A multi-layer multi-pass weld seam tracking system and method based on active and passive vision

Abstract

The present application belongs to the technical field of intelligent welding, and discloses a multi-layer multi-pass weld tracking method and system based on active and passive vision. The system of the present application comprises an active and passive vision weld tracking sensor, a welding robot and an industrial computer. The active and passive vision weld tracking sensor comprises an active vision device, a passive vision device and a linear laser emitter. The method of the present application generates an original groove model by detecting groove information with a structured light. Then, the active and passive vision are combined to detect the weld surface shape and the weld corner point position during welding, and a groove model of the current weld is established. The groove model of the current weld is compared with the original groove model to determine whether the weld is offset and measure the offset, so as to determine the position of the next weld and plan and generate a preset welding gun posture, thereby avoiding problems caused by heat loss estimation failure, weld offset and poor welding gun posture.

Description

Technical Field

[0001] This invention belongs to the field of intelligent welding technology, specifically, it relates to a multi-layer, multi-pass weld seam tracking system and method. Background Technology

[0002] Welding technology is one of the most important material manufacturing and processing technologies in modern manufacturing, widely used in almost all industrial manufacturing fields, including machining, oil pipelines, automobile manufacturing, aerospace, construction engineering, and microelectronics. In recent decades, with the rapid development of automatic control theory, computer technology, and artificial intelligence, automated welding methods, especially robotic welding technology, have developed rapidly. Due to their inherent stability and high precision, industrial robots, when applied to welding production, can significantly improve production efficiency while ensuring welding quality, eliminating the need for direct human contact with the welding environment and improving the working environment for welders. Currently, the most commonly used robots in industrial production lines are "teach-and-playback" type robots. These robots have a low level of intelligence, only capable of operating under pre-set modes, and cannot provide real-time feedback on welding paths and process parameters based on actual welding conditions. This leads to a decline in welding quality and may even prevent normal welding from being performed, significantly limiting the application of welding robots.

[0003] Based on whether an auxiliary light source is used, visual sensing can be divided into two categories: active vision, which uses auxiliary lighting such as lasers, and passive vision, which uses arc light and natural light as light sources. Due to the good monochromaticity, strong directionality, and high brightness of lasers, laser sensing, using lasers as an illumination source, has excellent anti-interference capabilities and is widely used in weld seam tracking technology. Currently, laser weld seam tracking technology has mature applications in butt joints and corner joints of medium and thin plates. However, in multi-layer, multi-pass welding, the large bevel depth, the varying number of layers and passes during welding, and the irregular weld seam surface all pose significant challenges to laser recognition. Therefore, various image processing techniques have been used to optimize images for weld seam information recognition. These include digital image processing techniques such as dark channel methods, binarization processes, and linear template matching methods, as well as deep learning methods based on laser stripe edge guidance networks. Active vision, assisted by lasers, can obtain weld seam dimensions and joint information, and this technology is relatively mature, with mature products available. However, because the laser projection position is in front of the welding torch and there is a distance, there is a lead detection error. Passive vision directly uses a camera to detect the condition of the weld. However, the lack of feature points such as laser lights makes image recognition and feature extraction very difficult.

[0004] For multi-layer, multi-pass welding of medium-thick plates, current adaptive weld seam tracking technology uses laser sensors. It calculates the cross-sectional shape of each weld pass by pre-defined welding current, power supply voltage, and wire feed speed using a pre-defined welding power supply (WPS) system. This determines the required number of layers and passes before welding. However, this system has several limitations. First, laser sensors cannot accurately detect the height of each weld pass; they can only approximate the height using welding current, power supply voltage, and wire feed speed. These parameters inevitably deviate from the preset values ​​in actual testing, resulting in inaccurate calculated weld cross-sectional areas. This deviation can also lead to weld pass misalignment due to varying heat input during welding. Addressing the conflict between this misalignment and the preset weld pass parameters is crucial. Finally, even if a self-adjusting welding power supply system can compensate for the error, heat loss during welding via spatter and thermal radiation varies with the environment and cannot be accurately determined. In general, many laser vision weld seam tracking sensors on the market support multi-layer, multi-pass weld seam tracking and detection, and some manufacturers have even customized dedicated multi-layer, multi-pass weld seam tracking systems. However, adaptive robotic welding systems, which require multiple functions such as bevel size detection, automatic welding 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. Summary of the Invention

[0005] This invention provides a multi-layer, multi-pass weld tracking method and system based on active and passive vision. It mainly solves the problem that a single camera cannot accurately detect weld bevel information during multi-layer, multi-pass welding, and aims to avoid problems caused by unpredictable heat loss, weld deviation, and poor welding torch posture.

