Multi-layer multi-pass welding real-time exhaust channel based on active and passive vision

Abstract

The invention relates to the technical field of welding automation, in particular to a multi-layer multi-pass welding real-time arrangement based on active and passive vision, which comprises an upper computer control system, a welding robot and a welding gun, and further comprises an active vision system and a passive vision system. Whether the technological parameters of the next welding bead need to be corrected or not is evaluated according to a bead pressing algorithm; when correction is needed, the upper computer outputs correction values of the poses of the welding robot and the welding gun, and welding of each welding bead is conducted in sequence till welding of the last welding bead is completed; and when correction is not needed, the next welding bead is directly welded. According to the method, the three-dimensional contour and the two-dimensional image of each welding bead are collected in real time after each welding bead is completed through the rear-mounted active and passive vision system, the technological parameters and the position of the next welding bead are dynamically adjusted based on the three-dimensional contour and the two-dimensional image, and pass-by-pass closed-loop control over the welding process is achieved.

Description

Technical Field

[0001] This invention relates to the field of welding automation technology, and in particular to a real-time arrangement of multi-layer, multi-pass welds based on active and passive vision. Background Technology

[0002] Welding technology, as a fundamental technique for joining metals or other materials, plays a crucial role in the production of various industrial products and structures. Multi-layer, multi-pass welding, as a special welding method, is a key technology in the welding of thick and extremely thick plate structural components. Taking the welding of marine engineering equipment as an example, the welding of thick plates still mainly relies on manual arc welding, submerged arc welding, and semi-automatic flux-cored wire arc welding. Manual welding involves harsh working environments; the welding arc light and fumes inevitably cause damage to the operators. Furthermore, individual workers have limited energy, resulting in low production efficiency and difficulty in controlling precision when welding multi-layer, multi-pass plates on-site. Robotic intelligent multi-layer, multi-pass welding technology can completely avoid these drawbacks.

[0003] Multi-layer, multi-pass welding bead planning models are based on the equal-height and equal-area methods. Each weld bead is pre-simplified into shapes such as triangles, parallelograms, and trapezoids. The equal-height or equal-area method is used to completely plan the welding of the entire bevel for multi-layer, multi-pass welding. Through the derivation of certain mathematical models and formulas, a suitable weld bead layout and corresponding process parameters for each weld bead are obtained, thus enabling robotic multi-layer, multi-pass welding. However, this bead layout logic ignores changes that occur during the welding process, including bevel gap changes caused by thermal deformation, weld bead offsets due to different heat inputs, and loose weld bead layouts caused by environmental factors. Approximating most weld bead cross-sectional areas into a uniform shape only guarantees accurate weld bead layout under stable welding conditions. However, in actual welding, it is impossible to predict when and where adjustments to the weld bead layout and parameters will be needed. Finally, even if each layer of multi-layer, multi-pass welding is planned separately with different welding parameters, it is still impossible to avoid poor welding results due to misalignment of a particular bead. In addition, some researchers have used laser vision sensors for multi-layer, multi-pass weld seam tracking, and some manufacturers have customized dedicated multi-layer, multi-pass weld seam tracking systems. However, adaptive robotic welding systems, which require unified functions such as bevel size detection, automatic weld path routing detection, adaptation to different bevel forms, and real-time communication with industrial robot control systems, have not been fully developed or widely applied in industry. Therefore, it is crucial to achieve real-time acquisition of the forming information of each pass in multi-layer, multi-pass welding and to correct the control process of the next weld pass based on the existing forming-control model. Combining active and passive vision can fully utilize the advantages of visual sensing and complement each other, thereby obtaining more weld seam information. Moreover, the welding information acquired using visual sensing technology can be compared with a preset model, providing real-time feedback on welding error information and employing real-time routing strategies to improve the applicability and accuracy of weld seam layout.

[0004] For research on adaptive pass arrangement technology for multi-layer and multi-pass welding of thick / very thick plates, most current research uses laser sensors and robotic systems to establish expert databases for different welding processes. During welding, the laser sensor scans the weld bead to generate a weld bead model, and the process in the expert database is matched with the current multi-layer and multi-pass weld bead arrangement and process of the entire bevel. Some researchers have changed the matching method in the database by dividing the multi-layer and multi-pass weld bead into root pass, fill pass, and cap pass sections, and then matching the process model in the expert database with them, thus realizing the zoned weld bead arrangement. However, this method still cannot accurately detect the height and width of each weld. Furthermore, the ambient temperature, gas flow rate, welding materials, and base material during welding affect the thermal deformation and morphology in multi-layer, multi-pass welding. If these parameters deviate from the requirements of the actual test environment, it will lead to different heat inputs during welding, causing weld deviation. Even if the error can be compensated by the self-adjusting system of the welding power source, heat will be lost during welding through spatter, thermal radiation, etc., and this proportion will vary with the environment and cannot be accurately known. Therefore, it is necessary to consider how to solve the conflict between deviation and preset weld in real time. Summary of the Invention

[0005] The main objective of this invention is to provide a real-time routing method for multi-layer, multi-pass welding based on active and passive vision, in order to solve the problems raised in related technologies.

