An automatic welding device for automobile sheet metal processing

Abstract

The application relates to the technical field of automobile sheet metal machining, in particular to an automatic welding device for automobile sheet metal machining, which comprises a base; a pneumatic clamping device is arranged at the top of the outer wall of the base; a supporting column is fixedly connected to the top of the outer wall of the base; a driving device is fixedly connected to the top of the outer wall of the supporting column through a fixing block; a threaded rotating shaft is arranged at the output end of the driving device; when the threaded rotating shaft is rotated to drive the threaded plate to descend, the threaded rod is rotated through a linkage mechanism, the moving plate, the square block, the rotating rod, the circular plate and the cleaning layer are moved, and the workpiece is cleaned; because the workpiece is cleaned, the welding effect of the welding device is better; because the cleaning mechanism and the welding mechanism realize cooperative work through the linkage mechanism, independent power sources and driving control systems need not be respectively arranged for the two, the hardware cost and the occupied space are reduced, the electrical wiring, control programming and maintenance of the whole system are more simple, and the maintenance cost and difficulty are reduced.

Description

Technical Field

[0001] This invention relates to the field of automotive sheet metal processing technology, specifically to an automated welding device for automotive sheet metal processing. Background Technology

[0002] Automated welding equipment for automotive sheet metal processing is a core process equipment in modern automobile manufacturing. The demand for automated welding in automotive sheet metal processing stems from the automotive industry's pursuit of ultimate efficiency, absolute consistency, and high flexibility, which are precisely what traditional manual welding cannot achieve. This makes automated welding equipment for automotive sheet metal processing an irreplaceable standard configuration and a key support for the automotive industry to move towards efficient, precise, flexible, and intelligent manufacturing.

[0003] In the existing technology, most automated welding equipment used in automotive sheet metal processing has a separate cleaning mechanism and a welding mechanism. This means that each of the separate cleaning mechanism and welding mechanism needs to be equipped with an independent power source and drive control system. This not only increases the hardware cost and space occupied by the power source equipment itself, but also significantly increases the complexity of electrical wiring, control programming and maintenance of the entire system, thus increasing maintenance costs and difficulty. Summary of the Invention

[0004] The purpose of this invention is to address the problem that most automated welding devices used in automotive sheet metal processing have independently set cleaning and welding mechanisms, requiring each mechanism to be equipped with its own power source and drive control system. This not only increases the hardware cost and space occupied by the power source equipment itself, but also significantly increases the complexity of electrical wiring, control programming, and maintenance of the entire system, thus increasing maintenance costs and difficulties. Therefore, this invention proposes an automated welding device for automotive sheet metal processing.

[0005] The objective of this invention can be achieved through the following technical solutions: An automated welding device for automotive sheet metal processing includes a base; a pneumatic clamping device is provided at the top of the outer wall of the base; a support column is fixedly connected to the top of the outer wall of the base; a driving device is fixedly connected to the top of the outer wall of the support column via a fixing block; a threaded shaft is provided at the output end of the driving device; a threaded plate is threadedly connected to the outer wall of the threaded shaft, and the threaded plate is slidably connected to the support column; a moving device is provided at the bottom of the outer wall of the threaded plate; a welding device is fixedly connected to the bottom of the outer wall of the moving device; a threaded rod is rotatably connected to the top of the outer wall of the base via a square plate; a linkage mechanism is provided on the outer wall of the threaded rod, and the threaded rod rotates through the linkage mechanism; a moving plate is threadedly connected to the outer wall of the threaded rod; a rotating rod is provided on one side of the outer wall of the moving plate via a square block; a cleaning layer is provided on the bottom of the outer wall of the rotating rod via a circular plate.

[0006] In a preferred embodiment of the present invention, the linkage mechanism includes a chain; a bevel gear is fixedly connected to the outer wall of the threaded shaft; a linkage rod is rotatably connected to one side of the outer wall of the support column via a block; a bevel gear is fixedly connected to one end of the outer wall of the linkage rod, and the bevel gear and the bevel gear mesh with each other; a sprocket is fixedly connected to the outer wall of both the linkage rod and the threaded rod, and a pair of sprockets are connected to each other via a chain; a limiting plate is fixedly connected to one side of the outer wall of the square plate, and the limiting plate is slidably connected to the moving plate.

[0007] In a preferred embodiment of the present invention, a circular rod is rotatably connected to one end of the outer wall of the square plate, and one end of the outer wall of the circular rod penetrates through the square plate and the movable plate; gear three is fixedly connected to the outer walls of both the circular rod and the threaded rod, and a pair of gear three mesh with each other; a square groove is formed on the outer wall of the circular rod; a helical gear four is rotatably connected to one side of the outer wall of the movable plate through an annular block, and the inner side wall of the helical gear four is slidably connected to the inner side wall of the square groove; an auxiliary rod is rotatably connected to one side of the outer wall of the movable plate through a block three; a helical gear five is fixedly connected to the bottom end of the outer wall of the auxiliary rod, and helical gear five meshes with helical gear four; a sprocket two is fixedly connected to the outer walls of both the auxiliary rod and the rotating rod, and a pair of sprocket two are connected by a chain two; the rotating rod is rotatably connected to the square block, and the rotating rod is fixedly connected to the circular plate, and the cleaning layer is disposed on the bottom end of the outer wall of the circular plate.

[0008] In a preferred embodiment of the present invention, a bevel gear six is ​​rotatably connected to one side of the outer wall of the movable plate, and the inner sidewall of the bevel gear six is ​​slidably connected to the inner sidewall of the square groove; a reciprocating rod is rotatably connected to one side of the outer wall of the movable plate via a block five; a bevel gear seven is fixedly connected to one end of the outer wall of the reciprocating rod, and the bevel gear seven meshes with the bevel gear six; a reciprocating plate is provided on the outer sidewall of the reciprocating rod, and the reciprocating plate is slidably connected to the movable plate; a vertical rod is provided on the inner sidewall of the reciprocating plate; a square seat is fixedly connected to the bottom end of the outer wall of the vertical rod; an auxiliary layer is provided on the bottom end of the outer wall of the square seat.

[0009] In a preferred embodiment of the present invention, the vertical rod is rotatably connected to the reciprocating plate; a gear eight is fixedly connected to the outer wall of the vertical rod; a rack eight is fixedly connected to one side of the outer wall of the moving plate through a connecting block; the gear eight and the rack eight mesh with each other.