[0006] This invention combines active and passive vision, which not only fully utilizes the advantages of both and complements each other to obtain more welding information, but also uses visual sensing technology to acquire more welding information, significantly improving the accuracy of welding deviation detection.

[0007] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0008] According to one aspect of the present invention, a multi-layer, multi-pass weld seam tracking system based on active and passive vision is provided, including a welding robot and an industrial control computer; it also includes an active and passive vision weld seam tracking sensor, wherein the active and passive vision weld seam tracking sensor communicates with the welding robot, the industrial control computer, and the welding torch via a network cable;

[0009] The active and passive vision weld seam tracking sensor includes an active vision device, a passive vision device, and a linear laser emitter. Relative to the welding movement direction, the active vision device is positioned behind the welding torch, the linear laser emitter is positioned behind the active vision device, and the passive vision device is positioned behind the linear laser emitter. The linear laser emitter has the same focal length as the active vision device and is angled to the axis of the active vision device, so that the image acquired by the active vision device contains linear structured light capable of image processing. The axis of the welding torch, the lens axis of the active vision device, and the lens axis of the passive vision device are all in the same plane.

[0010] The industrial control computer stores image processing and recognition algorithms, robot control algorithms, bevel models, and welding torch posture planning algorithms.

[0011] Furthermore, the active and passive vision weld seam tracking sensor also includes a package housing; the active vision device and the linear laser emitter are disposed inside the package housing, and the passive vision device is installed outside the package housing.

[0012] Furthermore, a cooling plate is attached to the outside of the encapsulation housing.

[0013] Furthermore, the active vision device and the passive vision device are configured such that the weld bead remains in the middle position of the images captured by the active vision device and the passive vision device.

[0014] Furthermore, a light-reducing filter is provided below the lens of the active vision device, and the light-reducing filter is coaxially mounted with the lens of the active vision device.

[0015] Furthermore, an arc light baffle is provided between the light-reducing filter device and the welding torch.

[0016] Furthermore, the image processing and recognition algorithm is used to obtain the bevel depth and bevel width based on the linear structured light acquired by the active vision device; to obtain the start position and end position of the bevel based on the change in the shape of the structured light acquired by the active vision device; and to obtain the weld surface shape information and weld corner position of the current weld based on the bevel image acquired by the active vision device and the passive vision device.

[0017] The robot control algorithm is used to obtain the robot's motion coordinate information based on the position of each weld bead, and to send instructions to the welding robot for control.

[0018] The bevel model and welding torch posture planning algorithm are used to obtain the cross-sectional area of ​​each weld bead according to preset welding parameters; to establish the original bevel model according to the bevel depth and bevel width; to fit the weld bead surface shape information and weld bead corner position information of the current weld bead to obtain the bevel model of the current weld bead; and to correct the position of the next weld bead according to the bevel model of the current weld bead.

[0019] According to another aspect of the present invention, a method for a multi-layer, multi-pass weld seam tracking system based on active and passive vision is provided, utilizing the aforementioned multi-layer, multi-pass weld seam tracking system based on active and passive vision, comprising the following steps:

[0020] (1) Input the preset welding parameters into the welding machine. The industrial control computer receives the preset welding parameters and stores them in the bevel model and welding gun posture planning algorithm. The industrial control computer obtains the cross-sectional area of ​​each weld bead according to the preset welding parameters through the bevel model and welding gun posture planning algorithm.

[0021] The active vision device acquires the linear structured light emitted by the linear laser emitter and transmits it to the industrial control computer; the industrial control computer obtains the bevel depth and bevel width through the image processing and recognition algorithm, and establishes the original bevel model through the bevel model and welding torch posture planning algorithm;

[0022] The welding robot drives the active and passive vision weld seam tracking sensor to scan the bevel, and the industrial control computer uses image processing and recognition algorithms to identify changes in the shape of the structured light to locate the start and end positions of the bevel.