[0006] To achieve the above objectives, according to one aspect of the present invention, a multi-layer, multi-pass welding real-time routing system based on active and passive vision is provided, comprising a host computer control system, a welding robot, and a welding torch, and further comprising: An active vision system is wirelessly connected to a host computer control system. It is used to scan the weld bevel before welding, obtain the bevel morphology, build a bevel model based on the bevel morphology, and then generate a complete offline welding process program based on the bevel model. A passive vision system, located behind the active vision system and wirelessly connected to the host computer control system, is used to acquire the shape and morphology of the weld bead after welding. During welding, the active and passive vision systems acquire the weld bead shape after a single weld bead is welded, and then the pressure algorithm is used to evaluate whether the process parameters of the next weld bead need to be corrected. If correction is needed, the host computer outputs correction values ​​for the welding robot and welding torch poses respectively, and each weld bead is welded sequentially until the last weld bead is completed. If no correction is needed, the next weld bead is welded directly.

[0007] Furthermore, the active vision system is located behind the welding torch and includes an active vision camera and a laser emitter.

[0008] Furthermore, the active vision camera employs a laser sensor and is installed within the active vision encapsulation housing.

[0009] Furthermore, the active vision packaging shell is fixed to the rear side of the welding gun via a connecting plate.

[0010] Furthermore, the laser emitter is a line laser emitter, used to acquire the three-dimensional cross-sectional profile of the weld bead.

[0011] Furthermore, the passive vision system includes a passive vision camera, which is fixedly mounted on the outside of the active vision package housing and located behind the active vision camera.

[0012] Furthermore, the field of view of the passive vision camera on the workpiece is a passive vision field of view.

[0013] Furthermore, the field of view of the active vision camera on the workpiece is the field of view of the active vision camera.

[0014] Furthermore, the laser emitted by the line laser emitter is projected onto the workpiece to form a laser line.

[0015] Furthermore, the host computer control system, the welding robot, and the welding torch are included. The welding torch is mounted on the welding robot, which drives the welding torch to perform welding tasks. The welding robot is wirelessly connected to the host computer control system and is used to receive instructions from the host computer control system.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This application achieves closed-loop control of the welding process by acquiring the 3D contour and 2D image of each weld pass in real time after completion using a post-positioned active and passive vision system. Based on this, a pass-by-pass algorithm is used to dynamically adjust the process parameters and position of the next weld pass, thus realizing closed-loop control of the welding process pass by pass. This method can effectively compensate for dynamic deviations caused by thermal deformation during the welding process, prevent defect accumulation, and thus significantly improve the forming accuracy and consistency of multi-layer, multi-pass welding.

[0017] 2. This application provides a more comprehensive and reliable data foundation for groove layout decisions by integrating the precise three-dimensional measurement capabilities of active vision with the two-dimensional texture and offset recognition capabilities of passive vision. This multi-information fusion perception strategy enhances adaptability to different groove shapes and dynamic welding environments, improving the system's intelligence and robustness. Attached Figure Description

[0018] Figure 1 This is a flowchart of the present invention; Figure 2 This is a schematic diagram of the installation location of the present invention; Figure 3These are images captured by the active vision camera of this invention; Figure 4 These are images captured by the passive vision camera of this invention; Figure 5 This is a side view of the welded result of the present invention; Figure 6 This is a top view of the post-welding result of the present invention.

[0019] Figure label: 1. Welding torch; 2. Workpiece; 3. Connecting plate; 4. Active vision packaging shell; 5. Linear laser emitter; 6. Passive vision camera; 7. Active vision camera; 8. Active vision camera field of view; 9. Laser line; 10. Passive vision field of view; 11. Weld seam. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structure, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0021] This embodiment provides a real-time arrangement of multi-layer, multi-pass welds based on active and passive vision, such as... Figure 1 As shown, its core components include a welding execution unit, a sensing unit, and a host computer control system. The welding execution unit consists of a welding robot, such as a Kuka robot, and a welding torch 1, such as an Aotai welding machine. The welding torch 1 is mounted on the end effector of the welding robot.

[0022] In this embodiment, the welding robot can be a Kuka robot, and the welding torch 1 can be an Aotai welding machine. Both the Kuka robot and the Aotai welding machine have stable performance and mature technology, which is conducive to ensuring welding quality.

