Pipeline all-position multi-layer multi-channel TIG welding bead morphology measuring device and method

Abstract

The invention discloses a pipeline all-position multi-layer multi-channel TIG welding bead morphology measuring device and method, and relates to the technical field of intelligent welding. A molten pool image collecting module is used for collecting morphology information of a molten pool, and the light emitting end of a laser transmitter is used for facing a welding bead; the active and passive vision module is used for collecting an image of light emitted by the laser emitter projected on a weld pool groove and identifying a weld bead corner opening position and a weld bead width, and the active and passive vision module and the weld pool image collecting module are both electrically connected with the control module; the control module can fuse image features collected by the active and passive vision module and the molten pool image collection module, the axes of the molten pool image collection module, the welding module and the active and passive vision module are located in the same plane, and the molten pool image collection module and the active and passive vision module are located on the two sides of the welding module respectively. According to the method, manual participation in the welding process can be reduced, the welding automation degree is improved, and meanwhile, the welding bead morphology measurement accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent welding technology, and in particular to a device and method for measuring the weld bead profile of pipeline all-position multi-layer multi-pass TIG welding. Background Art

[0002] Pipeline welding is a critical process in the industrial sector, particularly in multi-layer, multi-pass welding of high-pressure, thick-walled pipelines, where weld quality directly impacts safety and reliability. Large, thick-walled nuclear power pipelines primarily utilize all-position tungsten inert gas (TIG) welding, typically with a V-groove angle of 15°-50°. During automated welding using an all-position trolley, torch deviation correction and sidewall incomplete fusion defects are key challenges in quality control. Currently, these adjustments rely primarily on manual experience, making this a key issue that must be addressed in achieving full automation of the entire process.

[0003] During multi-pass welding, the shape of each weld bead (such as width, height, and interpass angle) directly affects the position and posture of the welding torch during the next weld. Improper adjustment can result in lack of interpass fusion or poor weld formation, which in turn affects subsequent welds. In particular, improper torch adjustment at the sidewall can lead to lack of fusion defects. Lack of interpass fusion and sidewall fusion defects are also major potential causes of pipeline weld failure.

[0004] With the rapid development of automatic control theory, computer technology, and artificial intelligence, rail-mounted welding trolleys are now the most common type of equipment used for pipeline welding. These trolleys are equipped with arc sensors or vision sensors. Traditional rail-mounted welding trolleys are divided into gas metal arc welding (GMAW) trolleys and tungsten inert gas welding (TIG) trolleys, each representing a different welding process for pipeline welding.

[0005] For the TIG welding process, the arc sensor is mainly used to detect the height direction of the welding gun. It is based on the principle that there is a corresponding relationship between the arc length and the welding voltage. However, it can only reflect the height information of the welding gun and cannot judge the left and right of the welding gun. As for visual sensors, the visual sensors currently used include high-dynamic cameras for photographing the molten pool and structured light cameras for photographing the groove. Among them, the molten pool camera is mainly used to photograph the shape and size of the molten pool. When the welder cannot directly observe the weld with the naked eye, the molten pool camera can be used to assist in observation. Structured light vision is mainly used to photograph the groove shape, that is, the groove shape is reconstructed by observing the reflection state of the laser stripes after photographing the groove.

[0006] In the existing technology, the focus is usually on realizing welding path planning and parameter control through active visual scanning, simplified weld bead models or offline point cloud analysis, but the irregular characteristics of the weld bead cross-section caused by the dynamic behavior of the molten pool are not fully considered. There is a general limitation that the model simplification does not match the actual weld bead morphology. It is not suitable for the recognition of multi-layer and multi-pass welds in pipelines, especially unable to realize the recognition of complex weld bead morphology caused by strong reflection, spatter interference and molten pool flow. The recognition accuracy is insufficient, which can easily lead to side wall unfusion defects. Summary of the Invention

[0007] The purpose of the present invention is to provide a device and method for measuring the weld bead profile of pipeline all-position multi-layer and multi-pass TIG welding to solve the problems existing in the above-mentioned prior art, reduce manual participation in the welding process, improve the degree of welding automation, and at the same time, improve the accuracy of weld bead profile measurement.

[0008] Avoid lack of fusion defects caused by welding gun position deviation.

[0009] To achieve the above object, the present invention provides the following solutions:

[0010] The present invention provides a pipeline all-position multi-layer and multi-pass TIG welding weld bead profile measurement device, comprising a molten pool image acquisition module, a welding module, an active and passive vision module, a laser emitter and a control module. The molten pool image acquisition module is used to collect morphological information of the molten pool, the light output end of the laser emitter is used to be set toward the weld bead, the active and passive vision modules are used to collect images of the light emitted by the laser emitter projected on the molten pool groove, and to identify the weld corner position and weld bead width. The active and passive vision modules and the molten pool image acquisition module are both electrically connected to the control module, and the control module can fuse the image features collected by the active and passive vision modules and the molten pool image acquisition module. The axes of the molten pool image acquisition module, the welding module and the active and passive vision modules are all located in the same plane, and the molten pool image acquisition module and the active and passive vision modules are respectively located on both sides of the welding module.

[0011] Preferably, the molten pool image acquisition module is a molten pool camera, which is installed above the weld of the molten pool through a camera bracket, and the molten pool camera is slidably connected to the camera bracket. After the molten pool camera slides into place on the camera bracket, it can be locked to the camera bracket.