[0023] (2) The industrial control computer combines the cross-sectional area of ​​each weld bead, the original groove model, the starting position and the ending position of the groove obtained in step (1), and performs layering and sorting of the groove through the groove model and welding gun posture planning algorithm to initially determine the initial planning position of each weld bead.

[0024] (3) Based on the initial planned position of each weld bead obtained in step (2), the industrial control computer robot control algorithm obtains the coordinate information and sends the coordinate information to the welding robot. The welding robot controls the welding gun to weld the first weld bead.

[0025] (4) During the welding process of the current weld bead, the active vision device and the passive vision device acquire the bevel image of the current weld bead and upload it to the industrial control computer; the industrial control computer obtains the surface shape information of the current weld bead and the position of the weld bead corner through the image processing and recognition algorithm.

[0026] (5) The industrial control computer fits the weld surface shape information and weld corner position information obtained in step (4) through the bevel model and welding gun posture planning algorithm to obtain the bevel model of the current weld.

[0027] (6) The industrial control computer compares the bevel model of the current weld obtained in step (5) with the original bevel model obtained in step (1) to detect the deviation between the actual position of the current weld and the initial planned position of the current weld.

[0028] If the deviation does not exceed the threshold, the position of the next weld bead is directly obtained based on the original bevel model.

[0029] If the position deviation exceeds the deviation threshold, the position of the next weld bead is corrected based on the bead model of the current weld bead and the welding gun posture planning algorithm.

[0030] (7) Based on the position of the next weld bead obtained in step (6), the industrial control computer obtains the coordinate information through the robot control algorithm and sends the coordinate information to the welding robot. The welding robot controls the welding gun to weld the next weld bead.

[0031] (8) Repeat steps (4)-(7) until welding is complete.

[0032] Further, in step (4), the weld surface shape is obtained by the image recognition algorithm through the recognition of a linear structured light acquired by the active vision device; the weld corner position includes the corner position between welds and the corner position between the weld and the sidewall; the weld corner position is obtained by the image recognition algorithm through the recognition of the image acquired by the passive vision device.

[0033] Further, in step (5), the correction method is as follows: the bevel model and welding torch posture planning algorithm add the value of the position deviation to the coordinates of the initial planned position of the next weld pass to generate the actual position coordinate information of the next weld pass; the industrial control computer sends motion commands to the welding robot according to the actual position coordinate information of the next weld pass.

[0034] The beneficial effects of this invention are:

[0035] This invention provides a multi-layer, multi-pass weld tracking method based on active and passive vision. First, structured light is used to detect bevel information and generate an original bevel model. Then, during welding, active and passive vision are combined. An active vision device detects the weld surface shape, and a passive vision device identifies the weld corner positions. The two are combined to establish the bevel model of the current weld. The current weld bevel model is compared with the original bevel model to determine if there is any weld offset and to measure the offset. This determines the position of the next weld and allows for the planning and generation of a preset welding torch posture. This avoids problems such as unpredictable heat loss, weld offset, and poor welding torch posture.

[0036] This invention provides a multi-layer, multi-pass weld seam tracking system based on active and passive vision. In multi-layer, multi-pass welding, a combination of active and passive vision devices is used for weld seam planning, weld seam positioning and tracking, and welding torch posture planning. A linear laser emitter is mainly used for bevel morphology recognition, while the active and passive vision devices need to identify the welding start position and weld seam surface morphology features. The images recognized by both devices are processed and reconstructed in 3D to jointly generate the current 3D bevel model, realizing the collaborative work of the welding robot, active and passive vision devices, and welding machine, thereby completing multi-layer, multi-pass adaptive welding. This technology avoids the previous method of approximating the bevel model through calculation. By recognizing and detecting feature points after each weld seam is completed, the accurate bevel model is updated in real time, and the specific position and welding torch posture of the next weld seam are planned based on this. Attached Figure Description

[0037] Figure 1 A schematic diagram of the structure of the multi-layer, multi-pass weld seam tracking system based on active and passive vision provided by the present invention;