[0023] The sensing unit is a combination of active and passive vision systems, which is fixedly mounted to the rear of the welding torch 1 via a rigid connecting plate 3. This mounting method ensures that the vision system can move synchronously with the welding torch 1 and that its field of view can cover the weld area behind the welding torch 1 that has just been welded. The active and passive vision systems are mounted on the active vision package housing 4 and are connected to the welding torch 1 through the active vision package housing 4 and the connecting plate 3.

[0024] like Figure 2 As shown, the core sensing elements of the active and passive vision system include a line laser emitter 5, an active vision camera 7, and a passive vision camera 6. The line laser emitter 5 projects a high-brightness linear laser beam onto the bevel surface or the surface of the formed weld bead on the workpiece 2. Figure 2 9. The red laser line in the middle. Figure 2The area within the blue box represents the field of view of the active vision camera 7, i.e., the active vision camera field of view 8. The active vision camera field of view 8 is used to capture the deformed image of the laser line 9 on the surface of the workpiece 2 due to the contour undulation. Figure 2 The area within the orange box represents the field of view of the passive vision camera 6, i.e., the passive vision field of view 10. The passive vision field of view 10 covers a wider area and is used to collect texture and color information of the weld bead and the surrounding base material area.

[0025] Before starting the welding process, system initialization is required. This includes calibrating the welding robot's tool coordinate system and performing hand-eye calibration to determine the spatial transformation relationship between the active and passive vision systems and the welding robot's base coordinate system. Simultaneously, a welding process expert database needs to be pre-built, storing optimal single-pass weld formation models and their welding process parameter ranges corresponding to different bevel geometry parameters.

[0026] The welding process begins in the offline planning phase. The robot is controlled to move, causing the laser line projected by the linear laser emitter 5 to scan the original bevel to be welded. For example... Figure 1 As shown, the active vision camera 7 synchronously acquires a series of laser contour images of the bevel and uploads them to the host computer. The host computer runs image processing algorithms based on the OpenCV library, including image filtering, laser stripe centerline extraction, and 3D point cloud reconstruction, to extract key geometric features of the bevel cross-section from the images, such as bevel angle, depth, and root gap. It also extracts the contour lines of the laser line 9 emitted by the linear laser emitter 5, the active vision camera's field of view 8, and the passive vision field of view 10 to obtain the weld bead cross-sectional contour data, thereby constructing an accurate digital bevel model. Subsequently, the host computer calls an expert database to match and fill the constructed bevel model with the single-pass weld bead forming models stored in the database, generating a complete offline welding program for that specific bevel. This program includes a preliminary weld bead arrangement sequence, robot motion path, and reference welding parameters for each weld bead.

[0027] The host computer control system uses a high-performance industrial computer as the core processor. It sends welding torch position adjustment commands to the welding robot controller through the robot's KUKA, Ethernet, KRL and other Ethernet communication interfaces. It interacts with the welding robot to exchange welding parameters through the Modbus TCP protocol and integrates a vision processing module based on OpenCV to realize real-time closed-loop control of the welding process.

[0028] After the offline program is generated, welding begins. The initial weld bead is welded first. Once a weld bead is completed, the robot moves the welding torch 1 and its rear vision system across that bead. During this movement, the linear laser emitter 5 projects a laser line onto the newly solidified weld surface, and the active vision camera 7 captures images of the laser stripes deformed by the uneven contours of the weld surface; simultaneously, the passive vision camera 6 captures a two-dimensional panoramic image of the weld area. The active and passive vision systems transmit the acquired image data to the host computer in real time.

[0029] The host computer processes and analyzes the received image data. For active vision images, the three-dimensional cross-sectional contour of the weld bead is reconstructed using the aforementioned algorithms, and key dimensions such as the weld bead height and width are accurately quantified and calculated. For passive vision images, edge detection and contour extraction algorithms are used to identify the boundary between the weld bead and the base material, and the horizontal offset of the weld bead centerline relative to the preset ideal path is calculated.

[0030] The host computer acquires weld bead forming data, including weld height, width, and horizontal offset, in real time and compares it with the preset weld bead model corresponding to the current weld layer. If the comparison results show that any one or more parameters exceed the preset tolerance threshold, it determines that the process parameters and / or position of the next weld bead need to be corrected, calculates the adjustment amount required to compensate for the current deviation, and finally outputs a correction command for welding torch 1 to adjust the corresponding welding current and voltage, and issues correction commands to the welding robot for adjusting the lateral offset of welding torch 1 and the welding speed, thereby adjusting the welding posture.