[0010] In a preferred embodiment of the present invention, the base is provided with a surface cleanliness feature extraction system for performing the following steps: S601: The initial spectral reflectance of the workpiece to be welded area is obtained by a multispectral sensor set on the side of the support column, and the micro-morphological feature vector of the surface to be welded is extracted by a laser displacement sensor. S602: The initial spectral reflectance and the micro-morphological feature vector are fused and used as input variables into the pre-constructed lightweight gradient boosting decision tree model. The model outputs a cleaning difficulty coefficient value that represents the degree of contamination of the surface to be treated.

[0011] In a preferred embodiment of the present invention, the driving device performs the following adjustment based on the cleaning difficulty coefficient value: S603: The cleaning difficulty coefficient value is mapped to the pulse control frequency of the drive device in real time. By dynamically adjusting the output speed of the threaded shaft, the unit area friction work done by the cleaning layer during the displacement of the moving plate is intervened in real time to achieve consistent pretreatment of different contaminated workpieces.

[0012] In a preferred embodiment of the present invention, a molten pool morphology recognition system is provided on the base for performing the following steps: S701: Real-time capture of dynamic heat distribution image sequence of the molten pool at the working point of the welding equipment using an infrared thermal imaging sensor; S702: A deep residual network is used to perform semantic segmentation on the image sequence, extract the molten pool boundary contour in real time, and calculate the geometric deviation value of the instantaneous geometric centroid of the molten pool relative to the theoretical weld centerline.

[0013] In a preferred embodiment of the present invention, the mobile device performs the following real-time gain control based on the geometric deviation value: S703: The geometric deviation value is fed into the long short-term memory network as a time-series input variable to predict the penetration depth shift trend in the next sampling period. The predicted value is converted into a compensation signal and fed back to the mobile device. By fine-tuning the walking speed of the welding equipment, the uneven heating caused by the residual micro-bumps after the cleaning in the aforementioned steps is offset.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The drive device moves the threaded shaft, threaded plate, mobile device, and welding equipment to a suitable height. The welding equipment then moves to weld the workpiece via the mobile device. Simultaneously, the threaded shaft rotates the threaded rod via a linkage mechanism. The rotation of the threaded rod moves the mobile plate, which in turn moves the rotating rod, circular plate, and cleaning layer via a square block. This cleaning layer cleans the workpiece's welding position, resulting in improved welding performance and significantly enhanced welding quality. Furthermore, because the cleaning and welding mechanisms work collaboratively via the linkage mechanism, separate power sources and drive control systems are unnecessary, effectively reducing hardware costs and space requirements. The electrical wiring, control programming, and maintenance of the entire system are also simplified, reducing maintenance costs and difficulty, and improving production efficiency.

[0015] 2. The circular rod rotates through the square slot, driving the bevel gear six to rotate. The moving plate also moves the bevel gear six, ensuring it is always engaged with the bevel gear seven. This causes the bevel gear six to drive the bevel gear seven to rotate, which in turn drives the reciprocating rod to rotate. Because the reciprocating rod and the reciprocating plate are in a reciprocating connection, the reciprocating rod drives the reciprocating plate to move back and forth. This causes the reciprocating plate to drive the vertical rod, the square seat, and the auxiliary layer to move back and forth, wiping and cleaning the workpiece surface prepared for welding after the cleaning layer has been cleaned. This further ensures the absolute cleanliness of the welding area and avoids welding defects caused by residual impurities. This application first uses a rotating cleaning layer for preliminary cleaning, and then uses a reciprocating auxiliary layer for fine wiping, further improving the welding quality. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a structural diagram of the main body of the present invention; Figure 2 This is a structural diagram of the drive device, threaded plate, welding device, and linkage rod of the present invention. Figure 3 This is a structural diagram of the threaded rod, moving plate, cleaning layer, and auxiliary layer of the present invention; Figure 4 This is a structural diagram of the circular rod, auxiliary rod, and cleaning layer of the present invention; Figure 5 This is a structural diagram of the reciprocating rod, reciprocating plate, and auxiliary layer of the present invention; In the diagram: 1. Base; 2. Pneumatic clamping device; 3. Support column; 4. Drive device; 5. Threaded shaft; 6. Threaded plate; 7. Mobile device; 8. Welding equipment; 9. Square plate; 10. Threaded rod; 11. Linkage mechanism; 12. Moving plate; 13. Limiting plate; 14. Square block; 15. Rotating rod; 16. Circular plate; 17. Cleaning layer; 111. Chain 1; 112. Bevel gear 1; 113. Linkage mechanism. 114. Moving rod; 115. Bevel gear 2; 116. Sprocket 1; 18. Round rod; 19. Gear 3; 20. Square slot; 21. Helical gear 4; 22. Auxiliary rod; 23. Helical gear 5; 24. Sprocket 2; 25. Chain 2; 26. Bevel gear 6; 27. Reciprocating rod; 28. Bevel gear 7; 29. ​​Reciprocating plate; 30. Vertical rod; 31. Square seat; 32. Auxiliary layer; 33. Gear 8; 34. Rack 8. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figures 1-5As shown, an automated welding device for automotive sheet metal processing includes a base 1; a pneumatic clamping device 2 is provided at the top of the outer wall of the base 1; a support column 3 is fixedly connected to the top of the outer wall of the base 1; a driving device 4 is fixedly connected to the top of the outer wall of the support column 3 via a fixing block; a threaded shaft 5 is provided at the output end of the driving device 4; a threaded plate 6 is threadedly connected to the outer wall of the threaded shaft 5, and the threaded plate 6 is slidably connected to the support column 3; a moving device 7 is provided at the bottom of the outer wall of the threaded plate 6; a welding device 8 is fixedly connected to the bottom of the outer wall of the moving device 7; a threaded rod 10 is rotatably connected to the top of the outer wall of the base 1 via a square plate 9; the outer wall of the threaded rod 10 is provided with... A linkage mechanism 11 is provided, and the threaded rod 10 rotates through the linkage mechanism 11; a movable plate 12 is threadedly connected to the outer wall of the threaded rod 10; a limiting plate 13 is fixedly connected to one side of the outer wall of the square plate 9, and the limiting plate 13 is slidably connected to the movable plate 12; a rotating rod 15 is provided on one side of the outer wall of the movable plate 12 through a square block 14; a cleaning layer 17 is provided at the bottom of the outer wall of the rotating rod 15 through a circular plate 16. By placing the workpiece on the base 1 and fixing it with a pneumatic clamping device 2, the threaded shaft 5 is driven by the drive device 4, causing the threaded shaft 5 to rotate and drive the threaded plate 6 to move downward, thereby driving the movable device 7 and the welding device. 8. Move to a suitable height. At this time, welding equipment 8 welds the workpiece. The moving equipment 7 drives the welding equipment 8 to move, so that the welding equipment 8 can weld while moving, and the workpiece welding is completed. When the threaded shaft 5 rotates and drives the threaded plate 6 to descend, the threaded shaft 5 rotates at the same time, and drives the threaded rod 10 to rotate through the linkage mechanism 11. The rotation of the threaded rod 10 drives the moving plate 12 to move. The moving plate 12 is limited by the limiting plate 13 and can only move but not rotate. When the moving plate 12 moves, it drives the rotating rod 15, the circular plate 16 and the cleaning layer 17 to move through the square block 14, so that the cleaning layer 17 cleans the position of the workpiece to be welded. When the welding equipment 8 moves to the appropriate position to prepare for welding, the moving plate 12 and the cleaning layer 17 move to the edge position, which does not hinder the welding equipment 8 from welding. Moreover, because the workpiece is cleaned at the welding position, the welding effect of the welding equipment 8 is better and the welding quality is significantly improved. At the same time, since the cleaning mechanism and the welding mechanism work together through the linkage mechanism 11, there is no need to equip them with independent power sources and drive control systems, thereby effectively reducing hardware costs and space occupation. In addition, the electrical wiring, control programming and maintenance of the entire system are also simplified, reducing maintenance costs and maintenance difficulty, and improving production efficiency.