[0038] Figure 2 A schematic diagram of the active and passive vision weld seam tracking sensor provided by the present invention;

[0039] Figure 3 A flowchart illustrating the multi-layer, multi-pass weld seam tracking method based on active and passive vision provided by this invention;

[0040] Figure 4 This is an active visual image of a weld-free bevel according to Embodiment 1 of the present invention;

[0041] Figure 5 This is an active vision image of the first weld bead in Embodiment 1 of the present invention;

[0042] Figure 6 This is a passive visual image of the first weld bead in Embodiment 1 of the present invention;

[0043] Figure 7 This is an active vision image of the third weld bead in Embodiment 2 of the present invention;

[0044] Figure 8 This is a passive visual image of the third weld bead in Embodiment 2 of the present invention;

[0045] Figure 9 This is an actual image of the third weld bead in Embodiment 2 of the present invention.

[0046] In the diagram: 1. Welding torch, 2. Active and passive vision weld seam tracking sensor, 3. Welding robot, 4. Industrial computer, 5. Cooling plate, 6. Light reduction filter device, 7. Arc light baffle, 8. Encapsulation shell, 9. Active vision device, 10. Linear laser emitter, 11. Passive vision device. Detailed Implementation

[0047] To provide a clearer understanding of the technical features, objectives, and effects of this invention, the technical solution of this invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0048] Existing weld seam tracking technologies typically place the welding torch 1 at the end of the welding robot 3, with the weld seam tracking sensor mounted in front of the welding torch 1 (relative to the welding movement direction). However, multi-layer, multi-pass welding of V-grooves involves dividing the groove cross-section into several layers, each layer further divided into several passes. The active and passive vision weld seam tracking sensor 2 of this invention is suitable for multi-layer, multi-pass welding. It requires scanning the groove model after each pass, and therefore, is mounted behind the welding torch 1. This allows for simultaneous scanning of the current weld pass and planning of the next pass's path, thereby reducing repetitive work while achieving the purpose of correction and tracking.

[0049] like Figure 1 As shown, the present invention provides a multi-layer, multi-pass weld seam tracking system based on active and passive vision, including an active and passive vision weld seam tracking sensor 2, a welding robot 3, and an industrial control computer 4. The active and passive vision weld seam tracking sensor 2 communicates with the welding torch 1, welding machine, welding robot 3, and industrial control computer via a network cable.

[0050] The industrial computer 4 stores image processing and recognition algorithms, robot control algorithms, bevel models, and welding torch posture planning algorithms.

[0051] The image processing and recognition algorithm is used to receive and process images acquired by the active and passive vision weld seam tracking sensors 2, including processing images acquired by the active vision device 9 and images acquired by the passive vision device 11. The active vision image processing flow generally includes: noise reduction, ROI region extraction, binarization, stripe thinning, and linear fitting. Specifically, this embodiment uses ROI region extraction on the original image, median filtering, Roberts edge detection on the processed image, edge scanning, morphological dilation, and finally Hoough detection. For passive vision image processing, deep learning is used, specifically a convolutional neural network (CNN). A CNN model consisting of two convolutional layers, two pooling layers, and two fully connected layers is built using the collected weld seam images for recognition of passive vision images.

[0052] The robot control algorithm is used to send commands to the welding robot 3 for control. The core of the algorithm consists of matrix operations and the writing of robot trajectory commands. Before welding, the camera image coordinate system and the robot coordinate system are linked by matrix calibration using a hand-eye calibration method. Specifically, Zhang's calibration method and the MATLAB calibration toolbox can be used. When the image processing and recognition algorithm finds the feature points of the weld bead, it transforms their position information in the image coordinate system into three-dimensional coordinates in the robot coordinate system through matrix operations. The robot control algorithm replaces the code of the position information based on the coordinate information in the robot coordinate system and transmits the motion commands to the robot controller via a PLC. The robot controller calculates the joint motion commands based on the predetermined path and target posture, ensuring that the welding robot 3 executes the plan.