[0031] When welding the next weld pass, the welding robot and welding torch 1 will perform the welding task according to the latest corrected parameters issued by the host computer. After the weld pass is completed, the above-mentioned closed-loop process of "welding-acquisition-evaluation-adjustment" is repeated: the active and passive vision systems collect information again, and the host computer evaluates again and decides whether to optimize the parameters for the next weld pass.

[0032] This cyclical process continues throughout the entire multi-layer, multi-pass welding filling process until the bevel is completely filled. This strategy of feedback and real-time correction per pass constitutes an adaptive control system that ensures the final weld bead formation closely approximates the preset ideal model, effectively responding to various dynamic disturbances during the welding process.

[0033] In one specific embodiment, a multi-layer, multi-pass butt welding experiment was conducted on a 400mm × 300mm × 40mm 60° V-groove S355 sheet metal. A Kuka robot was used as the walking mechanism, and an Aotai welding machine was used for welding. The root pass was welded using the STT process with a welding current set to 180A; the fill pass was welded using a pulse welding process with a current output range set to 160A-220A and a welding speed of 6mm / s. During this process, the original images captured by the active vision camera are shown below. Figure 3 As shown, the original images captured by the passive vision camera are as follows: Figure 4 As shown in the figure. After each weld is completed, weld bead information is collected using the above method, and the parameters for the next weld are dynamically adjusted. This process is repeated until the welding is finished. Post-weld observation of the multi-layer, multi-pass weld formation results is shown in the figure. Figure 5 and Figure 6 As shown in the figure. After welding, non-destructive testing was performed on weld 11 of the sample. The results showed that the multi-layer, multi-pass weld was well formed and without defects, with an evaluation level of Grade 1, which verified the effectiveness and superiority of the method.

[0034] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A multi-layer multi-pass welding real-time lane arrangement based on active and passive vision, comprising a host computer control system, a welding robot and a welding torch (1), characterized in that, Also include: An active vision system wirelessly connected with the host computer control system, used to scan the groove of the weld (11) before welding, obtain the groove profile, and establish a groove model according to the groove profile, and then generate a complete set of off-line welding process program according to the groove model; A passive vision system provided at the rear side of the active vision system and wirelessly connected with the host computer control system, used to collect the forming profile of the weld after welding; During welding, the active vision and the passive vision obtain the weld forming condition after welding of a single weld, and then evaluate whether the process parameters of the next weld need to be corrected according to the weld pressure algorithm; If correction is needed, the host computer outputs correction values for the welding robot and the welding gun pose respectively, and the welding of each weld is performed in turn until the welding of the last weld is completed; If correction is not needed, the next weld is directly welded.

2. The real-time lane assignment system for multi-pass multi-layer welding based on active and passive vision of claim 1, wherein, The active vision system is provided behind the welding gun (1) and includes an active vision camera (7) and a laser emitter.

3. The real-time lane assignment system for multi-pass multi-layer welding based on active and passive vision of claim 2, wherein, The active vision camera (7) is a laser sensor installed in the active vision packaging shell (4).

4. The real-time lane assignment system for multi-pass multi-layer welding based on active and passive vision of claim 3, wherein, The active vision packaging shell (4) is fixedly provided on the rear side of the welding gun (1) through the connecting plate (3).

5. The real-time lane assignment system for multi-pass multi-layer welding based on active and passive vision of claim 2, wherein, The laser emitter is a linear laser emitter (5) for obtaining the three-dimensional cross-sectional profile of the weld.

6. The real-time lane assignment system for multi-pass multi-layer welding based on active and passive vision of claim 3, wherein, The passive vision system includes a passive vision camera (6) fixedly installed outside the active vision packaging shell (4) and located at the rear side of the active vision camera (7).

7. The real-time lane assignment system for multi-pass multi-layered vision-based active-passive welding of claim 6, wherein, The field of view of the passive vision camera (6) on the workpiece (2) is the passive vision field of view (10).

8. The real-time lane assignment system for multi-pass multi-layer welding based on active and passive vision of claim 2, wherein, The field of view of the active vision camera (7) on the workpiece (2) is the active vision camera field of view (8).

9. The real-time lane assignment system for multi-pass multi-layer welding based on active and passive vision of claim 5, wherein, The laser emitted by the linear laser emitter (5) forms a laser line (9) on the workpiece (2).

10. The real-time lane assignment system for multi-pass multi-layered active- passive vision based welding of claim 1, wherein, The host computer control system, the welding robot, and the welding gun (1), the welding gun (1) is provided on the welding robot, the welding robot is used to drive the welding gun (1) to perform the welding task, and the welding robot is wirelessly connected with the host computer control system to receive the instructions of the host computer control system.

Images