Self-adaptive welding device based on machine vision

Abstract

The self-adaptive welding device comprises a robot, a robot controller, an industrial personal computer, a visual assembly and a welding gun, a visual processing module, a welding seam contour recognition module, a welding seam precision adjustment module and a track generation module are arranged in the industrial personal computer, and the visual assembly is used for conducting multi-angle shooting on a workpiece; the visual processing module is used for identifying geometrical characteristics of a main pipe and a branch pipe of the workpiece; the weld contour recognition module is used for recognizing a weld contour; the welding seam precision adjusting module is used for refining the precise position of the welding seam point at the junction of the main pipe and the branch pipe through a calibration method; the track generation module is used for generating a welding track; and the robot controller receives the welding track and controls the welding gun to move according to the welding track to execute welding operation. According to the method, the reasonable robot welding posture is automatically matched, the track continuity and the welding process stability are improved, through automatic track planning and execution, it is ensured that the quality of each welding seam is stable, and manual operation errors are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pipeline weld processing, in particular to a self-adaptive welding device based on machine vision. BACKGROUND

[0002] The saddle-shaped weld of the pipeline is widely used in the fields of oil and gas engineering, chemical equipment manufacturing, pressure vessel and boiler manufacturing, marine engineering, shipbuilding and nuclear power equipment, etc. It mainly appears at the position where the main pipeline and the branch pipeline intersect. Due to the intersection of two cylindrical pipelines in space forming a complex three-dimensional curve, the weld at the junction presents a special geometric feature similar to a saddle-shaped weld (Saddle Weld).

[0003] In various industrial production systems, the connection of the main pipe and the branch pipe is a common and key structure form. For example, the branch connection in the long-distance oil and gas pipeline, the process branch of the pipe network in the chemical plant, the branch of the cooling water system in the nuclear power plant, and the deck drainage and ventilation pipeline system of the ship, etc. all use this interface structure. The existence of the saddle-shaped weld makes the pipeline system more flexible and compact in layout, while meeting the requirements of fluid transportation, structural support and process.

[0004] In the traditional manufacturing process, the preparation of the saddle-shaped weld depends on experienced technicians, who manually draw the weld contour line, cut the groove, and pair weld, which is a tedious process and prone to deviation. With the development of industrial automation and intelligent manufacturing technology, the existing technology adds a laser vision sensor to the front of the welding gun to scan the saddle-shaped weld contour in real time. The system automatically adjusts the robot trajectory, posture and welding parameters (current, speed, etc.) according to the collected contour. The disadvantage of this method is that the posture requirement for scanning is relatively high, and the adaptability of the workpiece is poor. A set of posture can only be used for small-scale adjustment of the specifications and models, and only local feature recognition can be performed, which cannot use overall features to extract more information. On the other hand, the existing technology can also use teaching welding, that is, manual teaching of points, and then automatic welding by the robot. The disadvantage of this method is that the teaching-based method relies heavily on manual teaching, which is low in efficiency and slow in production rhythm.

[0005] Therefore, it is necessary to further research and improve the existing technology to solve the problems of the existing technology. SUMMARY

[0006] In view of the above technical problems, the present application provides a self-adaptive welding device based on machine vision, which aims to realize automatic welding and improve production efficiency.

[0007] The embodiment of the present application provides a machine vision-based adaptive welding device, which comprises a robot, a robot controller, an industrial computer, a visual assembly and a welding torch mounted at the tail end of the robot, the industrial computer is provided with a visual processing module, a weld seam contour identification module, a weld seam precision adjustment module and a trajectory generation module, the visual assembly is mounted at one end of the robot, and the visual assembly is used for multi-angle shooting of a workpiece to obtain surface point cloud data of the workpiece; the visual processing module is used for point cloud fusion of the point cloud data through space registration, three-dimensional point cloud data is formed, and the three-dimensional point cloud data is processed to identify geometric features of a main pipe and a branch pipe of the workpiece; the weld seam contour identification module is used for detecting and extracting weld seam edge data at the intersection of the main pipe and the branch pipe, and identifying a weld seam contour; the weld seam precision adjustment module is used for secondary fitting and optimization of the extracted weld seam edge data, and abnormal points are removed, so that the accurate positions of the weld seam points at the intersection of the main pipe and the branch pipe are refined through a calibration method; the trajectory generation module is used for processing the weld seam edge curve to extract welding points, automatically calculating or matching the best posture of each welding point, and generating a welding trajectory; and the robot controller receives the welding trajectory and controls the welding torch to move according to the welding trajectory to perform a welding operation.