[0020] The linkage mechanism 11 includes a chain 111; a bevel gear 112 is fixedly connected to the outer wall of the threaded shaft 5; a linkage rod 113 is rotatably connected to one side of the outer wall of the support column 3 via a block; a bevel gear 114 is fixedly connected to one end of the outer wall of the linkage rod 113, and the bevel gear 114 meshes with the bevel gear 112; sprockets 115 are fixedly connected to the outer walls of both the linkage rod 113 and the threaded rod 10, and a pair of sprockets 115 are connected by the chain 111. The rotation of the threaded shaft 5 drives the bevel gear 112 to rotate. Wheel 112 drives bevel gear 214 to rotate, bevel gear 214 drives linkage rod 113 to rotate, linkage rod 113 drives threaded rod 10 to rotate through sprocket 115 and chain 111, thereby driving moving plate 12 and cleaning layer 17 to move and clean the workpiece. By utilizing the principle of mechanical transmission, the cleaning mechanism and welding mechanism can be operated synchronously. Without the need for an additional power source or complex control system, the cleaning and welding can be seamlessly connected, improving the automation level and production efficiency of automotive sheet metal processing.

[0021] A bevel gear 26 is rotatably connected to one side of the outer wall of the movable plate 12, and the inner wall of the bevel gear 26 is slidably connected to the inner wall of the square groove 20; a reciprocating rod 27 is rotatably connected to one side of the outer wall of the movable plate 12 via a square block 5; a bevel gear 28 is fixedly connected to one end of the outer wall of the reciprocating rod 27, and the bevel gear 28 meshes with the bevel gear 26; a reciprocating plate 29 is provided on the outer wall of the reciprocating rod 27, and the reciprocating plate 29 is slidably connected to the movable plate 12; a vertical rod 30 is provided on the inner wall of the reciprocating plate 29; a square seat 31 is fixedly connected to the bottom end of the outer wall of the vertical rod 30; an auxiliary layer 32 is provided on the bottom end of the outer wall of the square seat 31; the circular rod 18 rotates through the square groove 20 to drive the bevel gear 26 to rotate, and the movable plate 12 drives the bevel gear 26 to rotate. The movement of wheel 6 26 keeps bevel gear 6 26 meshing with bevel gear 7 28, causing bevel gear 6 26 to drive bevel gear 7 28 to rotate, which in turn drives reciprocating rod 27 to rotate. Since reciprocating rod 27 and reciprocating plate 29 are in a reciprocating connection, reciprocating rod 27 drives reciprocating plate 29 to move back and forth, causing reciprocating plate 29 to drive vertical rod 30, square seat 31 and auxiliary layer 32 to move back and forth, wiping and cleaning the workpiece surface to be welded after cleaning by cleaning layer 17, further ensuring the absolute cleanliness of the welding area and avoiding welding defects caused by residual impurities. In this application, the rotating cleaning layer 17 is used for preliminary cleaning, and then the reciprocating auxiliary layer 32 is used for fine wiping, so that the welding quality is further improved.

[0022] A circular rod 18 is rotatably connected to one end of the outer wall of the square plate 9, and one end of the outer wall of the circular rod 18 passes through the square plate 9 and the movable plate 12; gears 19 are fixedly connected to the outer walls of both the circular rod 18 and the threaded rod 10, and a pair of gears 19 mesh with each other; a square groove 20 is formed on the outer wall of the circular rod 18; a helical gear 21 is rotatably connected to one side of the outer wall of the movable plate 12 via an annular block, and the inner side wall of the helical gear 21 is slidably connected to the inner side wall of the square groove 20; a square groove 20 is formed on one side of the outer wall of the movable plate 12 via a square rod 18. An auxiliary rod 22 is movably connected; a helical gear 23 is fixedly connected to the bottom of the outer wall of the auxiliary rod 22, and the helical gear 23 meshes with the helical gear 21; sprockets 24 are fixedly connected to the outer walls of both the auxiliary rod 22 and the rotating rod 15, and a pair of sprockets 24 are connected by a chain 25; the rotating rod 15 is rotatably connected to the square block 14, and the rotating rod 15 is fixedly connected to the circular plate 16; the cleaning layer 17 is set on the bottom of the outer wall of the circular plate 16; the rotation of the threaded rod 10 drives the circular rod 18 to rotate through a pair of gears 3 19. Because the inner wall of helical gear 4 21 is slidably connected to the inner wall of the square groove 20 on the circular rod 18, the circular rod 18 drives helical gear 4 21 to rotate. Since helical gear 4 21 is rotatably connected to the moving plate 12 via an annular block, the moving plate 12 drives helical gear 4 21 to move, and the circular rod 18 drives it to rotate, causing it to move and rotate simultaneously, always meshing with helical gear 5 23. This causes helical gear 4 21 to drive helical gear 5 23 to rotate, thereby driving auxiliary rod 22 to rotate. Auxiliary rod 22 is connected to sprocket 2 24 and chain... The second component 25 drives the rotating rod 15 to rotate, which in turn drives the circular plate 16 and the cleaning layer 17 to rotate. This causes the cleaning layer 17 to move and rotate simultaneously, thereby improving its cleaning effect. This not only enhances the cleaning ability of the cleaning layer 17 on the workpiece surface, ensuring the cleanliness of the workpiece surface before welding, but also provides high flexibility and adjustability. By adjusting the relative positions and transmission ratios of each component, it can adapt to the welding needs of automotive sheet metal parts of different sizes and shapes, thus improving the versatility and applicability of the equipment.