[0012] Preferably, the welding module comprises a welding gun and a wire feeding element, and the wire feeding end of the wire feeding element is arranged close to the welding end of the welding gun.

[0013] Preferably, the axis of the molten pool camera faces the welding end of the welding gun, and the angle between the axis of the molten pool camera and the axis of the welding gun is 60° to 65°.

[0014] Preferably, the active and passive vision modules include an active vision camera and a passive vision camera. The active vision camera is arranged above the weld, and the active vision camera is used to collect the image of the linear structured light emitted by the laser emitter projected on the molten pool groove, and analyze the image information to obtain the geometric dimensions of the weld; the passive vision camera is used to identify the weld corner position and weld width, and the axes of the active vision camera, the passive vision camera, the welding module and the molten pool image acquisition module are all located on the same plane.

[0015] Preferably, it further includes a shell, the active vision camera and the laser emitter are both installed in the shell, and the lower end of the shell is provided with openings corresponding to the positions of the active vision camera and the laser emitter, the laser emitter is located on the side of the active vision camera away from the welding module, and the angle between the axis of the active vision camera and the axis of the laser emitter is 10° to 30°, the passive vision camera is installed on the outer wall of the shell, and the laser emitter is located between the active vision camera and the passive vision camera.

[0016] Preferably, the shell is mounted on a track-type automatic welding robot through a clamp, and the clamp is connected to one side of a bracket, and the molten pool image acquisition module is mounted on the other side of the bracket. The clamp is provided with a long hole at one end for connecting to the shell, and the length direction of the long hole is perpendicular to the axial direction of the welding module. A sliding hole adjustment block is rotatably mounted on the side wall of the shell, and the sliding hole adjustment block is connected to the long hole by bolts, and the installation position of the sliding hole adjustment block on the long hole can be adjusted.

[0017] Preferably, a light reduction filter element is installed in the housing, and the light reduction filter element is coaxially installed below the lens of the active vision camera. An arc baffle is installed at the lower end of the housing, and the arc baffle is located below the light reduction filter element. The arc baffle is located between the light reduction filter element and the welding module, and the light reduction filter element includes a light reduction plate and a filter.

[0018] Preferably, a cooling plate is installed on the outer wall of the shell, and the cooling plate is located between the shell and the welding module. The interior of the cooling plate can be ventilated and cool the inside of the shell.

[0019] The present invention also provides a method for measuring the weld bead profile of a pipeline in all positions, multiple layers, and multiple passes of TIG welding, using the apparatus for measuring the weld bead profile of a pipeline in all positions, multiple layers, and multiple passes of TIG welding as described in any one of the above technical solutions, comprising the following steps:

[0020] S1. Pre-welding preparation: Secure the welding carriage to the surface of the pipe to be welded, ensure that the field of view of the active and passive vision modules and the molten pool image acquisition module covers the weld bead and sidewall area of ​​the molten pool, and adjust the angles of the active and passive vision modules and the molten pool image acquisition module. The control module stores the set channel model and corresponding pipe welding parameters. After preparation is complete, begin welding;

[0021] S2. Align the weld pool image with the active and passive vision modules. After welding starts, the weld pool image acquisition module and the active and passive vision cameras in the active and passive vision modules simultaneously acquire images at the same frequency and spatially align them so that the weld pool images, which are aligned with the active and passive vision modules, correspond to the same points.

[0022] S3. The molten pool image acquisition module identifies the first and last passes or the middle pass of the molten pool. During the welding process of the current weld bead, the molten pool image acquisition module captures the molten pool image and uses the image processing and recognition algorithm in the control module to determine the molten pool contour. It locates the geometric position of the molten pool sidewall through edge detection, calculates the pixel distance between the molten pool edge and the molten pool sidewall, and determines whether the current pass is the middle pass or the first and last pass based on this pixel distance.

[0023] S4. Image recognition: The active vision camera captures and recognizes the image of the linear structured light emitted by the laser emitter projected on the weld bead, and uses the image recognition algorithm in the control module to calculate the geometric parameters of the weld bead. At the same time, the passive vision camera captures the image of the current weld bead and uploads it to the control module. The image processing and recognition algorithm in the control module obtains the weld surface shape information of the current weld bead, as well as the position of the weld bead and the weld corner point. The image captured by the passive vision camera is recognized by the pre-trained neural network model to obtain the weld width information and calibrate the position of the weld corner point in the image. During the welding process of the current weld bead, the molten pool image acquisition module predicts the weld width after solidification and calculates the distance from the weld edge to the groove edge based on the shape and internal flow state of the molten pool, combined with the linear relationship between the molten pool size and the weld bead width. When the passive vision camera is interfered with by external factors and cannot accurately identify the weld bead width, the predicted molten pool width and distance are used instead.

[0024] S5. Image feature fusion: The laser stripe image obtained by the active vision camera and the corner point distribution map obtained by the passive vision camera are superimposed at the same scale and coordinate system. The corner point distribution obtained by the passive vision camera is used to characterize the weld edge and assist in calibrating the weld edge point in the laser stripe image. After fusion, the pixel distance between the weld edge point and the groove edge point is recorded and converted into the actual distance.