[0008] Optionally, the device further comprises a positioner, which is used for clamping the workpiece and controlling rotation and tilting of the workpiece.

[0009] Optionally, the workpiece is a saddle-shaped workpiece.

[0010] Optionally, the device further comprises a workbench, the workbench is provided with a ground rail, the positioner is located at one side of the workbench, the robot controller is located at the other side of the workbench, and the robot controller controls the robot to slide on the ground rail.

[0011] Optionally, the device further comprises a welding machine, and the welding machine is connected with the robot.

[0012] Optionally, the robot is further provided with a wire feeding mechanism, which is used for feeding welding wire to the robot.

[0013] Optionally, the trajectory generation module is configured to extract welding points from the weld seam edge curve at equal intervals or according to a specific density, each welding point comprises a three-dimensional spatial coordinate, the tangent direction of the weld seam is taken as a reference of the movement direction of the welding torch, the surface normal of the main pipe is taken as an auxiliary, the best posture of each welding point is automatically calculated or matched, the welding points and corresponding postures are sequentially connected, and a welding trajectory is generated.

[0014] Optionally, the weld seam contour identification module is further configured to calculate a theoretical intersection line through a fitting model of the main pipe and the branch pipe, compare the theoretical intersection line with actual point cloud, correct and extract weld seam edge data, and identify a weld seam contour.

[0015] Optionally, the device further comprises a communication module, which is in communication connection with the robot.

[0016] Optionally, the vision system comprises a 3D camera or a laser scanner.

[0017] The technical scheme provided by the embodiment of the present application comprises a robot, a robot controller, an industrial computer, a vision assembly and a welding gun mounted at the end of the robot, the industrial computer is provided with a visual processing module, a weld contour identification module, a weld precision adjustment module and a trajectory generation module, the vision assembly is mounted at one end of the robot, the vision assembly is used for multi-angle shooting of a workpiece, surface point cloud data of the workpiece is acquired, and in combination with an edge recognition algorithm, the pipe saddle-shaped weld can be quickly and efficiently welded, the ability of automatic welding of the pipe saddle-shaped workpiece of multiple specifications and models is realized, and therefore the automatic and efficient production of the saddle-shaped workpiece is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 Fig. 1 is a structural schematic diagram of the adaptive welding device based on machine vision. DETAILED DESCRIPTION

[0019] The technical scheme in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0020] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "connected", "connected" should be understood in a broad sense, for example, can be directly connected, can be indirectly connected through an intermediate medium, and can be connected inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances. In addition, the technical features involved in different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0021] The present application provides a kind of adaptive welding device based on machine vision, please refer to Figure 1As shown, the device comprises a robot 220, a robot controller 210, an industrial computer, a vision assembly 230 and a welding gun 240 mounted at the end of the robot 220, the industrial computer is provided with a vision processing module, a weld contour identification module, a weld precision adjustment module and a trajectory generation module, the vision assembly 230 is mounted at one end of the robot 220, and the vision assembly 230 is used for multi-angle shooting of the workpiece 100 to obtain surface point cloud data thereof; the vision processing module is used for point cloud fusion of the point cloud data through spatial registration to form three-dimensional point cloud data, and the three-dimensional point cloud data is processed to identify the geometric characteristics of the main pipe and the branch pipe of the workpiece 100; the weld contour identification module is used for detecting and extracting weld edge data at the intersection of the main pipe and the branch pipe to identify the weld contour; the weld precision adjustment module is used for secondary fitting and optimization of the extracted weld edge data, and abnormal points are removed to refine the accurate position of the weld points at the intersection of the main pipe and the branch pipe through the calibration method; the trajectory generation module is used for processing the weld edge curve to extract the welding points, automatically calculating or matching the best posture of each welding point, and generating the welding trajectory; the robot controller 10 receives the welding trajectory and controls the welding gun 240 to move according to the welding trajectory to perform the welding operation.