[0053] The bevel model and welding torch posture planning algorithm are used to calculate the weld cross-section and plan the weld position based on bevel information. The algorithm can calculate the cross-sectional area of ​​each weld bead based on welding parameters and plan the weld position using bevel information. It can also plan the welding torch posture based on the weld bead position, including placing the welding torch at the angle bisector of the bevel and the horizontal line when the weld bead is near the bevel, and placing the welding torch perpendicular to the horizontal line by default when the weld bead is in other positions within the bevel.

[0054] like Figure 2 As shown, the active and passive vision weld seam tracking sensor 2 includes a cooling plate 5, a light-reducing filter device 6, an arc light baffle 7, a package shell 8, an active vision device 9, a linear laser emitter 10, and a passive vision device 11. The active vision device 9 and the passive vision device 11 each consist of a set of cameras and lenses.

[0055] Relative to the welding movement direction, the active vision device 9 is positioned behind the welding torch 1, and the linear laser emitter 10 is positioned behind the active vision device 9. The linear laser emitter 10 has the same focal length as the active vision device 9 and is set at a certain angle to the axis of the active vision device 9. This ensures that the image acquired by the active vision device 9 contains a clear linear structured light beam, facilitating subsequent image processing. Generally, the angle between their axes is preferably between 10° and 30°, and both have the same focal length to ensure a clear linear structured light beam in the image acquired by the active vision device, facilitating subsequent image processing.

[0056] The active vision device 9 and the linear laser emitter 10 are housed inside the encapsulation shell 8. The shape of the encapsulation shell 8 is not limited, as long as it facilitates its installation and adjustment on the welding robot 3 and the installation and fixation of the internal devices; and it is unaffected by welding spatter and high temperature, thus achieving the effect of protecting the internal devices.

[0057] Relative to the welding movement direction, the passive vision device 11 is positioned behind the linear laser emitter 10 and mounted outside the packaging housing 8. The axis of the welding torch 1, the lens axis of the active vision device 9, and the lens axis of the passive vision device 11 must be in the same plane to ensure that the weld bead is always in the middle position of the images acquired by the active vision device 9 and the passive vision device 11.

[0058] The light-reducing filter device 6 includes a light-reducing film and a filter. The light-reducing filter device 6 is located below the lens of the active vision device 9 and is installed coaxially with the lens.

[0059] The arc light baffle 7 is located between the light reduction filter device 6 and the welding torch 1. The arc light baffle 7 is made of copper plate by cold extrusion and is fixedly connected to the bottom of the encapsulation shell 8 by screws and screw holes. It is used to block a large amount of arc light during the welding process and reduce the impact on image acquisition.

[0060] The cooling plate 5 is installed close to the package housing 8. This cooling plate 5 allows for ventilation and cooling, ensuring rapid heat dissipation and continuous operation of the active vision device 9 and the linear laser emitter 10 inside the package housing 8. The cooling plate 5 is generally installed on the side panel of the package housing 8, which is beneficial for heat dissipation and does not affect the operation of the sensor.

[0061] like Figure 3As shown, the present invention also provides a multi-layer, multi-pass weld tracking method based on active and passive vision. Before welding, an active vision device 9 is used to detect groove information and establish a groove model. On this basis, a passive vision device 11 is added to identify weld beads, determine whether there is horizontal offset of the weld bead, and measure its offset. The method includes: after planning the welding start position, welding the first weld bead, and after the first weld bead is welded, detecting the weld bead shape in real time, updating the groove model, and planning the start position and weld bead path of the next weld bead. This controls the welding torch 1 to weld the next weld bead. This process is repeated until all weld beads are welded, thereby greatly improving the accuracy of weld bead recognition in multi-layer, multi-pass welding.

[0062] A multi-layer, multi-pass weld seam tracking method based on active and passive vision, specifically including the following processes:

[0063] (1) Input the preset welding parameters into the welding machine. The welding parameters are received by the industrial control computer 4 via the network cable and stored in its stored bevel model and welding torch posture planning algorithm. Based on the welding parameters, the industrial control computer 4 can obtain the cross-sectional area of ​​each weld bead through the bevel model and welding torch posture planning algorithm.