[0025] S6. Secondary amplification to identify weld bead angles: After feature fusion of the laser streak image corner distribution image, two weld bead edge points are calibrated on the laser streak image. A secondary amplification strategy is then used to extract the image area near the calibrated edge points as a region of interest (ROI). Image analysis is then performed again. The direction of the laser streaks on either side of the edge point is analyzed, and the angle of the laser streaks in this area is calculated using the slope of the laser streaks on both sides. This angle is then proportionally converted to the actual weld bead angle, achieving accurate characterization of the weld bead geometric features.

[0026] S7. Generate the weld bead morphology at the current position. The depth, width and angle of the groove are obtained through the above steps, and the size of the groove is obtained. The weld height is identified by the active vision camera, and the relative height position of the weld is obtained. The passive vision camera obtains the width of the weld. After fusing the image features, the distance from the edge point of the weld to the edge point of the groove can be recorded to obtain the horizontal position information of the weld bead morphology, and then the position relationship of the weld relative to the groove is obtained. Then, a second zoom is performed to identify the angle between the current weld and the previous weld or groove. Combined with the width and height information of the weld, a more refined weld bead morphology feature is fitted, and then the operation is repeated until the welding is completed.

[0027] Compared with the prior art, the present invention has achieved the following technical effects:

[0028] The present invention provides a pipeline all-position multi-layer multi-pass TIG welding weld bead morphology measurement device and method, which include a molten pool image acquisition module, a welding module, an active and passive vision module, a laser emitter and a control module. The molten pool image acquisition module is used to acquire the morphology information of the molten pool. The light-emitting end of the laser emitter is used to be set toward the weld. The active and passive vision modules are used to acquire the image of the light emitted by the laser emitter projected on the molten pool groove, and to identify the weld corner position and weld width. The active and passive vision modules and the molten pool image acquisition module are both electrically connected to the control module, and the control module can fuse the image features collected by the active and passive vision modules and the molten pool image acquisition module. Through the coordination and intelligent control of the molten pool image acquisition module and the active and passive vision modules, the manual participation in the welding process is reduced, the degree of welding automation is improved, and a secondary amplification strategy is adopted to achieve refined identification of weld bead morphology. Based on the fusion of the active and passive visual features of the module, the area near the edge of the weld is magnified twice, the precise direction of the laser stripes is extracted, the straight line equation is fitted and the angle between the two laser stripes is calculated, and finally it is converted into the weld angle morphology to achieve refined weld morphology recognition. At the same time, combined with the distance between the edge of the molten pool and the side wall of the groove in the molten pool image acquisition module and the weld width predicted by the molten pool image acquisition module and the weld width identified by the passive camera, a comprehensive judgment of the side wall unfusion defect is made and an early warning is issued. It can also effectively improve the accuracy of weld morphology measurement during pipeline welding. The axes of the molten pool image acquisition module, welding module and active and passive vision module are all located in the same plane, and the molten pool image acquisition module and the active and passive vision modules are respectively located on both sides of the welding module, so that the weld is always in the middle position of the images captured by the molten pool image acquisition module and the active and passive vision modules, thereby improving the image acquisition effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0030] Figure 1 Schematic diagram of the structure of the device for measuring the weld bead profile of multi-layer and multi-pass TIG welding of pipelines at all positions in Example 1;

[0031] Figure 2 This is a partial structural diagram of the device for measuring the weld bead profile of multi-layer and multi-pass TIG welding of pipelines at all positions in Example 1;

[0032] Figure 3 This is a diagram showing the angle relationship among the molten pool camera, welding gun, active vision camera, and passive vision camera in Example 1;

[0033] Figure 4 This is a flow chart of the method for measuring the weld bead profile of multi-layer and multi-pass TIG welding of pipelines at all positions in Example 2;

[0034] In the figure: 1-molten pool camera, 2-wire feeding element, 3-industrial computer, 4-track automatic welding robot, 5-active and passive vision modules, 6-welding pipe, 7-fixture, 8-camera bracket, 9-welding gun, 10-slide hole adjustment block, 11-cooling plate, 12-housing, 13-laser transmitter, 14-passive vision camera, 15-active vision camera, 16-light reduction filter element, 17-arc baffle. DETAILED DESCRIPTION

[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0036] The purpose of the present invention is to provide a device and method for measuring the weld bead profile of pipeline all-position multi-layer and multi-pass TIG welding to solve the problems existing in the prior art, reduce manual participation in the welding process, improve the degree of welding automation, and at the same time, improve the accuracy of weld bead profile measurement.