[0023] The vertical rod 30 is rotatably connected to the reciprocating plate 29; a gear 8 33 is fixedly connected to the outer wall of the vertical rod 30; a rack 8 34 is fixedly connected to one side of the outer wall of the moving plate 12 through a connecting block; the gear 8 33 and the rack 8 34 mesh with each other, and when the reciprocating plate 29 moves back and forth, it drives the vertical rod 30 and the gear 8 33 to move back and forth. Because the gear 8 33 and the rack 8 34 mesh, and the rack 8 34 is fixed to the moving plate 12 through the connecting block, when the gear 8 33 moves, it rotates through the rack 8 34, thereby driving the vertical rod 30 to rotate, which in turn drives the square seat 31 and the auxiliary layer 32 to rotate. The auxiliary layer 32 moves back and forth while rotating, which performs a more comprehensive and detailed cleaning treatment on the surface of the workpiece. It not only makes full use of the wide range of the cleaning layer 17's rotational cleaning, but also uses the deep cleaning ability of the auxiliary layer 32's reciprocating rotational wiping to remove impurities from the workpiece's welding position more thoroughly.

[0024] In use, the workpiece is placed on the base 1 and fixed by the pneumatic clamping device 2. At this time, the drive device 4 drives the threaded shaft 5, and the rotation of the threaded shaft 5 causes the threaded plate 6 to move down. The threaded plate 6 then moves the moving device 7 and the welding device 8 to a suitable height. At this time, the welding device 8 welds the workpiece, and the moving device 7 drives the welding device 8 to move, so that the welding device 8 can perform welding while moving, and the workpiece welding is completed.

[0025] As the threaded shaft 5 rotates, causing the threaded plate 6 to descend, the rotation of the threaded shaft 5 also drives the first bevel gear 112 to rotate. The first bevel gear 112 drives the second bevel gear 114 to rotate, which in turn drives the linkage rod 113 to rotate. The linkage rod 113, through the first sprocket 115 and the first chain 111, drives the threaded rod 10 to rotate. The rotation of the threaded rod 10 causes the moving plate 12 to move. The moving plate 12 is limited by the limiting plate 13, allowing it to move but not rotate. When the moving plate 12 moves, it drives the rotating rod 15, the circular plate 16, and the cleaning layer 17 to move through the square block 14, so that the cleaning layer 17 cleans the workpiece at the position to be welded. When welding... When the equipment 8 is moved to the appropriate position to prepare for welding, the moving plate 12 and the cleaning layer 17 move to the edge position, which does not hinder the welding equipment 8 from welding. Moreover, because the workpiece is cleaned at the welding position, the welding effect of the welding equipment 8 is better and the welding quality is significantly improved. At the same time, since the cleaning mechanism and the welding mechanism work together through the linkage mechanism 11, there is no need to equip them with independent power sources and drive control systems, thereby effectively reducing hardware costs and space occupation. In addition, the electrical wiring, control programming and maintenance of the entire system are also simplified, reducing maintenance costs and maintenance difficulty, and improving production efficiency.

[0026] The rotation of the threaded rod 10 drives the circular rod 18 to rotate via a pair of gears 19. Because the inner wall of the helical gear 21 is slidably connected to the inner wall of the square groove 20 on the circular rod 18, the circular rod 18 drives the helical gear 21 to rotate. Furthermore, because the helical gear 21 is rotatably connected to the moving plate 12 via an annular block, the moving plate 12 drives the helical gear 21 to move, causing the circular rod 18 to rotate, thus rotating it while moving, always meshing with the helical gear 23. This causes the helical gear 21 to drive the helical gear 23 to rotate, thereby driving the auxiliary rod 22 to rotate. The auxiliary rod 22 drives the rotating rod 15 to rotate via the second sprocket 24 and the second chain 25. The rotating rod 15 drives the circular plate 16 and the cleaning layer 17 to rotate, so that the cleaning layer 17 moves and rotates at the same time, thereby improving the cleaning effect of the cleaning layer 17. This not only enhances the cleaning ability of the cleaning layer 17 on the surface of the workpiece and ensures the cleanliness of the workpiece surface before welding, but also has high flexibility and adjustability. By adjusting the relative position and transmission ratio of each component, it can adapt to the welding needs of automotive sheet metal parts of different sizes and shapes, thus improving the versatility and applicability of the equipment.

[0027] The circular rod 18 rotates through the square slot 20, driving the bevel gear 26 to rotate. The moving plate 12 also drives the bevel gear 26 to move, ensuring that the bevel gear 26 is always engaged with the bevel gear 28. This causes the bevel gear 26 to drive the bevel gear 28 to rotate, which in turn drives the reciprocating rod 27 to rotate. Since the reciprocating rod 27 and the reciprocating plate 29 are in a reciprocating connection, the reciprocating rod 27 drives the reciprocating plate 29 to move back and forth. This causes the reciprocating plate 29 to drive the vertical rod 30, the square seat 31, and the auxiliary layer 32 to move back and forth. This process wipes and cleans the workpiece surface prepared for welding after cleaning by the cleaning layer 17, further ensuring the absolute cleanliness of the welding area and avoiding welding defects caused by residual impurities. In this application, the rotating cleaning layer 17 performs preliminary cleaning, and then the reciprocating auxiliary layer 32 performs fine wiping, further improving the welding quality.