[0064] The active vision device 9 acquires the linear structured light emitted by the linear laser emitter 10 and transmits it to the image processing and recognition algorithm in the industrial control computer 4 via a network cable. Since the relative positions of the welding robot 3, welding torch 1, active vision device 9, and passive vision device 11 are fixed, their positions in the coordinate system of the welding robot 3 have been pre-tested. The image processing and recognition algorithm identifies the number of pixels between bevel feature points, then performs unit conversion to obtain the bevel depth and bevel width, thereby establishing the original bevel model through the bevel model and welding torch posture planning algorithm.

[0065] Welding robot 3, driven by active and passive vision weld seam tracking sensor 2, scans the bevel. The linear structured light generated by linear laser emitter 10 forms a V-shape within the bevel, but not elsewhere. The image processing and recognition algorithm in industrial computer 4 identifies the changes in the structured light shape to pinpoint the start and end points of the bevel. The start point of the bevel refers to the transition point from no bevel features to the presence of bevel features, and the end point refers to the transition point from the presence of bevel features to the absence of bevel features.

[0066] (2) The industrial control computer 4 combines the cross-sectional area of ​​each weld bead, the original groove model, the starting position and the ending position of the groove obtained in step (1), and divides the entire groove into layers and channels through the groove model and the welding gun posture planning algorithm to initially determine the initial planning position of each weld bead.

[0067] (3) Based on the initial planned position of each weld bead obtained in step (2), the industrial control computer 4 uses the robot control algorithm to obtain the coordinate information and sends the coordinate information to the welding robot 3. The welding robot 3 controls the welding gun 1 to weld the first weld bead.

[0068] (4) During the welding process of the current weld bead, the active vision device 9 and the passive vision device 11 located behind the welding torch 1 acquire the bevel image of the current weld bead and upload it to the industrial control computer 4. The image processing and recognition algorithm in the industrial control computer 4 obtains the surface shape information of the current weld bead and the position of the weld bead corner. The specific steps are as follows:

[0069] The image recognition algorithm in the industrial computer 4 identifies the bevel image of the current weld bead. The identified features include the surface shape of the weld bead and the position of the weld bead corners. The position of the weld bead corners includes the corner positions between weld beads and the corner positions between the weld bead and the sidewall.

[0070] The weld surface shape is obtained by an image recognition algorithm through the recognition of a linear structured light collected by the active vision device 9.

[0071] However, in image processing using linear structured light, the feature extraction of weld bead corner points is often inaccurate due to the divergence and reflection of the structured light itself. Therefore, image processing and recognition algorithms obtain weld bead corner point location information by recognizing images acquired by passive vision devices. The specific method for recognizing images acquired by passive vision devices is to utilize a neural network model (e.g., a CNN convolutional neural network model). The neural network model is trained using a large number of weld bead images. By feeding the passive vision images into the neural network model, the position of the weld bead corner points in the image can be determined, and the weld bead corner point location information can be obtained through coordinate transformation.

[0072] (5) The industrial control computer 4 fits the surface shape information and corner position information of the weld obtained in step (4) through the bevel model and welding gun posture planning algorithm to obtain the bevel model of the current weld.

[0073] (6) The industrial control computer 4 compares the current weld bevel model obtained in step (5) with the original bevel model obtained in step (1) to detect the deviation between the actual position of the current weld bevel and the initial planned position of the current weld bevel.

[0074] If the deviation does not exceed the threshold, the position of the next weld bead can be directly obtained based on the original bevel model.

[0075] If the position deviation exceeds the deviation threshold, the position of the next weld bead is corrected based on the bevel model of the current weld bead and the welding gun posture planning algorithm. Specifically, the bevel model and welding gun posture planning algorithm add the position deviation value to the coordinates of the initial planned position of the next weld bead to generate the actual position coordinate information of the next weld bead. The industrial control computer 4 sends motion commands to the welding robot 3 based on the actual position coordinate information of the next weld bead, and the position adjustment of the next weld bead is realized when welding again.

[0076] The threshold value can be adjusted in the bevel model and welding torch posture planning algorithm, with the reference of not having fusion defects. The value range is generally 0.3mm to 0.6mm.

[0077] (7) Based on the position of the next weld bead, the industrial control computer 4 obtains the coordinate information through the robot control algorithm and sends the coordinate information to the welding robot 3. The welding robot 3 controls the welding gun 1 to weld the next weld bead.

[0078] (8) Repeat steps (4)-(7) until welding is complete.