[0037] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] Example 1

[0039] like Figure 1-Figure 3As shown, this embodiment provides a pipeline all-position multi-layer multi-pass TIG welding weld bead morphology measurement device, including a molten pool image acquisition module, a welding module, an active and passive vision module 5, a laser emitter 13 and a control module. The molten pool image acquisition module is used to collect the morphology information of the molten pool. The light-emitting end of the laser emitter 13 is used to be set toward the weld and to emit a linear structured light. The active and passive vision module 5 is used to collect the image of the light emitted by the laser emitter 13 projected on the molten pool groove, and to identify the weld corner position and weld width. The control module is preferably an industrial computer 3, and the industrial computer 3 establishes communication through a network cable. The active and passive vision module 5 and the molten pool image acquisition module are both electrically connected to the control module, and the control module can fuse the image features collected by the active and passive vision module 5 and the molten pool image acquisition module. Through the coordination and intelligent control of the molten pool image acquisition module and the active and passive vision module 5, the manual participation in the welding process is reduced, and the degree of welding automation is improved. In addition, a secondary amplification strategy is adopted to achieve refined identification of weld bead morphology (weld bead morphology refers to the shape of the weld bead during welding). The geometric shape and surface features of the single weld formed are the core parameters for evaluating welding quality). Based on the fusion of the active and passive vision features of the active and passive vision module 5, the area near the edge of the weld is amplified twice, the precise direction of the laser stripes is extracted, the linear equation is fitted, and the angle between the two laser stripes is calculated. Finally, it is converted into the weld angle profile, realizing refined weld profile recognition. At the same time, the distance between the edge of the molten pool and the side wall of the groove in the molten pool image acquisition module and the weld width predicted by the molten pool image acquisition module are compared with the weld width identified by the passive camera. The side wall unfusion defect (i.e., the molten pool and the side wall of the groove are not fully fused) is comprehensively judged and an early warning is issued. This can also effectively improve the accuracy of weld profile measurement during pipeline welding. The axes of the molten pool image acquisition module, the welding module, and the active and passive vision module 5 are all located in the same plane, and the molten pool image acquisition module and the active and passive vision module 5 are respectively located on both sides of the welding module, so that the weld is always in the middle position of the images captured by the molten pool image acquisition module and the active and passive vision module 5, thereby improving the image acquisition effect.

[0040] Specifically, the molten pool image acquisition module is a molten pool camera 1. The molten pool camera 1 is a special visual device installed at the front end of the welding gun 9 of the welding module. It is equipped with a narrow-band filter and a high-speed imaging module. It can monitor the molten pool state during the welding process in real time and capture parameters such as the shape, size and position of the molten pool in time. The molten pool camera 1 is installed above the weld of the molten pool through the camera bracket 8, and can monitor the molten pool state in real time. The molten pool camera 1 is slidably connected to the camera bracket 8, so that the distance between the molten pool camera 1 and the weld can be adjusted. After the molten pool camera 1 slides into place on the camera bracket 8, it can be locked to the camera bracket 8, thereby fixing the position of the molten pool camera 1 to avoid the position displacement of the molten pool camera 1 during welding and affecting the image acquisition effect.

[0041] The welding module includes a welding gun 9 and a wire feeding element 2. The wire feeding end of the wire feeding element 2 is arranged close to the welding end of the welding gun 9. The position setting of the wire feeding element 2 must ensure that it does not interfere with the lens of the molten pool camera 1.

[0042] The axis of the molten pool camera 1 is oriented toward the welding end of the welding gun 9, and the angle between the axis of the molten pool camera 1 and the axis of the welding gun 9 is 60° to 65°. That is, the molten pool camera 1 is placed at a certain angle to the welding direction and far enough away from the arc so that the molten pool camera 1 can clearly observe the entire molten pool. By placing the molten pool camera 1 at the front end of the welding gun 9 (relative to the welding direction), the distance between the edge of the molten pool and the side wall of the groove can be measured in real time, and the weld bead width recognized by the passive vision camera 14 of the active and passive vision module 5 can be compared with the weld bead width predicted by the molten pool to determine whether there is a side wall unfused defect in the molten pool size. When the passive vision camera 14 is blocked by arc light or spatter, the weld bead width predicted by the molten pool camera 1 and the distance from the weld bead to the side wall can be used instead of calculation.

[0043] The active and passive vision module 5 includes an active vision camera 15 and a passive vision camera 14. The active vision camera 15 is set above the weld, and the active vision camera 15 is used to collect the image of the linear structured light emitted by the laser emitter 13 projected on the molten pool groove, and analyze these images with an image processing and recognition algorithm. By combining the feature points to analyze the deformed laser stripe image, the geometric dimensions such as the weld height are calculated according to the triangular geometric relationship. For the active vision camera 15, it uses laser structured light (such as the laser emitter 13) to project onto the weld surface, and obtains the groove depth, width and other geometric dimensions through three-dimensional point cloud modeling. The passive vision camera 14 is used to identify the weld corner position and weld width, which can compensate for the inaccurate extraction of weld corner position features due to structured light divergence when the active vision camera 15 processes images. A neural network model trained with a large number of weld images is used to identify the images collected by the passive vision camera 14, thereby predicting the weld width. At the same time, the passive vision camera 14 relies on ambient light or welding arc light to collect images and uses a pre-trained neural network model (such as CNN) to identify the weld edge and corner position. The axes of the active vision camera 15, passive vision camera 14, welding gun 9, and weld pool camera 1 are all located on the same plane.

[0044] By designing the active and passive vision module 5 as an active vision camera 15 + a passive vision camera 14, a laser emitter 13 is used to project structured light onto the groove surface, and the active vision camera 15 is used to collect the deformed light band image, extract the three-dimensional geometric parameters such as the groove depth and width, and locate the starting point and the end point; the passive vision camera 14 uses a deep learning algorithm to identify the weld corner point position and the weld width, and finally the control module fuses the dual-source data to obtain the weld morphology characteristics; at the same time, on the basis of the active vision camera 15 and the passive vision camera 14, a molten pool camera 1 is added to complement the parameters, and integrated into a composite sensing system. Through multimodal data fusion, all-round monitoring of the welding process can be achieved, which can improve the equipment's weld morphology recognition accuracy and warn of side wall unfusion defects, providing a reliable solution for pipeline welding under complex working conditions.