[0022] The device also comprises a workbench 200, the workbench 200 is provided with a ground rail 270, one side of the workbench 200 is provided with a positioner 50, the robot controller 210 is located on the other side of the workbench 200, the robot controller 210 controls the robot 220 to slide on the ground rail 270, the positioner 250 is used for clamping the workpiece 100 and controlling the posture of the workpiece 100 such as rotation and inclination, a bracket 280 for supporting the workpiece 100 is also arranged below the workpiece 100, one end of the workpiece 100 is clamped in the positioner 250, and the other end is placed on the bracket 280, the workpiece 100 can be a saddle-shaped workpiece, and a welding machine 260 is also arranged on the other side of the workbench 200, the welding machine 260 can be placed close to the robot controller 210, and the welding machine 260 is connected with the robot 220.

[0023] Specifically, the device also comprises a base 221, the robot 220 is fixed on the base 221, and the robot 220 slides on the ground rail 270 through the base 221, and a wire feeding mechanism 290 is also arranged on the robot 220 and used for feeding welding wire to the robot 220. In one embodiment of the present application, the vision assembly 230 is a 3D camera, the 3D camera is placed above the workpiece 100 and mounted at the end of the robot 220, and the welding gun 240 is also mounted at the end of the robot 220.

[0024] The application can shoot and generate high-resolution three-dimensional point cloud data through a high-precision 3D camera, which is used for accurate modeling and measurement. The point cloud is a data set composed of a large number of three-dimensional coordinate points, which is used to describe the spatial form of the object surface. The workpiece includes a main pipe and a branch pipe, wherein the main pipe is a pipe that plays a main conveying role in the pipeline system; the branch pipe is an auxiliary pipe that is branched from the main pipe for shunting or branch transmission.

[0025] When the visual component 230 is a 3D camera, the high-precision 3D camera is used to shoot the main pipe and the branch pipe intersection area of the workpiece 100 from multiple angles to obtain comprehensive surface point cloud data. When used, reasonable exposure and collection parameter settings are adopted to ensure that the weld area contour is clear and the data loss caused by reflection or shadow is reduced.

[0026] In order to avoid occlusion and shadow during shooting, multiple cameras can also be used to shoot different parts of the main pipe and the branch pipe intersection area to be welded. When shooting, automatic exposure, automatic white balance and other functions should be used as much as possible to ensure that the brightness, color and other characteristics of each image are relatively consistent. The whole shooting process needs to be carried out under good lighting conditions, and a fill light or a reflector can be used to enhance the lighting.

[0027] The visual component 230 of the application can also be a laser scanner. The laser scanner is used to collect point cloud data and obtain three-dimensional topographic data of the weld area, which is used for subsequent weld extraction and trajectory generation. The laser scanner has higher precision but slower speed, and is suitable for higher requirement scenarios.

[0028] The visual processing module is used to fuse the point cloud data shot from multiple angles through spatial registration to form three-dimensional point cloud data, and process the three-dimensional point cloud data to identify the geometric features of the main pipe and the branch pipe. Point cloud fusion is the process of aligning and merging point cloud data from multiple sources to generate a complete three-dimensional model. The geometric features include plane, cylinder, sphere and other features.

[0029] The application fuses the point cloud data obtained from different angles through spatial registration (such as ICP algorithm) to form unified and complete three-dimensional scene point cloud data. Through filtering, denoising and resampling preprocessing, the overall quality and continuity of the point cloud are further improved, which lays a foundation for subsequent identification and extraction.

[0030] The ICP algorithm (Iterative Closest Point Algorithm) is a commonly used point cloud registration method for spatial registration (alignment) of two point clouds based on iterative closest point matching. The goal of the ICP algorithm is to accurately adjust the alignment of image data and point cloud data by minimizing the distance between the two sets of point clouds, so that they coincide as accurately as possible in space. The ICP algorithm is an algorithm that optimizes the alignment of point cloud data by minimizing the distance between corresponding points in the point cloud, and finds the best transformation parameters (including rotation and translation) through iteration, so that the point cloud data and target data are as aligned as possible.