[0028] When the reciprocating plate 29 moves back and forth, it drives the vertical rod 30 and the gear 8 33 to move back and forth. Since the gear 8 33 meshes with the rack 8 34 and the rack 8 34 is fixed on the moving plate 12 through the connecting block, the gear 8 33 moves and rotates through the rack 8 34, thereby driving the vertical rod 30 to rotate. The vertical rod 30 drives the square seat 31 and the auxiliary layer 32 to rotate. The auxiliary layer 32 moves back and forth and rotates at the same time, so as to carry out a more comprehensive and detailed cleaning treatment on the surface of the workpiece. It not only makes full use of the wide range of cleaning layer 17, but also uses the deep cleaning ability of the auxiliary layer 32 to wipe back and forth, so that the impurities at the position of the workpiece to be welded are removed more thoroughly.

[0029] Base 1 is equipped with a surface cleanliness feature extraction system for performing the following steps: S601: The initial spectral reflectance of the workpiece to be welded area is obtained by a multispectral sensor set on the side of the support column 3, and the micro-morphological feature vector of the surface to be welded is extracted by a laser displacement sensor. The specific implementation process of step S601 is as follows: The multispectral sensor and the laser displacement sensor are fixed to the side wall of the support column 3 by the mounting bracket, so that their measuring optical axis points to the workpiece to be welded area fixed on the base 1 by the pneumatic clamping device 2. Under the illumination of the matching standard light source, the multispectral sensor collects the reflected light energy of the workpiece surface in multiple discrete wavelength bands. The initial spectral reflectance is obtained by calculating the ratio of reflected energy to incident energy. This initial spectral reflectance is used to characterize the degree of oil stain coverage, oxide scale thickness and inherent spectral response characteristics of the metal material on the sheet metal surface. The specific method for extracting the micro-morphological feature vector of the surface to be welded using the laser displacement sensor is as follows: The laser displacement sensor emits a laser beam to reciprocate and scan the area to be welded, and obtains the three-dimensional spatial coordinate information of each pixel in the sampling area. The coordinate data was then detrended to eliminate the influence of the workpiece's macroscopic slope, resulting in height field data reflecting the microscopic undulations of the surface. Extracting surface arithmetic mean height from height field data Surface root mean square height and surface kurtosis Multiple roughness parameters, including those mentioned above, are combined into a multidimensional array in a preset order to form a micro-morphological feature vector. This vector serves as a geometric feature descriptor describing the physical texture and unevenness of the workpiece surface. Together with the initial spectral reflectance, it constitutes a multidimensional input source for subsequent evaluation of the workpiece cleaning difficulty.

[0030] S602: The initial spectral reflectance and micro-morphological feature vector are fused and used as input variables into a pre-constructed lightweight gradient boosting decision tree model. The model outputs a cleaning difficulty coefficient value that represents the degree of contamination on the surface to be treated.

[0031] The specific implementation of step S602 is as follows: The initial spectral reflectance sequence composed of multispectral data obtained in step S601 and the micro-morphological feature vector composed of roughness parameters are normalized. The data of different dimensions are mapped to the [0,1] interval by the deviation normalization method to eliminate the difference in numerical magnitude between features. The detailed process is as follows: For each band value in the original spectral reflectance sequence extracted in step S601 and each dimension of roughness parameter in the micro-morphological feature vector... , The specific source and acquisition process are as follows: The multispectral sensor integrates... Narrowband filters and photoelectric detection arrays of specific wavelengths are used to synchronously sample the area of ​​the workpiece to be welded before welding, obtaining the reflected energy intensity corresponding to each band. This intensity is then compared with a preset reference reflected energy to obtain discrete reflectivity values ​​corresponding to different center wavelengths. The reflectivity value of each band is defined as a component, i.e., a parameter, in the original spectral reflectivity sequence. The spectral source is determined by the laser. Secondly, the laser displacement sensor utilizes the principle of laser triangulation, emitting a linear laser and receiving the reflected light spot from the workpiece surface through an internal photosensitive element, thereby calculating the spatial height coordinates of each sampling point on the surface of the area to be welded. A raw height field dataset reflecting the micro-undulations of the surface was constructed. Based on this height field dataset and according to the surface roughness characterization standard, multiple evaluation indicators reflecting the surface morphology characteristics were calculated using integral operations and statistical methods. Specifically, the arithmetic mean height was obtained by integrating the absolute height with the area. The root mean square height is obtained by performing a root mean square operation on the square of the height. And the skewness reflecting the symmetry of the height distribution. and kurtosis, which reflects the sharpness of the surface. Each of the statistical index values ​​obtained after algorithmic processing constitutes a one-dimensional component of the micromorphological feature vector, i.e., a parameter. The morphological source components and the spectral source components are arranged together according to a preset topological order to form a high-dimensional original feature set. Any independent element in this set is uniformly abstracted into a variable to be normalized when entering the mathematical processing logic of step S602. It retrieves the historical extreme value, i.e. the maximum value, corresponding to the specific feature from the preset feature reference database. and minimum value Using the linear transformation formula Numerical calculations were performed to obtain the normalized eigencomponents. Among them, the molecular end Used to calculate the increment of the current observation relative to the baseline minimum, denominator side The dynamic evolution of this feature spans the entire range, and the ratio of the two ensures that both large-scale spectral energy values ​​and extremely small micrometer-scale morphology parameters are proportionally scaled to the dimensionless unified numerical space [0,1]. The scalar value is calculated by the output layer of the lightweight gradient boosting decision tree model and defined as the cleaning difficulty coefficient value. The detailed process is as follows: input the concatenated fused feature input vector into the system... In the LightGBM cascade structure composed of regression decision trees, the input vector searches downwards from the root node within each tree according to the feature segmentation rules, eventually falling into the corresponding leaf node. Each leaf node of the decision tree stores the weighted weight values ​​calculated based on negative gradient fitting. The model performs a linear summation operation on the weight values ​​output by the decision tree to calculate the initial comprehensive score. In order to give the rating a quantifiable physical meaning of cleaning difficulty, the initial comprehensive rating will be... Substitute into the Sigmoid activation function A normalization mapping is performed, and the scalar value located in the closed interval [0,1] is calculated as the cleaning difficulty coefficient. This value directly quantifies the mapping relationship between the adhesion of contaminants on the workpiece surface and the mechanical cleaning resistance. Its output is sent to the control system on the base 1, which adjusts the output speed of the drive device 4 to dynamically change the action parameters of the threaded shaft 5, thereby controlling the cleaning layer 17 to adjust its speed according to the workpiece surface during cleaning operations. The size is automatically matched to the optimal cleaning kinetic energy, and the normalized spectral features and morphological features are vector-stitched in a preset order to construct a dimension of [dimensional value missing]. The fused feature input vector, where The sum of the number of spectral channels and the total number of morphological parameters is used to input the fused feature vector into a lightweight gradient boosting decision tree model pre-trained and deployed in the control system on base 1. This model employs a histogram-based decision tree algorithm and a leaf-growth strategy. Utilizing weight parameters trained offline on sample sets labeled with different oil types, oxidation levels, and metal substrates, it performs multi-level decision reasoning on the input fused features, identifying the combined influence of the adhesion energy of contaminants on the workpiece surface and the micro-texture. A scalar value, defined as the cleaning difficulty coefficient, is calculated from the output layer of the lightweight gradient boosting decision tree model. Its value range is set to real numbers between 0 and 1, where The closer the value is to 1, the stronger the adhesion between the contaminants on the workpiece surface and the substrate, and the deeper the micro-pits are deposited. This means that the cleaning layer 17 needs to perform a stronger cleaning effort. Conversely, the value is lower, indicating that the surface is easier to clean. This provides a precise quantitative input basis for the operation in step S603 of dynamically changing the descent rate of the threaded shaft 5 by adjusting the output speed of the drive device 4.