[0045] This embodiment also includes a housing 12, in which an active vision camera 15 and a laser emitter 13 are both installed. The active vision camera 15 and the laser emitter 13 are both composed of a group of cameras and lenses. In the welding direction, the laser emitter 13 is located behind the active vision camera 15, and the two have the same focal length. Openings are provided at the lower end of the housing 12 corresponding to the positions of the active vision camera 15 and the laser emitter 13. The laser emitter 13 is located on the side of the active vision camera 15 away from the welding module, and the angle between the axis of the active vision camera 15 and the axis of the laser emitter 13 is 10° to 30°. This design enables the active vision camera 15 to collect clearer linear structured light, which is convenient for subsequent processing. The passive vision camera 14 is installed on the outer wall of the housing 12, and the laser emitter 13 is located between the active vision camera 15 and the passive vision camera 14, that is, in the welding direction, the passive vision camera 14 is installed behind the housing 12, and the passive vision camera 14 is also composed of a group of cameras and lenses. The design of the housing 12 not only secures the active vision camera 15 , the laser transmitter 13 and the passive vision camera 14 , but also protects them from high temperatures and welding spatter.

[0046] The shell 12 is installed on the track-type automatic welding robot 4 through the clamp 7, and the clamp 7 is connected to one side of a bracket, and the molten pool image acquisition module is installed on the other side of the bracket. The bracket and the camera bracket 8 can be the same component. The clamp 7 is provided with a long hole at one end for connecting to the shell 12. The length direction of the long hole is perpendicular to the axial direction of the welding module. A sliding hole adjustment block 10 is rotatably installed on the side wall of the shell 12. The sliding hole adjustment block 10 is connected to the long hole by bolts, and the installation position of the sliding hole adjustment block 10 on the long hole can be adjusted, thereby adjusting the angle of the shell 12 and the distance between the shell 12 and the welding gun 9.

[0047] A light-reduction filter element 16 is installed in the housing 12. The light-reduction filter element 16 is coaxially installed below the lens of the active vision camera 15. An arc baffle 17 is installed at the lower end of the housing 12. The arc baffle 17 is made of cold-extruded copper plate. The arc baffle 17 is connected to the housing 12 by screws and is arranged corresponding to the bottom of the light-reduction filter element 16. The arc baffle 17 is located between the light-reduction filter element 16 and the welding module to block a large amount of arc light during welding and reduce the impact on image acquisition. The light-reduction filter element 16 includes a light-reduction plate and a filter.

[0048] A cooling plate 11 is installed on the outer wall of the shell 12. The cooling plate 11 is located between the shell 12 and the welding module. The inside of the cooling plate 11 can be ventilated and cool the inside of the shell 12, ensuring that the active vision camera 15 and laser emitter 13 inside the shell 12 can quickly dissipate heat for continuous use.

[0049] In this embodiment, the angle relationship between the molten pool camera 1, the active vision camera 15, the laser emitter 13, the passive vision camera 14 and the central axis of the welding gun 9 is shown as follows: Figure 3 As shown: the central axis of the molten pool camera 1 is directed toward the bottom A of the welding gun 9, and the angle θ with the axis of the welding gun 9 is 60°-65°; the central axis of the active vision camera 15 is directed toward the center of the welding pipe 6, and is perpendicular to the point B on the surface of the welding pipe 6 to obtain the best weld bead morphology. The central axis of the active vision camera 15 is at an angle α to the axis of the welding gun 9, and the degree of α varies with the radius R of the welding pipe 6, and is generally 25°-35°; like the active vision camera 15, the central axis of the passive vision camera 14 is also directed toward the center of the welding pipe 6, and is perpendicular to the point B on the surface of the welding pipe 6. Perpendicular to point C of the welding pipe 6, the central axis of the passive vision camera 14 and the central axis of the active vision camera 15 form an angle β, which also changes with the change of R and is generally 10°; the laser emitter 13 irradiates point B on the surface of the welding pipe 6 and forms an angle γ with the central axis of the active vision camera 15. In order to ensure that the image captured by the active vision camera 15 has clear linear structured light, which is convenient for subsequent image processing, generally speaking, the central axis of the laser emitter 13 and the central axis of the active vision camera 15 preferably have an angle range of 10°-30°.

[0050] In this embodiment, the image processing and recognition algorithm is stored in the program of the industrial computer 3, and the image processing and recognition algorithm includes image processing of the molten pool camera 1, active visual image processing and passive visual image processing.

[0051] For active visual image processing, the John Canny edge detection algorithm (Canny edge detection algorithm) is generally used to identify the edge position of the laser stripe in the image, and then the grayscale centroid method is used to calculate the center line of the laser stripe. Finally, by analyzing the deformed laser stripe image, the geometric dimensions such as the weld height are calculated based on the triangular geometric relationship. The role of the passive visual image processing algorithm is to identify the weld corner position and then the weld width, which is mainly accomplished by inputting the preprocessed image into the pre-trained convolutional neural network (UNet neural network) model.