[0079] The following describes several preferred embodiments of the present invention to make the technical content clearer and easier to understand, but the scope of protection of the present invention is not limited to the embodiments described herein.

[0080] Example 1

[0081] Welding is performed on the first weld bead of a 60° V-groove Q235 steel plate butt joint. Preset parameters are input: welding current 150A, current voltage 23V, welding speed 5mm / s, welding wire diameter 1.2mm, shielding gas 98% Ar + 2% CO2, gas flow rate 14L / min. Welding robot 3 first scans the groove, identifies the weld bead, generates a groove model, and acquires images as follows: Figure 4 As shown, the welding position is calculated and welding begins. The active camera acquires images as follows. Figure 5 As shown, passively acquired images are as follows: Figure 6 As shown, after welding is completed, the acquired images are processed and identified to plan the next welding operation. The weld width monitoring error is better than ±0.6mm, and the weld height monitoring error is better than ±0.6mm.

[0082] Example 2

[0083] Welding was performed on the third weld bead of a 60° V-groove Q235 steel plate butt joint. Preset parameters were input: welding current 170A, current voltage 22.5V, welding speed 5mm / s, welding wire diameter 1.2mm, shielding gas 98% Ar + 2% CO2, and gas flow rate 14L / min. Welding robot 3 performed the welding after determining the welding position, and the active camera captured images as follows: Figure 7 As shown, passively acquired images are as follows: Figure 8 As shown in the image, the actual weld bead after welding is as follows: Figure 9 As shown. After welding is completed, the acquired images are processed and identified to plan the next welding operation. The weld width monitoring error is better than ±0.6mm, and the weld height monitoring error is better than ±0.6mm.

[0084] This invention is not limited to the embodiments described above. The above description of specific embodiments is intended to illustrate and explain the technical solutions of this invention. The specific embodiments described above are merely illustrative and not restrictive. Without departing from the spirit and scope of the claims, those skilled in the art can make many specific modifications based on the teachings of this invention, and these modifications all fall within the scope of protection of this invention.