[0032] Drive device 4 performs the following adjustments based on the cleaning difficulty coefficient value: S603: The cleaning difficulty coefficient value is mapped to the pulse control frequency of the drive device 4 in real time. By dynamically adjusting the output speed of the threaded shaft 5, the unit area friction work done by the cleaning layer 17 during the displacement process with the moving plate 12 is intervened in real time to achieve consistent pretreatment of different contaminated workpieces.

[0033] The implementation method of step S603 is as follows: the control system calculates the cleaning difficulty coefficient value output in step S602. Through a preset linear or nonlinear mapping function Calculate the target pulse frequency, where The gain function ensures that the higher the cleaning difficulty, the higher the corresponding output pulse frequency. The larger the pulse frequency, the more likely the drive device 4 is a stepper motor or servo motor. Its driver receives the pulse frequency signal and controls the rotational speed of the motor output shaft, making the rotational angular velocity of the threaded shaft 5 proportional to the pulse frequency. The specific method of using the threaded shaft 5 to dynamically intervene in the frictional work per unit area is as follows: by adjusting the rotational speed of the threaded shaft 5, the power distribution between the threaded rod 10 and the rotating rod 15 is changed synchronously by the linkage mechanism 11, so that the horizontal displacement speed of the moving plate 12 and the rotational tangential speed of the cleaning layer 17 change in coordination. The so-called frictional work per unit area refers to the mechanical energy generated by friction in the unit scanning area of ​​the cleaning layer 17 on the workpiece surface. The sum, the size of which depends on the product of the normal pressure, friction coefficient and relative rubbing distance of the cleaning layer 17, increases the rotational sweeping frequency of the cleaning layer 17 without changing the total stroke of the moving plate 12 by increasing the rotational speed of the threaded shaft 5. This increases the frictional distance when the cleaning difficulty coefficient is high, compensating for the additional cleaning kinetic energy required due to the high adhesion of contaminants. Through real-time intervention of the power source output parameters, the surface cleanliness of sheet metal workpieces with different levels of contamination after the pretreatment stage reaches the preset welding standard, ensuring the stability of the molten pool and the consistency of the weld quality during subsequent welding by the welding equipment 8.

[0034] A molten pool morphology recognition system is installed on base 1 to perform the following steps: S701: Real-time capture of dynamic heat distribution image sequences of the molten pool at 8 working points of the welding equipment using an infrared thermal imaging sensor; The specific implementation of step S701 is as follows: An infrared thermal imaging sensor with long-wave infrared sensing function is fixed to the side of the mobile device 7 using a high-temperature resistant mounting bracket. The focal length of its optical lens and the detection optical axis are adjusted so that its field of view center is locked at the intersection of the electrode axis of the welding device 8 and the workpiece surface, which is defined as the working point. The working point is the physical area where the electric arc or high-energy beam exchanges heat with the metal substrate during welding, inducing material melting. The liquid metal pool formed within this area is defined as the molten pool. The specific method for real-time capture is as follows: Simultaneously with the start of the driving device 4 and the entry of the welding device 8 into the arc-starting welding state, the infrared thermal imaging sensor is synchronously triggered to capture a preset image. The sampling frequency, such as 50Hz to 120Hz, is used to continuously scan the working point with thermal radiation. The sensed radiation energy signal is converted into a two-dimensional pixel matrix representing the instantaneous temperature field distribution. The two-dimensional pixel matrix and its corresponding timestamp together constitute a dynamic thermal distribution image sequence. The gray value or pseudo-color value of each pixel in the image is directly mapped to the real temperature value corresponding to the spatial coordinate point after blackbody radiation correction. The geometric deformation, thermal field propagation and heat-affected zone evolution characteristics of the molten pool during the welding process are recorded in real time. This provides an image data source with high temporal resolution and temperature sensitivity for the subsequent step S702 to extract molten pool features using a deep residual network.

[0035] S702: A deep residual network is used to perform semantic segmentation on the image sequence, extract the molten pool boundary contour in real time, and calculate the geometric deviation value of the instantaneous geometric centroid of the molten pool relative to the theoretical weld centerline.