[0052] The molten pool image captured by the molten pool camera 1 is subjected to median filtering, grayscale stretching, edge detection, thinning, and curve fitting to obtain a complete molten pool contour. Geometric parameters such as the molten pool area, perimeter, and major axis length are calculated. The groove edge contour is then extracted using edge detection and other algorithms. The distance between the molten pool edge and the sidewall is calculated. Finally, based on this distance, it is determined whether there is a sidewall unfused defect.

[0053] In addition, the industrial computer 3 also stores a molten pool width prediction algorithm, a sidewall unfusion defect determination algorithm, and a first, last, and middle pass determination algorithm.

[0054] Through the above-mentioned design, this embodiment solves the problem of manual assistance in adjusting the offset of the welding gun 9 after each weld is completed, due to the inability of the all-position welding carriage to collect the shape of each weld (shape contour, angle between welds) in the multi-layer and multi-pass welding process during automated all-position tungsten inert gas shielded welding (TIG) of nuclear power large thick-walled pipelines (wall thickness above 12 mm). In addition, for the first weld and the last weld of each layer, the existing weld bead morphology and the side wall morphology can be simultaneously identified to avoid the occurrence of unfusion defects due to the offset position of the welding gun 9.

[0055] Example 2

[0056] like Figure 4 As shown, this embodiment provides a pipeline all-position multi-layer multi-pass TIG welding weld bead morphology measurement method, using the pipeline all-position multi-layer multi-pass TIG welding weld bead morphology measurement device in Example 1, including the following steps:

[0057] S1. Preparation before welding: fix the welding carriage to the surface of the pipe to be welded, ensure that the field of view of the active and passive vision module 5 and the molten pool camera 1 covers the weld bead and side wall area of ​​the molten pool, and adjust the angles of the active and passive vision module 5 and the molten pool camera 1 to meet the requirements. Figure 3 The industrial computer 3 stores the set channel model and the corresponding pipe welding parameters, and starts welding after preparation is completed;

[0058] S2. Align the image between the molten pool and the active and passive vision modules 5. After welding starts, the molten pool camera 1 and the active and passive vision cameras 15 and 14 in the active and passive vision modules 5 collect images at the same time and frequency. Due to the different installation positions and angles of each camera, the images collected by them will be different in space. Therefore, spatial alignment processing is required, such as Figure 3 As shown, with point A, the direct shot of the molten pool camera 1, as time zero, the active vision camera 15 and the passive vision camera 14 also start to collect images. However, they are not actually data from point A. At this time, it is necessary to calculate the distance between arc AB and arc BC, and calculate the time it takes for the active vision camera 15 and the passive vision camera 14 to reach point A based on the angular velocity of the welding gun 9. Then, the arrival time is compared with the camera frame rate to obtain the offset of the number of images, so that the images of the molten pool, active and passive cameras, etc., correspond to the same point.

[0059] S3. Molten pool camera 1 identifies the first and last passes or the middle pass of the molten pool. During the welding process of the current weld bead, molten pool camera 1 captures the molten pool image and uses the image processing and recognition algorithm in industrial computer 3 to determine the molten pool contour. It then uses edge detection to locate the geometric position of the molten pool sidewalls, calculates the pixel distance between the molten pool edge and the molten pool sidewalls, and uses this pixel distance to determine whether the weld is currently in the middle pass or the first and last pass.

[0060] S4. Image recognition: The active vision camera 15 captures and recognizes the image of the linear structured light projected on the weld by the laser emitter 13. The image recognition algorithm in the industrial computer 3 is used to calculate geometric parameters such as the weld height, groove width, groove depth, and groove angle. Simultaneously, the passive vision camera 14 captures an image of the current weld and uploads it to the industrial computer 3. The image processing and recognition algorithm in the industrial computer 3 is used to obtain the weld surface shape information of the current weld and the positions of the weld and its corners. Due to the divergence and reflection of the structured light itself, the feature extraction of the weld corner positions is often inaccurate. Therefore, by recognizing the images captured by the passive vision camera 14 through a pre-trained neural network model (e.g., a UNet neural network model), the width information of the weld bead can be obtained and the position of the weld bead corner point in the image can be calibrated. During the welding process of the current weld bead, the molten pool camera 1 predicts the width of the weld bead after solidification and calculates the distance from the edge of the weld bead to the edge of the groove based on the shape and size of the molten pool and the internal flow state, combined with the linear relationship between the molten pool size and the weld bead width. When the passive vision camera 14 is interfered with by external factors and cannot accurately identify the weld bead width, the predicted width and distance of the molten pool are used instead.

[0061] S5. Image feature fusion: The laser stripe image obtained by the active vision camera 15 and the corner point distribution map obtained by the passive vision camera 14 are superimposed at the same scale and coordinate system. The corner point distribution obtained by the passive vision camera 14 is used to characterize the weld edge and assist in calibrating the weld edge point in the laser stripe image. After fusion, the pixel distance from the weld edge point to the groove edge point is recorded and converted into the actual distance;

[0062] S6. Secondary amplification to identify the weld angle: After feature fusion of the laser stripe image corner point distribution image, two edge points of the weld are calibrated on the laser stripe image. A secondary amplification strategy is used to extract the image area near the calibrated edge point as the region of interest (ROI). Image analysis is performed again. The direction of the laser stripes on both sides of the edge point is analyzed, and the angle of the laser stripes in this area is calculated based on the slope of the laser stripes on both sides. This is then proportionally converted into the actual weld angle to achieve accurate characterization of the weld geometric features. In this step, it should be noted that there is an angle on both sides of the weld, and this angle may be the angle between the current weld and the previous weld (middle weld), the angle between the current weld and the groove, or the first and last welds.