[0031] In one embodiment of the present application, the method for processing three-dimensional point cloud data to identify the geometric features of the main pipe and the branch pipe comprises:

[0032] The three-dimensional point cloud data is processed by point cloud segmentation to achieve segmentation of the main pipe and the branch pipe. The geometric features of the main pipe and the branch pipe are identified by extracting the geometric features of the segmented point cloud data.

[0033] Point cloud segmentation is the division of point cloud into several regions or objects with specific meanings for further processing. Geometric feature extraction is a technique for extracting geometric properties (such as planes, cylinders, spheres, etc.) from point cloud data.

[0034] The present application identifies the geometric shapes of the main pipe and the branch pipe based on point cloud segmentation and geometric feature extraction techniques (such as the RANSAC algorithm). The fitting results are classified and judged to determine the spatial relationship between the main pipe and the branch pipe, including key parameters such as the direction of the main pipe, the insertion angle and position of the branch pipe. The present application randomly selects some points from a large number of matching points to determine whether the matching results meet the matching of most points. If some matching points are incorrectly paired, the RANSAC algorithm will automatically remove them to ensure that the remaining matching points can provide the most accurate alignment results.

[0035] The RANSAC algorithm (Random Sample Consensus Algorithm) is an iterative method for finding the optimal model fitting result in data containing a large amount of noise.

[0036] In one embodiment of the present application, the method for processing three-dimensional point cloud data to identify the geometric features of the main pipe and the branch pipe further comprises:

[0037] The geometric features of the main pipe and the branch pipe are identified by matching the CAD model with the actual point cloud.

[0038] The weld contour identification module is configured to detect and extract the weld edge data at the intersection of the main pipe and the branch pipe, and identify the weld contour.

[0039] The weld contour identification module is configured to detect and extract the weld edge data at the intersection of the main pipe and the branch pipe, and identify the weld contour.

[0040] The weld precision adjustment module is configured to perform secondary fitting and optimization on the extracted weld edge data, remove abnormal points, improve the continuity and smoothness of the weld edge curve, and refine the accurate position of the weld point at the intersection of the main pipe and the branch pipe through the calibration method.

[0041] The weld precision adjustment module is configured to perform secondary fitting and optimization on the extracted weld edge data, remove abnormal points, improve the continuity and smoothness of the weld edge curve, and refine the accurate position of the weld point at the intersection of the main pipe and the branch pipe through the calibration method.

[0042] The trajectory generation module is configured to process the weld edge curve to extract the welding point, automatically calculate or match the best pose of each welding point, and generate the welding trajectory.

[0043] The trajectory generation module is configured to process the weld edge curve to extract the welding point, automatically calculate or match the best pose of each welding point, and generate the welding trajectory.

[0044] In order to obtain the best pose of the welding point, the trajectory generation module is configured to process the weld edge curve to extract the welding point, automatically calculate or match the best pose of each welding point, and generate the welding trajectory.

[0045] The welding point position and the corresponding posture are sequentially connected to generate a complete and continuous welding trajectory. According to the process requirements, the trajectory is optimized in terms of speed, acceleration, angle smoothing, etc., and necessary arc striking and arc collecting action instructions are inserted. The generated welding trajectory is converted into a motion program format (such as a specific brand robot instruction set output by an offline programming tool) recognizable by the robot. The trajectory program is downloaded to the robot control system through a network or data interface, and the program deployment is completed. The robot automatically performs the full-process welding of the saddle-shaped weld according to the downloaded trajectory program.

[0046] The machine vision-based adaptive welding device also comprises a communication module connected to the robot.

[0047] The present application includes at least one of the following beneficial technical effects:

[0048] Full-process automation: The present application realizes complete automation from point cloud collection to trajectory delivery and robot welding execution, greatly reducing manual intervention and improving production efficiency.