[0036] The specific implementation of step S702 is as follows: The infrared thermal distribution image obtained in step S701 is input into a deep residual network deployed in the local controller. The deep residual network uses stacked residual modules and skip connection structures to extract features from the gradient changes of different temperature zones in the image using convolutional kernels, effectively solving the gradient vanishing problem in deep network training. Semantic segmentation refers to using the deep residual network to perform pixel-level binary classification on the pixels in the thermal distribution image, labeling the pixels in the molten pool area above the liquidus temperature as target pixels, and labeling the remaining substrate areas as background pixels, outputting the corresponding binary mask image. The specific method for real-time extraction of the molten pool boundary contour is as follows: the closed contour tracking of the binary mask image is performed using edge operators or connected component extraction algorithms to obtain a series of discrete spatial coordinate points. The specific process for constructing the melt pool boundary curve is as follows: Connectivity analysis is performed on the binarized mask image output by the deep residual network to identify target regions with a pixel value of 1 and calculate the area of ​​each region. The connected region with the largest area is selected as the target melt pool region. Median filtering is used to remove isolated noise pixels around this region. A Moore's neighborhood tracking algorithm is employed, starting from the top left corner of the mask image and scanning line by line. When the first abrupt change from background pixel 0 to target pixel 1 is detected, that pixel is set as the tracking starting point. ,by Centered on its Within the neighborhood, probe the next target pixel in a clockwise direction, iterating through the search path until the search path returns to the starting point. This forms a closed loop surrounding the molten pool. The physical coordinates of the boundary pixels visited sequentially during the tracking process are extracted and arranged in order, generating a series of discrete spatial coordinate points. The resulting sequence is the molten pool boundary curve, where the subscripts are... The topological index representing the boundary pixel is used. This curve accurately depicts the geometrically broadened edge of the liquid metal pool under the instantaneous temperature field. The specific method for calculating the instantaneous geometric centroid is as follows: count the coordinates of all target pixels within the area enclosed by the molten pool boundary curve, and use the arithmetic mean method to calculate the coordinates of the geometric center point of the molten pool region in the current frame image. The theoretical weld centerline refers to the reference straight line equation established in the image coordinate system based on the processing trajectory preset by the welding equipment 8. ; by calculating the geometric center point To the equation of the line The geometric deviation value is obtained by measuring the vertical Euclidean distance or normal vector displacement. The specific process is as follows: Obtain the theoretical centerline data of the welding device 8 in the physical space coordinate system of the base 1; using the pre-calibrated intrinsic and extrinsic projection matrices of the infrared thermal imaging sensor, project the three-dimensional trajectory points in the physical space onto the two-dimensional image coordinate system plane corresponding to the sensor's photosensitive array; and use the least squares method to linearly fit the projected discrete trajectory points to construct the reference straight line equation. ,in These are the x and y coordinates of the image, respectively. To reflect the preset orientation and tilt of the weld bead in the image field of view, the geometric center point calculated in step S702 is used as a constant. Substitute the coordinate values ​​into the formula for calculating the perpendicular distance from a point to a line, that is, by calculating... Obtain the geometric center point to the reference line The vertical Euclidean distance, defined as the geometric deviation value. Its absolute value quantitatively characterizes the physical displacement of the current molten pool energy center from the theoretical welding centerline. This geometric deviation value will be input as an instruction to the control system of the mobile device 7 to compensate for the travel trajectory of the welding equipment 8 or to perform subsequent dynamic gain adjustment on the processing speed. It is used to quantitatively describe the degree to which the melting center deviates from the preset trajectory due to the workpiece's thermal deformation or residual micro-unevenness on the surface during the welding process, and serves as the core input feature quantity for the long short-term memory network to predict the evolution trend of the weld depth in step S703.

[0037] Mobile device 7 performs the following real-time gain control based on the geometric deviation value: S703: The geometric deviation value is fed into the long short-term memory network as a time-series input variable to predict the penetration depth shift trend in the next sampling period. The predicted value is converted into a compensation signal and fed back to the mobile device 7. The uneven heating caused by the residual micro-bumps after the cleaning in the previous steps is offset by fine-tuning the walking speed of the welding equipment 8.

[0038] The specific implementation of step S703 is as follows: the control system extracts the previous data from the cache module. Geometric deviation value sequence of consecutive sampling periods As a time-series feature input, the data is scaled to a scaled value using a linear normalization function. Within the interval, to eliminate dimensional fluctuations in sensor signals caused by electromagnetic interference, the sequence is... The data is fed into the hidden layers of the Long Short-Term Memory (LSTM) network sequentially over time steps. Within each time step, a forgetting gate is used to calculate forgetting weights by combining the current input with the hidden state from the previous time step. This weights are used to remove non-characteristic transient noise caused by mechanical vibration or arc flicker. The input gate selectively retains features reflecting the slow evolution of the melt pool boundary due to heat accumulation or changes in surface roughness, updating the cell state at the current sampling time. The output gate calculates the hidden state vector based on the updated cell state. The vector This incorporates the deep temporal evolution of the melt pool morphology from its initial steady state to its current deviation state, and utilizes a pre-defined fully connected mapping layer on the hidden state vector. The mapping logic for linear regression calculation is as follows: ,in This is the weight matrix. The bias coefficient is the scalar value obtained from the calculation. This refers to the weld penetration deviation trend, which indicates a dynamic fluctuation in the absorption rate of welding heat on the workpiece surface due to residual micro-roughness differences or uneven local oxides after the pretreatment by the cleaning layer 17 in the aforementioned steps. This fluctuation induces an unexpected shrinkage or expansion of the weld pool volume. The control system maps the weld penetration deviation trend into a pulse frequency correction signal and feeds this signal back to the mobile device 7 in real time as a compensation signal. The mobile device 7 is fixed to the bottom of the threaded plate 6. After receiving the compensation signal, the control terminal of its internal drive motor changes the speed of its output shaft in real time, thus affecting the welding equipment 8 mounted on the mobile device 7. The walking speed is slightly adjusted. The specific walking speed adjustment logic is as follows: when the long short-term memory network predicts that the weld penetration is becoming shallower, the moving speed of the mobile device 7 is reduced to prolong the dwell time of the welding arc at the current position and increase the heat input per unit length. Conversely, the moving speed is increased to reduce heat accumulation. In this way, the processing speed is dynamically adjusted to counteract the instantaneous uneven heating caused by the micro-unevenness of the surface residue. Through this closed-loop timing prediction and speed feedback mechanism, the welding equipment 8 can maintain a constant molten pool penetration state throughout the entire welding stroke on the base 1, ensuring the high consistency of the weld formation and the connection strength of the automotive sheet metal parts.