[0063] S7. Generate the weld bead morphology at the current position. The depth, width and angle of the groove are obtained through the above steps, and the size of the groove is obtained. The weld bead height is identified in combination with the active vision camera 15, and the relative height position of the weld bead is obtained. The passive vision camera 14 obtains the width of the weld bead. After fusing the image features, the distance from the recorded weld bead edge point to the groove edge point can obtain the horizontal position information of the weld bead morphology, and then the position relationship of the weld bead relative to the groove is obtained. Then, a second zoom is performed to identify the angle between the current weld bead and the previous weld bead or groove. Combined with the width and height information of the weld bead, a more refined weld bead morphology feature is fitted, and then the operation is repeated until the welding is completed.

[0064] In this embodiment, the molten pool camera 1 can also be used to determine the unfused defects in the molten pool:

[0065] Condition (1): The molten pool-groove distance exceeds the limit: the minimum Euclidean distance between the edge of the molten pool and the side wall of the parent material is greater than the threshold value.

[0066] Condition (2): The predicted width of the molten pool is abnormally reduced: the predicted width of the molten pool in the liquid state is significantly smaller than the neural network recognition width after solidification.

[0067] Of the above two conditions, when both conditions (1) and (2) are met, it is indicated that there is a sidewall unfusion defect in the current molten pool; when only one of the conditions is met, it is indicated that there is a secondary warning of a sidewall unfusion defect in the current molten pool.

[0068] This embodiment achieves the following objectives through multi-sensor collaboration (active vision camera 15 + passive vision camera 14 + melt pool camera 1), secondary amplification strategy and dynamic parameter compensation technology:

[0069] A secondary magnification strategy is used to achieve refined recognition of weld angle and other morphological parameters: based on the fusion of active and passive visual features, secondary magnification is performed on the area near the weld edge point (ROI), the precise direction of the laser stripes is extracted, its straight line equation is fitted, and the angle between the two laser stripes is calculated. Finally, it is converted into the weld angle morphology to achieve refined weld morphology recognition.

[0070] The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A device for measuring the weld bead profile of multi-layer and multi-pass TIG welding of pipelines at all positions, characterized by: It includes a molten pool image acquisition module, a welding module, an active and passive vision module, a laser emitter and a control module. The molten pool image acquisition module is used to collect the morphological information of the molten pool. The light-emitting end of the laser emitter is used to be set toward the weld. The active and passive vision modules are used to collect the image of the light emitted by the laser emitter projected on the molten pool groove, and to identify the weld corner position and weld width. The active and passive vision modules and the molten pool image acquisition module are both electrically connected to the control module, and the control module can fuse the image features collected by the active and passive vision modules and the molten pool image acquisition module. The axes of the molten pool image acquisition module, the welding module and the active and passive vision modules are all located in the same plane, and the molten pool image acquisition module and the active and passive vision modules are respectively located on both sides of the welding module.

2. The device for measuring the weld bead profile of pipeline all-position multi-layer multi-pass TIG welding according to claim 1, characterized in that: The molten pool image acquisition module is a molten pool camera, which is installed above the weld of the molten pool through a camera bracket, and the molten pool camera is slidably connected to the camera bracket. After the molten pool camera slides into place on the camera bracket, it can be locked to the camera bracket.

3. The device for measuring the weld bead profile of pipeline all-position multi-layer multi-pass TIG welding according to claim 2, characterized in that: The welding module includes a welding gun and a wire feeding element, wherein the wire feeding end of the wire feeding element is arranged close to the welding end of the welding gun.

4. The device for measuring the profile of multi-pass, multi-layer TIG welding welds at all positions of a pipeline according to claim 3, characterized in that: The axis of the molten pool camera faces the welding end of the welding gun, and the angle between the axis of the molten pool camera and the axis of the welding gun is 60° to 65°.

5. The device for measuring the weld bead profile of pipeline all-position multi-layer multi-pass TIG welding according to claim 1, characterized in that: The active and passive vision module includes an active vision camera and a passive vision camera. The active vision camera is arranged above the weld, and the active vision camera is used to collect the image of the linear structured light emitted by the laser emitter projected on the molten pool groove, and analyze the image information to obtain the geometric dimensions of the weld; the passive vision camera is used to identify the weld corner position and weld width, and the axes of the active vision camera, the passive vision camera, the welding module and the molten pool image acquisition module are all located on the same plane.

6. The device for measuring the weld bead profile of multi-layer and multi-pass TIG welding of pipelines at all positions according to claim 5 is characterized in that: It also includes a shell, the active vision camera and the laser emitter are both installed in the shell, and the lower end of the shell is provided with openings corresponding to the positions of the active vision camera and the laser emitter, the laser emitter is located on the side of the active vision camera away from the welding module, and the angle between the axis of the active vision camera and the axis of the laser emitter is 10° to 30°, the passive vision camera is installed on the outer wall of the shell, and the laser emitter is located between the active vision camera and the passive vision camera.

