gap is ≤0.2mm, the position of the filler nozzle is finely adjusted. When the gap is 0.2~0.8mm, the laser focus, wire feeding speed and the power of the rear laser are finely adjusted simultaneously. When the gap is 0.8~1.5mm, the double-sided wire feeding and rear laser enhanced stirring mode are switched. When the misalignment is >0.5mm, the system issues a warning and reduces the speed, thereby ensuring that the laser beam and the welding wire are aligned throughout the process and the tracking accuracy is ±0.1mm. S4. Dual-beam laser filler wire welding steps: Based on the center coordinates of the weld, start the dual-beam laser filler wire welding unit. The front laser beam (1) focuses on the root of the weld with a power of 6~10kW and a small spot to achieve deep penetration and form a stable keyhole. The rear laser beam (2) covers the upper part of the molten pool with a power of 2~4kW and a large spot. It makes circular or linear oscillations with an amplitude of 1~3mm and a frequency of 50~200Hz to melt the welding wire fed on both sides and stir the molten pool. At the same time, high-purity argon gas is introduced through the front protective gas device (7) and the root protective gas device (8) for full protection. S5. Online intelligent detection steps for molten pool: During the welding process, the online intelligent detection unit for molten pool, integrated at 50~150mm behind the welding head, synchronously collects data. The high-speed camera system (10) collects the keyhole shape and the surface morphology of the molten pool, and the infrared thermal imaging system (11) collects the temperature field distribution of the molten pool. The keyhole diameter, aspect ratio of the molten pool, edge fluctuation amplitude, oscillation frequency, temperature peak, temperature gradient and other characteristic parameters of the molten pool are extracted in real time. Each characteristic data is accompanied by a millisecond-level timestamp and the absolute coordinates of the steel strip. S6. Deep Learning Stability Assessment and Defect Early Warning Steps: Input the molten pool feature parameters, temperature field data, and current process parameters into the CNN+LSTM hybrid architecture model of the deep learning control unit. The CNN extracts the spatial morphological features of the keyhole and the molten pool, and the LSTM captures their temporal evolution. The classification results and defect probabilities of the current welding state are output. When any defect probability is ≥30%, the absolute coordinates of the steel strip, timestamp, corresponding molten pool image, infrared image, feature parameters, and process parameters are packaged to generate defect marking information and written into the coordinate list to be re-inspected. At the same time, when the defect probability is <30% but the features deviate from the normal range, the laser power is automatically fine-tuned by ±5%, the wire feed speed by ±10%, or the oscillation parameters to stabilize the molten pool. S7. Post-weld X-ray point verification steps: After the steel strip continues to move forward through a 10~20m natural cooling section, it arrives at the post-weld X-ray verification unit (13); This unit receives the list of coordinates to be re-inspected, drives the X-ray generator and flat panel detector to move to the corresponding coordinates, and performs point scanning imaging on the marked area; The automatic defect recognition algorithm identifies the type, size and location of porosity, slag inclusion, incomplete penetration and cracks, and classifies the weld quality; If it is determined to be an excessive defect, the verification result is fed back to the deep learning control unit to incrementally update the model weights; S8. Data storage and full-process traceability steps: The data storage and traceability unit will store and associate all the data generated in steps S2 to S7 in a structured manner according to the unique product code of the steel strip, so as to realize reverse traceability of quality and forward iteration of process at any time and any location from pre-welding preparation, welding process to post-welding inspection.

Images