[0039] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, 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. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An automated welding device for automotive sheet metal processing, comprising a base (1); a pneumatic clamping device (2) is provided at the top of the outer wall of the base (1); a support column (3) is fixedly connected to the top of the outer wall of the base (1); a driving device (4) is fixedly connected to the top of the outer wall of the support column (3) via a fixing block; a threaded shaft (5) is provided at the output end of the driving device (4); a threaded plate (6) is threadedly connected to the outer wall of the threaded shaft (5), and the threaded plate (6) is slidably connected to the support column (3); a moving device (7) is provided at the bottom of the outer wall of the threaded plate (6); a welding device (8) is fixedly connected to the bottom of the outer wall of the moving device (7); characterized in that, The top of the outer wall of the base (1) is rotatably connected to a threaded rod (10) via a square plate (9); the outer wall of the threaded rod (10) is provided with a linkage mechanism (11), and the threaded rod (10) rotates through the linkage mechanism (11); the outer wall of the threaded rod (10) is threadedly connected to a movable plate (12); one side of the outer wall of the movable plate (12) is provided with a rotating rod (15) via a square block (14); the bottom of the outer wall of the rotating rod (15) is provided with a cleaning layer (17) via a circular plate (16).

2. The automated welding device for automotive sheet metal processing according to claim 1, characterized in that, The linkage mechanism (11) includes a chain (111); a bevel gear (112) is fixedly connected to the outer wall of the threaded shaft (5); a linkage rod (113) is rotatably connected to one side of the outer wall of the support column (3) through a block; a bevel gear (114) is fixedly connected to one end of the outer wall of the linkage rod (113), and the bevel gear (114) meshes with the bevel gear (112); a sprocket (115) is fixedly connected to the outer wall of both the linkage rod (113) and the threaded rod (10), and a pair of sprockets (115) are connected to each other through the chain (111); a limiting plate (13) is fixedly connected to one side of the outer wall of the square plate (9), and the limiting plate (13) is slidably connected to the moving plate (12).

3. The automated welding device for automotive sheet metal processing according to claim 2, characterized in that, A circular rod (18) is rotatably connected to one end of the outer wall of the square plate (9), and one end of the outer wall of the circular rod (18) passes through the square plate (9) and the movable plate (12); gears (19) are fixedly connected to the outer walls of both the circular rod (18) and the threaded rod (10), and a pair of gears (19) mesh with each other; a square groove (20) is opened on the outer wall of the circular rod (18); a helical gear (21) is rotatably connected to one side of the outer wall of the movable plate (12) through an annular block, and the inner wall of the helical gear (21) is slidably connected to the inner wall of the square groove (20); the movable plate (12) has an auxiliary rod (22) rotatably connected to one side of the outer wall via a block three; the bottom of the outer wall of the auxiliary rod (22) is fixedly connected to a helical gear five (23), and the helical gear five (23) meshes with the helical gear four (21); the outer walls of the auxiliary rod (22) and the rotating rod (15) are both fixedly connected to a sprocket two (24), and a pair of sprocket two (24) are connected by a chain two (25); the rotating rod (15) is rotatably connected to the block (14), and the rotating rod (15) is fixedly connected to the circular plate (16), and the cleaning layer (17) is set on the bottom of the outer wall of the circular plate (16).

4. The automated welding device for automotive sheet metal processing according to claim 3, characterized in that, A bevel gear six (26) is rotatably connected to one side of the outer wall of the movable plate (12), and the inner side wall of the bevel gear six (26) is slidably connected to the inner side wall of the square groove (20); a reciprocating rod (27) is rotatably connected to one side of the outer wall of the movable plate (12) through a square five; a bevel gear seven (28) is fixedly connected to one end of the outer wall of the reciprocating rod (27), and the bevel gear seven (28) meshes with the bevel gear six (26); a reciprocating plate (29) is provided on the outer side wall of the reciprocating rod (27), and the reciprocating plate (29) is slidably connected to the movable plate (12); a vertical rod (30) is provided on the inner side wall of the reciprocating plate (29); a square seat (31) is fixedly connected to the bottom end of the outer wall of the vertical rod (30); an auxiliary layer (32) is provided on the bottom end of the outer wall of the square seat (31).

5. An automated welding device for automotive sheet metal processing according to claim 4, characterized in that, The vertical rod (30) is rotatably connected to the reciprocating plate (29); a gear eight (33) is fixedly connected to the outer wall of the vertical rod (30); a rack eight (34) is fixedly connected to one side of the outer wall of the moving plate (12) through a connecting block; the gear eight (33) and the rack eight (34) mesh with each other.

6. An automated welding device for automotive sheet metal processing according to claim 5, characterized in that, The base (1) is equipped with a surface cleanliness feature extraction system for performing the following steps: S601: The initial spectral reflectance of the workpiece to be welded area is obtained by a multispectral sensor set on the side of the support column (3), and the micro-morphological feature vector of the surface to be welded is extracted by a laser displacement sensor. S602: The initial spectral reflectance and the micro-morphological feature vector are fused and used as input variables into the pre-constructed lightweight gradient boosting decision tree model. The model outputs a cleaning difficulty coefficient value that represents the degree of contamination of the surface to be treated.

7. An automated welding device for automotive sheet metal processing according to claim 6, characterized in that, The drive device (4) performs the following adjustments based on the cleaning difficulty coefficient value: S603: The cleaning difficulty coefficient value is mapped to the pulse control frequency of the drive device (4) in real time. By dynamically adjusting the output speed of the threaded shaft (5), the unit area friction work of the cleaning layer (17) during the displacement process with the moving plate (12) is intervened in real time to achieve consistent pretreatment of different contaminated workpieces.

8. An automated welding device for automotive sheet metal processing according to claim 7, characterized in that, The base (1) is equipped with a molten pool morphology recognition system, which is used to perform the following steps: S701: Real-time capture of dynamic heat distribution image sequence of the molten pool at the working point of the welding equipment (8) using an infrared thermal imaging sensor; S702: A deep residual network is used to perform semantic segmentation on the image sequence, extract the molten pool boundary contour in real time, and calculate the geometric deviation value of the instantaneous geometric centroid of the molten pool relative to the theoretical weld centerline.

9. An automated welding device for automotive sheet metal processing according to claim 8, characterized in that, The mobile device (7) performs real-time gain control based on the geometric deviation value, including: S703: The geometric deviation value is fed into the long short-term memory network as a time-series input variable to predict the melting depth shift trend in the next sampling period. The predicted value is converted into a compensation signal and fed back to the mobile device (7). By fine-tuning the walking speed of the welding equipment (8), the uneven heating caused by the residual micro-bumps after the cleaning in the aforementioned steps is offset.

Images