Claims

1. A multi-layer, multi-pass weld seam tracking method based on active and passive vision, characterized in that, The system is based on a multi-layer, multi-pass weld seam tracking system using active and passive vision. The system includes a welding robot, an industrial computer, and active and passive vision weld seam tracking sensors. The active and passive vision weld seam tracking sensors communicate with the welding robot, the industrial computer, and the welding torch via a network cable. The active and passive vision weld seam tracking sensor includes an active vision device, a passive vision device, and a linear laser emitter. Relative to the welding movement direction, the active vision device is positioned behind the welding torch, the linear laser emitter is positioned behind the active vision device, and the passive vision device is positioned behind the linear laser emitter. The linear laser emitter has the same focal length as the active vision device and is set at an angle to the axis of the active vision device, so that the image acquired by the active vision device contains linear structured light capable of image processing; the axis of the welding torch, the lens axis of the active vision device, and the lens axis of the passive vision device are in the same plane. The industrial control computer stores image processing and recognition algorithms, robot control algorithms, bevel models, and welding torch posture planning algorithms. The image processing and recognition algorithm is used to obtain the bevel depth and bevel width based on the linear structured light collected by the active vision device. Used to obtain the starting and ending positions of the bevel based on the changes in the shape of the structured light collected by the active vision device; Used to acquire the bevel image of the current weld bead based on the active vision device and the passive vision device, and obtain the weld bead surface shape information and weld bead corner position of the current weld bead; The robot control algorithm is used to obtain the robot's motion coordinate information based on the position of each weld bead, and to send instructions to the welding robot for control. The bevel model and welding torch posture planning algorithm are used to obtain the cross-sectional area of ​​each weld bead based on preset welding parameters. Used to create the original bevel model based on the bevel depth and bevel width; This is used to fit the weld surface shape information and weld corner position information of the current weld to obtain the bevel model of the current weld; and to correct the position of the next weld based on the bevel model of the current weld. The multi-layer, multi-pass weld seam tracking method based on active and passive vision includes the following steps: (1) Input the preset welding parameters into the welding machine. The industrial control computer receives the preset welding parameters and stores them in the bevel model and welding torch posture planning algorithm. The industrial control computer obtains the cross-sectional area of ​​each weld bead according to the preset welding parameters through the bevel model and welding torch posture planning algorithm. The active vision device acquires the linear structured light emitted by the linear laser emitter and transmits it to the industrial control computer; the industrial control computer obtains the bevel depth and bevel width through the image processing and recognition algorithm, and establishes the original bevel model through the bevel model and welding torch posture planning algorithm; The welding robot drives the active and passive vision weld seam tracking sensor to scan the bevel, and the industrial control computer uses image processing and recognition algorithms to identify changes in the shape of the structured light to locate the start and end positions of the bevel. (2) The industrial control computer combines the cross-sectional area of ​​each weld bead, the original groove model, the starting position and the ending position of the groove obtained in step (1), and performs layering and sorting of the groove through the groove model and welding gun posture planning algorithm to initially determine the initial planning position of each weld bead. (3) Based on the initial planned position of each weld bead obtained in step (2), the industrial control computer obtains the coordinate information through the robot control algorithm and sends the coordinate information to the welding robot. The welding robot controls the welding gun to weld the first weld bead. (4) During the welding process of the current weld bead, the active vision device and the passive vision device acquire the bevel image of the current weld bead and upload it to the industrial control computer; the industrial control computer obtains the surface shape information of the current weld bead and the position of the weld bead corner through the image processing and recognition algorithm; (5) The industrial control computer fits the weld surface shape information and weld corner position information obtained in step (4) through the bevel model and welding gun posture planning algorithm to obtain the bevel model of the current weld. (6) The industrial control computer compares the bevel model of the current weld obtained in step (5) with the original bevel model obtained in step (1) to detect the deviation between the actual position of the current weld and the initial planned position of the current weld. If the deviation does not exceed the threshold, the position of the next weld bead is directly obtained based on the original bevel model. If the position deviation exceeds the deviation threshold, the position of the next weld bead is corrected based on the bead model of the current weld bead and the welding gun posture planning algorithm. (7) Based on the position of the next weld bead obtained in step (6), the industrial control computer obtains the coordinate information through the robot control algorithm and sends the coordinate information to the welding robot. The welding robot controls the welding gun to weld the next weld bead. (8) Repeat steps (4)-(7) until welding is complete.

2. The multi-layer, multi-pass weld seam tracking method based on active and passive vision according to claim 1, characterized in that, The active and passive vision weld seam tracking sensor also includes a packaged housing; the active vision device and the linear laser emitter are disposed inside the packaged housing, and the passive vision device is installed outside the packaged housing.

3. The multi-layer, multi-pass weld seam tracking method based on active and passive vision according to claim 2, characterized in that, A cooling plate is attached to the outside of the encapsulation shell.

4. The multi-layer, multi-pass weld seam tracking method based on active and passive vision according to claim 1, characterized in that, The active vision device and the passive vision device are configured such that the weld bead remains in the middle position of the images captured by the active vision device and the passive vision device.

5. The multi-layer, multi-pass weld seam tracking method based on active and passive vision according to claim 1, characterized in that, A light-reducing filter is provided below the lens of the active vision device, and the light-reducing filter is coaxially mounted with the lens of the active vision device.

6. The multi-layer, multi-pass weld seam tracking method based on active and passive vision according to claim 5, characterized in that, An arc light baffle is provided between the light-reducing filter device and the welding torch.

7. The multi-layer, multi-pass weld seam tracking method based on active and passive vision according to claim 1, characterized in that, In step (4), the weld surface shape is obtained by the image processing recognition algorithm through the recognition of the linear structured light acquired by the active vision device; the weld corner position includes the corner position between welds and the corner position between weld and sidewall; the weld corner position is obtained by the image processing recognition algorithm through the recognition of the image acquired by the passive vision device.

8. The multi-layer, multi-pass weld seam tracking method based on active and passive vision according to claim 1, characterized in that, In step (5), the correction method is as follows: the bevel model and welding torch posture planning algorithm add the value of the position deviation to the coordinates of the initial planned position of the next weld to generate the actual position coordinate information of the next weld; the industrial control computer sends motion commands to the welding robot according to the actual position coordinate information of the next weld.

Images