7. The device for measuring the weld bead profile of multi-layer and multi-pass TIG welding of pipelines at all positions according to claim 6, characterized in that: The shell is mounted on a track-type automatic welding robot through a clamp, and the clamp is connected to one side of a bracket, and the molten pool image acquisition module is mounted on the other side of the bracket. The clamp is provided with a long hole at one end for connecting to the shell, and the length direction of the long hole is perpendicular to the axial direction of the welding module. A sliding hole adjustment block is rotatably mounted on the side wall of the shell, and the sliding hole adjustment block is connected to the long hole by bolts, and the installation position of the sliding hole adjustment block on the long hole can be adjusted.

8. The device for measuring the weld bead profile of multi-layer and multi-pass TIG welding of pipelines at all positions according to claim 6, characterized in that: A light reduction filter element is installed in the housing, and the light reduction filter element is coaxially installed below the lens of the active vision camera. An arc baffle is installed at the lower end of the housing, and the arc baffle is located below the light reduction filter element. The arc baffle is located between the light reduction filter element and the welding module. The light reduction filter element includes a light reduction plate and a filter.

9. The device for measuring the profile of multi-layer and multi-pass TIG welding welds at all positions of a pipeline according to claim 6, characterized in that: A cooling plate is installed on the outer wall of the shell. The cooling plate is located between the shell and the welding module. The interior of the cooling plate can be ventilated and cool the inside of the shell.

10. A method for measuring the weld bead profile of multi-layer and multi-pass TIG welding of pipelines at all positions, characterized by: The device for measuring the weld bead profile of a pipeline all-position multi-layer multi-pass TIG welding according to any one of claims 1 to 9 comprises the following steps: S1. Pre-welding preparation: Secure the welding carriage to the surface of the pipe to be welded, ensure that the field of view of the active and passive vision modules and the molten pool image acquisition module covers the weld bead and sidewall area of ​​the molten pool, and adjust the angles of the active and passive vision modules and the molten pool image acquisition module. The control module stores the set channel model and corresponding pipe welding parameters. After preparation is complete, begin welding; S2. Align the weld pool image with the active and passive vision modules. After welding starts, the weld pool image acquisition module and the active and passive vision cameras in the active and passive vision modules simultaneously acquire images at the same frequency and spatially align them so that the weld pool images, which are aligned with the active and passive vision modules, correspond to the same points. S3. The molten pool image acquisition module identifies the first and last passes or the middle pass of the molten pool. During the welding process of the current weld bead, the molten pool image acquisition module captures the molten pool image and uses the image processing and recognition algorithm in the control module to determine the molten pool contour. It locates the geometric position of the molten pool sidewall through edge detection, calculates the pixel distance between the molten pool edge and the molten pool sidewall, and determines whether the current pass is the middle pass or the first and last pass based on this pixel distance. S4. Image recognition: The active vision camera captures and recognizes the image of the linear structured light emitted by the laser emitter projected on the weld bead, and uses the image recognition algorithm in the control module to calculate the geometric parameters of the weld bead. At the same time, the passive vision camera captures the image of the current weld bead and uploads it to the control module. The image processing and recognition algorithm in the control module obtains the weld surface shape information of the current weld bead, as well as the position of the weld bead and the weld corner point. The image captured by the passive vision camera is recognized by the pre-trained neural network model to obtain the weld width information and calibrate the position of the weld corner point in the image. During the welding process of the current weld bead, the molten pool image acquisition module predicts the weld width after solidification and calculates the distance from the weld edge to the groove edge based on the shape and internal flow state of the molten pool, combined with the linear relationship between the molten pool size and the weld bead width. When the passive vision camera is interfered with by external factors and cannot accurately identify the weld bead width, the predicted molten pool width and distance are used instead. S5. Image feature fusion: The laser stripe image obtained by the active vision camera and the corner point distribution map obtained by the passive vision camera are superimposed at the same scale and coordinate system. The corner point distribution obtained by the passive vision camera is used to characterize the weld edge and assist in calibrating the weld edge point in the laser stripe image. After fusion, the pixel distance between the weld edge point and the groove edge point is recorded and converted into the actual distance. S6. Secondary amplification to identify weld bead angles: After feature fusion of the laser streak image corner distribution image, two weld bead edge points are calibrated on the laser streak image. A secondary amplification strategy is then used to extract the image area near the calibrated edge points as a region of interest (ROI). Image analysis is then performed again. The direction of the laser streaks on either side of the edge point is analyzed, and the angle of the laser streaks in this area is calculated using the slope of the laser streaks on both sides. This angle is then proportionally converted to the actual weld bead angle, achieving accurate characterization of the weld bead geometric features. S7. Generate the weld bead morphology at the current position. The depth, width and angle of the groove are obtained through the above steps, and the size of the groove is obtained. The weld height is identified by the active vision camera, and the relative height position of the weld is obtained. The passive vision camera obtains the width of the weld. After fusing the image features, the distance from the edge point of the weld to the edge point of the groove can be recorded to obtain the horizontal position information of the weld bead morphology, and then the position relationship of the weld relative to the groove is obtained. Then, a second zoom is performed to identify the angle between the current weld and the previous weld or groove. Combined with the width and height information of the weld, a more refined weld bead morphology feature is fitted, and then the operation is repeated until the welding is completed.

Images