[0049] High-precision recognition and positioning: Based on 3D point cloud fusion and fitting algorithm, the present application realizes high-precision recognition and positioning of the main pipe, branch pipe and weld position, ensuring welding quality.

[0050] Adaptive posture planning: The present application automatically matches reasonable robot welding postures according to the change of weld curve, improving the continuity of trajectory and stability of welding process.

[0051] Adapting to complex weld shape: It can handle complex intersection area welds such as saddle-shaped welds, and is not affected by pipe diameter and angle changes, with wide application range.

[0052] High consistency of welding quality: Through automatic trajectory planning and execution, the quality of each weld is stable, and human error is reduced.

[0053] Shorten the development cycle: The standardized algorithm process can be quickly deployed on different types of pipelines and robot platforms, improving system scalability and reusability.

[0054] Reduce the overall cost: By reducing the welding preparation time and manual labor, the overall manufacturing cost is significantly reduced.

[0055] The above examples are only used to illustrate the technical solutions of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that they can modify the technical solutions described in the foregoing examples, or make equivalent substitutions for part of the technical features; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An adaptive welding device based on machine vision, characterized in that, The device includes a robot, a robot controller, an industrial computer, a vision component, and a welding torch mounted on the robot's end effector. The industrial computer is equipped with a vision processing module, a weld contour recognition module, a weld precision adjustment module, and a trajectory generation module. The vision component is mounted on one end of the robot and is used to capture multi-angle images of the workpiece to obtain its surface point cloud data. The vision processing module is used to fuse the point cloud data through spatial registration to form three-dimensional point cloud data, and to process the three-dimensional point cloud data to identify the geometric features of the main pipe and branch pipes of the workpiece. The weld contour recognition module is used to detect and extract weld edge data in the intersection area of ​​the main pipe and branch pipe, and identify the weld contour; the weld accuracy adjustment module is used to perform secondary fitting and optimization on the extracted weld edge data, eliminate outliers, and refine the precise position of the weld points at the junction of the main pipe and branch pipe through calibration; the trajectory generation module is used to process the weld edge curve to extract the welding points, automatically calculate or match the best posture of each welding point, generate the welding trajectory, and transmit it to the robot controller; the robot controller receives the welding trajectory and controls the welding torch to move according to the welding trajectory to perform the welding operation.

2. The machine vision-based adaptive welding device according to claim 1, characterized in that, The device also includes a positioner for clamping the workpiece and controlling its rotation and tilting.

3. The machine vision-based adaptive welding device according to claim 1, characterized in that, The workpiece is saddle-shaped.

4. The machine vision-based adaptive welding device according to claim 2, characterized in that, The device also includes a workbench with a ground rail. The positioner is located on one side of the workbench, and the robot controller is located on the other side of the workbench. The robot controller controls the robot to slide on the ground rail.

5. The machine vision-based adaptive welding device according to claim 4, characterized in that, The device also includes a welding machine, which is connected to the robot.

6. The machine vision-based adaptive welding device according to claim 4, characterized in that, The robot is also equipped with a wire feeding mechanism for feeding welding wire to the robot.

7. The machine vision-based adaptive welding device according to claim 1, characterized in that, The trajectory generation module is configured to extract welding points at equal intervals or at a specific density from the weld edge curve. Each welding point includes three-dimensional spatial coordinates. The tangent direction of the weld is used as a reference for the direction of the welding gun movement, and the normal of the main pipe surface is used as an aid. The module automatically calculates or matches the optimal posture of each welding point and connects the welding points and their corresponding postures in sequence to generate a welding trajectory.

8. The machine vision-based adaptive welding device according to claim 1, characterized in that, The weld contour recognition module is also configured to calculate the theoretical intersection line through the fitting model of the main pipe and the branch pipe, compare the theoretical intersection line with the actual point cloud, correct and extract the weld edge data, and recognize the weld contour.

9. The machine vision-based adaptive welding device according to claim 1, characterized in that, The device also includes a communication module, which is connected to the robot for communication.

10. The machine vision-based adaptive welding device according to claim 1, characterized in that, The vision components include a 3D camera or a laser scanner.

Images